US Pat. No. 9,445,531

AIR WASHING FOR OPEN AIR COOLING OF DATA CENTERS

BAIDU USA LLC, Sunnyvale...

1. A data center system, comprising:
a housing to house a plurality of servers operating therein, the servers being arranged in a plurality of electronic racks;
an air supply system to receive contaminated air from an external environment of the housing, the contaminated air containing
a plurality of particles;

a misting and cooling system coupled to the air supply system to receive the contaminated air and to clean and cool the contaminated
air, wherein the misting and cooling system includes

a cooling and cleaning chamber to receive the contaminated air from the air supply system, and
a plurality of nozzles arranged to spray a plurality of liquid droplets into the contaminated air within the cooling and cleaning
chamber, wherein the liquid droplets directly contact the particles of the contaminated air to collect and remove at least
a portion of the particles contained in the contaminated air to generate non-contaminated air; and

an airflow delivery system coupled to the misting and cooling system to generate an airflow from the non-contaminated air,
to cause the airflow to travel through the servers of the electronic racks to exchange heat generated by the servers due to
operations of the servers, and to exhaust the airflow carrying the exchanged heat to the external environment outside of the
housing.

US Pat. No. 9,635,785

LIQUID-ASSISTED BOTTOM AIR COOLING OF ELECTRONIC RACKS IN DATA CENTERS

BAIDU USA LLC, Sunnyvale...

1. A data center system, comprising:
a container to contain a plurality of electronic racks of information technology (IT) components operating therein; and
a cooling unit having one or more cooling modules disposed underneath the electronic racks of the IT components to receive
first liquid from a chiller unit, to exchange heat generated from the IT components using the first liquid, to transform the
first liquid into a second liquid with a higher temperature, and to transmit the second liquid carrying the exchanged heat
back to the chiller unit,

wherein each of the electronic racks comprises
a housing to house one or more of the IT components arranged in a stack,
a first rack aisle formed on a first side of the one or more IT components to direct cool air received from the cooling unit
upwardly, and

a second rack aisle formed on a second side of the one or more IT components to direct hot air to the cooling unit downwardly,
the hot air being transformed from the cool air from the first rack aisle by flowing through an air space between the one
or more IT components.

US Pat. No. 9,848,516

LIQUID-ASSISTED BOTTOM AIR COOLING OF ELECTRONIC RACKS IN DATA CENTERS

BAIDU USA LLC, Sunnyvale...

1. An electronic rack, comprising:
a housing to house one or more IT components arranged in a stack;
a first rack aisle formed on a first side of the one or more IT components to direct cooler air upwardly;
a second rack aisle formed on a second side of the one or more IT components to direct warmer air downwardly, the warmer air
being transformed from the cooler air from the first rack aisle by flowing through an airspace between the one or more IT
components; and

a cooling unit disposed underneath the IT components to receive first liquid from an external chiller system, to exchange
heat carried by the warmer air using the first liquid to generate the cooler air, to transform the first liquid into a second
liquid with a higher temperature, and to transmit the second liquid carrying the exchanged heat back to the external chiller
system.

US Pat. No. 9,881,234

SYSTEMS AND METHODS FOR END-TO-END OBJECT DETECTION

Baidu USA LLC., Sunnyval...

1. A method to detect an object in an image, the method comprising:
receiving an input image;
inputting the input image into a fully convolutional neural network (FCN) without requiring a pre-processing step to convert
the input image into a canonical size, the FCN comprising:

a series of convolution layers and pooling layers that generate feature maps; and
an upsampling layer that provides input to a first set of convolution layers to detect confidence scores and a second set
of convolution layers to detect position;

an output feature map that comprises confidence scores and corresponding positions;
converting the output feature map to bounding boxes, each bounding box being associated with a position in the input image
and a confidence score for containing the object within the bounding box; and

outputting a set of bounding boxes having confidence scores that meet a threshold level that classifies the object as being
contained in each bounding box.

US Pat. No. 10,136,559

LIQUID-ASSISTED AIR COOLING OF ELECTRONIC RACKS WITH MODULAR FAN AND HEAT EXCHANGERS

BAIDU USA LLC, Sunnyvale...

1. An electronic rack, comprising:a stack of a plurality of information technology (IT) components disposed therein to form a frontend and a backend of the electronic rack;
a fan unit detachably mounted on the backend of the electronic rack, the fan unit generating an air flow flowing from the frontend to the backend past through an air space of the stack of the IT components to reduce a temperature of the IT components; and
a heat exchange unit detachably mounted on the backend of the electronic rack to exchange heat generated from the IT components, the heat exchange unit having one or more tubes forming a heat exchange screen, the one or more tubes allow a cool liquid to travel therein to exchange the heat carried by the air flow past through the heat exchange screen,
wherein the fan unit comprises a plurality of fan modules, and the heat exchange unit comprises a plurality of heat exchange modules, each of the heat exchange modules corresponding to one or more of the fan modules,
wherein each of the heat exchange modules is mounted on a heat exchange frame,
wherein each of the fan modules includes a pair of clamps to attach each of the fan modules onto one of the heat exchange modules by clamping onto the heat exchange frame to form a cooling module, the pair of clamps includes a first clamp disposed on a first side of each of the fan modules and a second clamp disposed on a second side of each of the fan modules opposite to the first side, the first clamp extends outwardly to clamp onto a first edge of the heat exchange frame, and the second clamp extends outwardly to clamp onto a second edge of the heat exchange frame opposite to the first edge,
wherein each of the fan modules further comprises one or more springs attached thereon and extended outwardly in an identical direction of the pair of clamps associated with each of the fan modules, and
wherein the cooling module is then mounted onto the backend of the electronic rack.

US Pat. No. 9,968,008

METHOD OF IMPROVING EFFICIENCY OF FAN IN INFORMATION TECHNOLOGY HARDWARE

BAIDU USA LLC, Sunnyvale...

1. An electronic rack, comprising:a plurality of information technology (IT) components arranged in a stack, wherein the electronic rack is one of a plurality of electronic racks of a data center, and wherein each of the IT components comprises a computer server having at least a processor and a memory;
a rack power supply having a voltage converter and a voltage regulator, the rack power supply to receive an input voltage from a power source, wherein the voltage converter is to convert the input voltage into an intermediate voltage, wherein the voltage regulator is to regulate the intermediate voltage to generate a regulated voltage to be supplied to the plurality of the IT components; and
a fan system having one or more fan units coupled to the rack power supply to provide cooling air to reduce an operating temperature of the IT components, wherein each of the fan units is to receive the intermediate voltage without going through the voltage regulator and to provide power to a fan of the fan unit, wherein each fan unit if mounted on a backend of one of the IT components.

US Pat. No. 9,932,038

SPEED CONTROL FOR A FULL STOP OF AN AUTONOMOUS DRIVING VEHICLE

BAIDU USA LLC, Sunnyvale...

1. A computer-implemented method for operating an autonomous vehicle, the method comprising:receiving a stop request to decelerate an autonomous vehicle from a first location and to stop at a second location;
determining, using a processor, a first zone and a second zone within a distance from the first location to the second location;
decelerating the autonomous vehicle based on a first deceleration rate from a current speed at the first location to a predetermined speed at a first intermediate location during the first zone, wherein the first intermediate location is between the first location and the second location; and
decelerating the autonomous vehicle based on a second deceleration rate from the predetermined speed at the first intermediate location to a stop at the second location during the second zone, wherein the first deceleration rate and the second deceleration rate are different.

US Pat. No. 9,947,145

SYSTEM AND METHOD FOR PROVIDING INTER-VEHICLE COMMUNICATIONS AMONGST AUTONOMOUS VEHICLES

BAIDU USA LLC, Sunnyvale...

1. A computer-implemented method for inter-vehicle communications amongst autonomous vehicles, the method comprising:transmitting a vehicle identifier (ID) and vehicle information of a first autonomous vehicle in transit to a cloud server over a network, wherein the cloud server is communicatively coupled to a plurality of autonomous vehicles;
receiving a list of one or more vehicle identifiers (IDs) from the cloud server, the vehicle IDs identifying one or more autonomous vehicles are within a predetermined geographic proximity of the first autonomous vehicle, wherein the one or more autonomous vehicles are identified by the cloud server using a nearest neighbor search algorithm; and
directly communicating with a second autonomous vehicle in transit that is selected from the list of one or more autonomous vehicles via a wireless local area network to exchange an operational status of the first autonomous vehicle with the second autonomous vehicle.

US Pat. No. 9,874,871

METHOD TO DYNAMICALLY ADJUSTING STEERING RATES OF AUTONOMOUS VEHICLES

BAIDU USA LLC, Sunnyvale...

17. A data processing system, comprising:
a processor; and
a memory coupled to the processor to store instructions, which when executed by the processor, cause the processor to perform
operations, the operations including determining a plurality of steering rate candidates for a steering control command of
operating an autonomous vehicle;

for each of the steering rate candidates,
calculating a plurality of individual costs for the steering rate candidate by applying a plurality of cost functions, each
cost function corresponding to one of a plurality of cost categories, and

calculating a total cost for the steering rate candidate based on the individual costs produced by the cost functions;
selecting one of the steering rate candidates having a lowest total cost as a target steering rate; and
generating a steering control command based on the selected steering rate candidate to control a steering wheel of the autonomous
vehicle.

US Pat. No. 10,031,526

VISION-BASED DRIVING SCENARIO GENERATOR FOR AUTONOMOUS DRIVING SIMULATION

BAIDU USA LLC, Sunnyvale...

1. A computer-implemented method for generating scenario data for a driving simulation, comprising:receiving image data captured by a camera of a vehicle during a live driving scenario;
identifying an object within a sequence of images of the image data;
determining movement characteristics of the identified object relative to the vehicle based on a size of the object within the sequence of images;
determining movement characteristics of the identified object relative to a base map based on location information received from a global positioning system (GPS) of the vehicle, wherein the location information is associated with one or more of the sequence of images used to determine the movement characteristics of the identified object relative to the vehicle; and
providing driving scenario data representing at least the movement characteristics of the identified object within the base map during the sequence of images.

US Pat. No. 10,118,639

METHOD AND SYSTEM FOR STEERING CONTROL OF AN AUTONOMOUS VEHICLE USING PROPORTIONAL, INTEGRAL, AND DERIVATIVE (PID) CONTROLLER

BAIDU USA LLC, Sunnyvale...

1. A computer-implemented method for steering controlling of an autonomous vehicle, the method comprising:calculating a first steering angle based on a target directional angle and an actual directional angle of the autonomous vehicle to maneuver an obstacle course;
calculating a second steering angle based on a target lateral position and an actual lateral position of the autonomous vehicle to maneuver the obstacle course;
determining a next target steering angle based on the first and second steering angles; and
controlling a subsequent steering angle of the autonomous vehicle based on the next target steering angle.

US Pat. No. 9,956,876

SYSTEM AND METHOD FOR PROVIDING CONTENT IN AUTONOMOUS VEHICLES BASED ON REAL-TIME TRAFFIC INFORMATION

BAIDU USA LLC, Sunnyvale...

1. A computer-implemented method for content delivery in an autonomous vehicle, the method comprising:obtaining location and route information of a route associated with the autonomous vehicle;
obtaining real-time traffic information of the route based on the location and route information;
obtaining first point of interest (POI) information of the route based on the location and route information;
determining a driving mode of the autonomous vehicle based on the location and route information and the real-time traffic information;
selecting a content item from a list of content items obtained from a content database based on the determined driving mode and the first POI information, including extracting a first set of features from the location and route information,
extracting a second set of features from the real-time traffic information,
extracting a third set of features from the first POI information, and
applying a content selection model to the first set, the second set, and the third set of features, generating an output for content selection; and
displaying the selected content item on a display device within the autonomous vehicle.

US Pat. No. 9,965,705

SYSTEMS AND METHODS FOR ATTENTION-BASED CONFIGURABLE CONVOLUTIONAL NEURAL NETWORKS (ABC-CNN) FOR VISUAL QUESTION ANSWERING

Baidu USA LLC, Sunnyvale...

1. A computer-implemented method of improving accuracy in generating an answer to a question input related to an image input, the method comprising:receiving an image input;
receiving a question input related to the image input;
inputting the question input and the image input into an Attention-Based Configurable Convolutional Neural Networks (ABC-CNN) framework to generate an answer, the ABC-CNN framework comprising:
an image feature map extraction component comprising a CNN that extracts an image feature map from the image input;
a semantic question embedding component that obtains question embeddings from the question input;
a question-guided attention map generation component that receives the image feature map and the question embeddings and that obtains a question-guided attention map focusing on a region or regions asked by question input; and
an answer generation component that obtains an attention weighted image feature map by weighting image feature map using the question-guided attention map and generates answers based on a fusion of the image feature map, the question embeddings, and the attention weighted image feature map.

US Pat. No. 10,053,091

SPRING SYSTEM-BASED CHANGE LANE APPROACH FOR AUTONOMOUS VEHICLES

BAIDU USA LLC, Sunnyvale...

1. A computer-implemented method for operating an autonomous vehicle, the method comprising:in response to a request for changing lane, perceiving one or more objects surrounding an autonomous vehicle;
for each of the perceived objects, assigning, by a spring analyzer executed by a processor, a virtual spring connecting the object and the autonomous vehicle, wherein each virtual spring is associated with a specific spring model to generate a force based on relative positions of an associated object and the autonomous vehicle;
aggregating, by a lane analyzer executed by the processor, one or more forces generated from one or more virtual springs corresponding to the one or more surrounding objects to generate an aggregated force;
determining one or more lane-changing parameters for the autonomous vehicle based on the aggregated force and a direction of the aggregated force, including determining a turning angle of changing lane based on the direction of the aggregated force; and
controlling by a control module the autonomous vehicle to change lane from a first lane to a second lane based on the one or more lane-changing parameters.

US Pat. No. 10,133,698

HOT-PLUG CAPABLE INPUT AND OUTPUT (IO) SUBSYSTEM

BAIDU USA LLC, Sunnyvale...

1. An input and output (IO) subsystem chassis, comprising:a plurality of IO modules;
a housing having a frontend and a backend, the housing including a plurality of IO slots to receive the plurality of IO modules inserted from the frontend of the housing; and
a baseboard disposed within the housing, the baseboard including a plurality of first connectors corresponding to the IO slots to receive and connect the plurality of IO modules, wherein each of the IO modules can be coupled to one of a plurality of servers via the backend panel using a cable,
wherein each of the IO modules comprises
an IO card having a peripheral device mounted thereon,
a card holder including
a printed circuit board (PCB),
a first receiving socket disposed perpendicularly on a surface of the PCB to receive and hold the IO card plugged in vertically and downwardly with respect to the surface of the PCB and the baseboard,
a second connector disposed on a backend edge of the PCB parallel with the surface of the PCB to engage with or disengage from a corresponding one of the first connectors of the baseboard horizontally with respect to the surface of the PCB and the baseboard, when the IO module is inserted into or removed from a corresponding IO slot from the frontend, without having to remove the housing, and
wherein the IO subsystem chassis and a server chassis housing the plurality of servers are disposed within an electronic rack.

US Pat. No. 10,049,328

GROUP DRIVING STYLE LEARNING FRAMEWORK FOR AUTONOMOUS VEHICLES

BAIDU USA LLC, Sunnyvale...

1. A computer-implemented method for operating an autonomous vehicle, the method comprising:perceiving driving behaviors of one or more vehicles surrounding a first autonomous vehicle;
transmitting one or more driving style elements representing the driving behaviors of the surrounding vehicles from the first autonomous vehicle to a remote server over a network;
receiving from the remote server a driving style that was determined based on the driving style elements at the remote server, the driving style including information describing how the first autonomous vehicle should drive in view of the surrounding vehicles at a point in time;
generating planning and control data based on the driving style; and
controlling and driving the first autonomous vehicle based on the planning and control data.

US Pat. No. 10,114,374

EMERGENCY HANDLING SYSTEM FOR AN AUTONOMOUS DRIVING VEHICLE (ADV)

Baidu USA LLC, Sunnyvale...

1. A computer-implemented method, comprising:providing autonomous driving control for a vehicle using a primary decision system, wherein the primary decision system uses a first set of driving rules to provide the autonomous driving control;
providing an emergency decision system operating standby as a backup system to the primary decision system, wherein the emergency decision system uses a second set of driving rules to control the vehicle, wherein the second set of driving rules is different than the first set of driving rules;
recognizing an occurrence of an emergency handling event associated with the vehicle; and
in response to recognizing the occurrence of the emergency handling event, switching from the primary decision system to the emergency decision system to continue providing the autonomous driving control for the vehicle, without using the primary decision system, wherein the emergency decision system prioritizes a second set of driving rules over the first set of driving rules to provide the autonomous driving control.

US Pat. No. 10,015,537

SYSTEM AND METHOD FOR PROVIDING CONTENT IN AUTONOMOUS VEHICLES BASED ON PERCEPTION DYNAMICALLY DETERMINED AT REAL-TIME

BAIDU USA LLC, Sunnyvale...

1. A computer-implemented method for content delivery to an autonomous vehicle, the method comprising:performing an image analysis on an image captured using a camera mounted on an autonomous vehicle, the image representing an exterior environment of an autonomous vehicle;
obtaining localization information surrounding the autonomous vehicle at a point in time;
determining a perception associated with an audience external to the autonomous vehicle based on the image analysis and the localization information;
receiving a plurality of content items from one or more content servers over a network in response to the perception associated with the audience;
ranking the plurality of content items using a predetermined ranking algorithm to select a first content item and a second content item from the plurality of content items;
displaying the first content item selected from the one or more content items on a first display device mounted on an exterior surface of the autonomous vehicle;
superimposing the second content item onto the image captured by the camera, generating an augmented image augmented with the second content item; and
displaying the augmented image on a second display device mounted on an interior surface of the autonomous vehicle, wherein the second display device includes a touch screen to allow a passenger to interact with one or more choices presented by the second content item, wherein the second display device is configured to appear as an ordinary window of the autonomous vehicle, and wherein when the passenger looks at the second display device, an object of the augmented image is displayed as a virtual object in a virtual reality manner as if the passenger looked at a physical object through the ordinary window of the autonomous vehicle.

US Pat. No. 10,019,008

SIDESLIP COMPENSATED CONTROL METHOD FOR AUTONOMOUS VEHICLES

BAIDU USA LLC, Sunnyvale...

1. A computer-implemented method for operating an autonomous vehicle, the method comprising:defining a plurality of driving scenarios for a plurality of types of autonomous vehicles, each driving scenario corresponding to a specific movement of a particular type of autonomous vehicles;
for each of the driving scenarios of each type of autonomous vehicles, obtaining driving statistics of each autonomous vehicle, including
recording one or more driving parameters used to control and drive each autonomous vehicle of the type,
capturing a driving condition corresponding the driving scenario, and
calculating an average sideslip caused by the driving parameters and the driving condition based on sideslips of a number of autonomous vehicles of the type under identical driving conditions; and
storing a driving scenario to a sideslip (scenario/sideslip) mapping table having a plurality of mapping entries to map the driving scenario to the calculated average sideslip, wherein the scenario/sideslip mapping table is utilized to compensate a sideslip under a similar driving scenario, the method further including
determining a driving scenario based on a driving condition of an autonomous vehicle;
identifying an entry matching the driving scenario from the scenario/sideslip mapping table;
obtaining an average sideslip from the matching entry; and
adjusting a steering angle of the autonomous vehicle using the obtained average sideslip.

US Pat. No. 10,007,264

AUTONOMOUS VEHICLE HUMAN DRIVER TAKEOVER MECHANISM USING ELECTRODES

BAIDU USA LLC, Sunnyvale...

1. A computer-implemented method for operating an autonomous vehicle, the method comprising:operating an autonomous vehicle in an autonomous mode, wherein during the autonomous mode, the autonomous vehicle operates without influence of a human driver riding in the autonomous vehicle;
detecting an electrical current flowing from a first electrode to a second electrode mounted on a steering wheel of the autonomous vehicle, wherein the first electrode is disposed on a first half of the steering wheel and significantly over a surface of the first half of the steering wheel, and wherein the second electrode is disposed on a second half of the steering wheel and significantly over a surface of the second half of the steering wheel;
detecting that the electrical current has been flowing from the first electrode to the second electrode for a predetermined of time;
in response to the detecting the electrical current flowing from the first electrode to the second electrode for the predetermined period of time, transitioning the autonomous vehicle from the autonomous mode to a manual mode; and
operating the autonomous vehicle in the manual mode in response to user actions from the human driver.

US Pat. No. 10,120,944

SYSTEM AND METHOD FOR ROUTING DATA AND CONNECTING USERS BASED ON USER INTERACTIONS WITH A MACHINE-READABLE CODE OF CONTENT DATA

BAIDU USA LLC, Sunnyvale...

1. A computer-implemented method for routing data and connecting users based on user interactions with content, the method comprising:in response to a first request for query received at a server from a user device over a network, the query including one or more keywords, generating by a search engine hosted by the server a search identifier (ID) identifying a search transaction corresponding to the query;
extracting the one or more keywords from the query;
performing, by the search engine, a search within a content database based on the extracted one or more keywords, generating a set of one or more content items;
for at least one of the content items,
encoding, by an encoder, the search ID and a content ID identifying the content item into a machine-readable code using a predetermined encoding algorithm, and
attaching the machine-readable code to the corresponding content item, the machine-readable code having the search ID and the corresponding content ID encoded therein;
generating a search result page by incorporating the content items therein, at least one of the content items in the search result page including a corresponding machine-readable code attached therein;
transmitting the search result page to the user device over the network such that the search result page is displayed on the user device, wherein detailed information of the at least one of the content items and the corresponding machine-readable code from the search result page are displayed on the user device when a user of the user device clicks on the at least one of the content items, the corresponding machine-readable code being scannable from the user device;
receiving at the server a second request from an agent device of an agent over the network, the second request including identifying information identifying a first machine-readable code associated with a first content item of the content items in the search result page, wherein the first machine-readable code is transmitted from the user device to the agent device over a communication path;
in response to the second request,
extracting by an interaction logging module the search ID and a first content ID from the identifying information;
determining a user ID identifying the user of the user device who initiated the first request based on the search ID and the first content ID; and
compiling by a user profile compiler user information of the user based on information obtained from a user information database based on the user ID.

US Pat. No. 10,054,945

METHOD FOR DETERMINING COMMAND DELAYS OF AUTONOMOUS VEHICLES

BAIDU USA LLC, Sunnyvale...

1. A computer-implemented method for determining command delays of driving commands operating an autonomous vehicle, the method comprising:receiving, by a data collector executed by a processor, driving statistics of an autonomous vehicle, the driving statistics including driving commands issued and responses from the autonomous vehicle at different points in time;
determining, by a command delay determination module executed by the processor, a plurality of command delay candidates for the autonomous vehicle, each of the command delay candidates represented by one or more command cycles associated with the autonomous vehicle;
for each of the command delay candidates, calculating, by the command delay determination module, a percentage of driving commands that resulted in a response of the autonomous vehicle conforming to the driving commands based on the driving statistics associated with the autonomous vehicle; and
selecting, by the command delay determination module, one of the command delay candidates having the highest percentage of conformity as the command delay for the autonomous vehicle, wherein the command delay is utilized to plan and control subsequent operations of the autonomous vehicle.

US Pat. No. 10,139,818

VISUAL COMMUNICATION SYSTEM FOR AUTONOMOUS DRIVING VEHICLES (ADV)

BAIDU USA LLC, Sunnyvale...

1. A computer-implemented method of providing autonomous driving control for a first autonomous vehicle, comprising:determining an initial driving maneuver for the first autonomous driving vehicle in response to the first autonomous driving vehicle perceiving a second vehicle;
initiating a first visual communication signal provided by the first autonomous driving vehicle, wherein the first visual communication signal represents the initial driving maneuver;
detecting a second visual communication signal that is provided by the second vehicle in response to the first visual communication signal, including emitting light according to a predetermined lighting pattern using a lighting mechanism mounted on the second vehicle in response to the first visual communication signal, the predetermined lighting pattern representing the second visual communication signal;
identifying a type of communication represented by the second visual communication signal based on the predetermined lighting pattern; and
determining a subsequent driving maneuver for the first autonomous driving vehicle based on the identified type of communication represented by the second visual communication signal.

US Pat. No. 10,137,896

METHOD AND SYSTEM FOR OPERATING AUTONOMOUS DRIVING VEHICLES USING GRAPH-BASED LANE CHANGE GUIDE

BAIDU USA LLC, Sunnyvale...

1. A computer-implemented method of operating an autonomous vehicle to change lanes from a source lane to a target lane, the method comprising:selecting a first reference node of a topological graph associated with the source lane, the first reference node being a node in the source lane at which the autonomous vehicle can change lanes to a second reference node in the target lane;
determining, with respect to the first reference node of the topological graph, an earliest node in the source lane at which it is first possible for the autonomous vehicle to change lanes from the source lane to the target lane;
determining, with respect to the first reference node of the topological graph, a last node in the source lane after which it is no longer possible for the autonomous vehicle to change lanes from the source lane to the target lane;
determining a range of the source lane for which the autonomous vehicle can change lanes from the source lane to the target lane, wherein the range is from the earliest node in the source lane to the last node in the source lane; and
generating a trajectory to autonomously control the autonomous vehicle to change lanes from the source lane the target lane within the range, including determining time, a speed, and a distance to change lanes.

US Pat. No. 10,162,792

METHOD AND SYSTEM FOR HIGH PRECISION TIME SYNCHRONIZATION

BAIDU USA LLC, Sunnyvale...

1. A data processing system, comprising:a processor;
a Peripheral Component Interface Express (PCIe) link; and
a time processing device coupled to the processor via the PCIe link, the time processing device including time processing logic and a time register, wherein the time processing logic is to receive time data representing date and time from a time data source, and store the time data in the time register,
wherein the time processing device comprises
a first input to receive a first signal from the time data source, the first signal indicating that an update of the time data is available from the time data source, and
a second input to receive a message having the updated time data from the time data source,
wherein the time processing logic is to
detect the first signal received from the first input,
in response to the first signal, decode the time data received from the second input, and
store the decoded time data in the time register, and
wherein the processor retrieves the time data from the time register via a single read instruction over the PCIe link and updates a system clock of the data processing system based on the time data.

US Pat. No. 10,162,396

METHOD AND SYSTEM FOR REMOVING HEAT USING HEAT REMOVAL LIQUID BASED ON WORKLOAD OF SERVER COMPONENTS OF ELECTRONIC RACKS

BAIDU USA LLC, Sunnyvale...

1. An electronic rack, comprising:a heat removal liquid manifold to provide heat removal liquid;
a plurality of server blades contained in a plurality of server slots respectively, each server blade including an information technology (IT) component representing a server therein, wherein each of the server blades is coupled to the heat removal liquid manifold to receive cooler liquid from the heat removal liquid manifold to remove heat generated by the IT component and to transmit warmer liquid carrying the heat exchanged from the IT component back to the heat removal liquid manifold;
a workload calculation module coupled to the server blades to receive workload information from the server blades, wherein the workload calculation module is configured to calculate a total workload of the sever blades;
a coolant distribution unit (CDU) coupled to the heat removal liquid manifold and the server blades, wherein the CDU is to control a liquid flow rate of the heat removal liquid supplied to the heat removal liquid manifold, wherein the CDU includes
a heat exchanger to exchange heat carried by the warmer liquid with an external cooler liquid supplied by an external heat removal system,
a liquid pump to pump the external cooler liquid into the heat removal liquid manifold, and
a pump controller coupled to control the liquid pump to control the liquid flow rate; and
a rack management controller (RMC) coupled to each of the server blades and the workload calculation module, wherein the RMC is configured to control the pump controller of the CDU based on the total workload in response to a signal representing the total workload received from the workload calculation module.

US Pat. No. 10,191,493

VEHICLE POSITION POINT FORWARDING METHOD FOR AUTONOMOUS VEHICLES

BAIDU USA LLC, Sunnyvale...

1. A computer-implemented method for operating an autonomous vehicle (ADV) having a set of front wheels as both driving wheels and turning wheels, practiced on the ADV, the ADV having a processor coupled to a memory programmed with executable instructions that, when executed, implement the method on the ADV, the method comprising:determining a first position associated with a set of rear wheels of the autonomous vehicle based on global positioning system (GPS) data received from a GPS source;
determining a moving direction of the autonomous vehicle based on directional data received from an inertial measurement unit (IMU), wherein the IMU is onboard the autonomous vehicle;
calculating a second position associated with the set of front wheels of the autonomous vehicle based on the first position and the moving direction of the autonomous vehicle;
planning, by the ADV, a route segment of a route based on the second position associated with the set of front wheels as a current position of the autonomous vehicle, generating, by the ADV, planning and control data for the route segment utilizing the set of front wheels as both the driving wheels and the turning wheels, wherein the generated planning and control data includes generating speed and direction data to control the ADV along the route segment, and includes location and direction data of obstacles relative to the ADV along the route segment; and
controlling and driving the autonomous vehicle along the route segment based on the planning and control data.

US Pat. No. 10,162,354

CONTROLLING ERROR CORRECTED PLANNING METHODS FOR OPERATING AUTONOMOUS VEHICLES

BAIDU USA LLC, Sunnyvale...

1. A computer-implemented method for operating an autonomous vehicle, the method comprising:receiving motion planning and control data indicating that an autonomous vehicle is to move from a first point to a second point of a path, the motion planning and control data describing a plurality of routes from the first point to the second point within the path;
for each of the routes,
performing a simulation of the route in view of physical characteristics of the autonomous vehicle to generate a simulated route, and
computing a controlling error representing a discrepancy between the route and the simulated route;
selecting one of the routes based on controlling errors between the routes and associated simulated routes; and
operating the autonomous vehicle moving from the first point to the second point according to the selected route.

US Pat. No. 10,238,011

OPTIMAL CONTROLLER FOR HYBRID LIQUID-AIR COOLING SYSTEM OF ELECTRONIC RACKS OF A DATA CENTER

BAIDU USA LLC, Sunnyvale...

1. An electronic rack of a data center, comprising:a plurality of computing nodes arranged in a stack, each computing node containing at least one processor attached to a liquid cold plate, wherein the liquid cold plate is coupled to a heat removal system, separate from the plurality of cooling fans, and the heat removal system cools the liquid received from the liquid cold plate;
a plurality of cooling fans, each of the cooling fans corresponding to one of the computing nodes to provide air cooling to the corresponding computing node, wherein a fan speed of each of the plurality of cooling fans is individually controllable;
a coolant distribution unit (CDU) coupled to each of the computing nodes to provide liquid cooling to a corresponding processor via a corresponding liquid cold plate, wherein the CDU includes a liquid pump to provide cooling liquid to remove heat generated from the processor of each computing node; and
a rack management controller (RMC) coupled to the CDU, the cooling fans, and the computing nodes, wherein the RMC is configured to collect real-time operating data from each of the computing nodes, the cooling fans, and the CDU, perform an optimization using a predetermined optimization function to determine an optimal pump speed of the liquid pump and determine an optimal fan speeds for each of the plurality of the cooling fans based on the collected real-time operating data, such that total power consumption of the liquid pump and the plurality of cooling fans is minimized, control a pump speed of the liquid pump based on the optimal pump speed, and control each fan speed of the plurality of cooling fans based on the optimal fan speed corresponding to each of the plurality of cooling fans, respectively.

US Pat. No. 10,183,641

COLLISION PREDICTION AND FORWARD AIRBAG DEPLOYMENT SYSTEM FOR AUTONOMOUS DRIVING VEHICLES

BAIDU USA LLC, Sunnyvale...

1. A computer-implemented method for operating an autonomous driving vehicle, the method comprising:receiving perception data perceiving a driving environment surrounding an autonomous driving vehicle (ADV), including perceiving an object in front of the ADV;
determining a relative speed and distance between the object and ADV based on the perception data;
determining whether a collision between the ADV and the object is unavoidable based on the relative speed and distance between the object and the ADV, in view of a first physical driving limitation associated with the ADV and a second physical driving limitation associated with the object; and
deploying an airbag equipped with the ADV prior to the collision with the object, in response to determining that the collision is unavoidable.

US Pat. No. 10,169,513

METHOD AND SYSTEM FOR DESIGNING FPGA BASED ON HARDWARE REQUIREMENTS DEFINED IN SOURCE CODE

BAIDU USA LLC, Sunnyvale...

1. A computer-implemented method for programming a field programmable gate array (FPGA), the method comprising:parsing, by a compiler executed by a processor, a source code to identify a first routine to perform a first function and a second routine to perform a second function;
determining, by the compiler, a control signaling topology between the first routine and the second routine based on one or more statements associated with the first routine and the second routine defined in the source code;
allocating, by the compiler, a first logic block describing a first hardware configuration representing the first function of the first routine;
allocating, by the compiler, a second logic block describing a second hardware configuration representing the second function of the second routine; and
generating, by the compiler, a register-transfer level (RTL) netlist based on the first logic block and the second logic block, wherein the second logic block is to perform the second function dependent upon the first function performed by the first logic block based on the control signaling topology, wherein the RTL netlist is provided for subsequent design and manufacturing the FPGA.

US Pat. No. 10,299,413

MODULAR SELF-ALIGNING LIQUID HEAT REMOVAL COUPLING SYSTEM FOR ELECTRONIC RACKS

BAIDU USA LLC, Sunnyvale...

1. An electronic rack, comprising:a back panel disposed on a backend of the electronic rack, wherein the back panel includes a heat removal liquid manifold assembly to provide heat removal liquid from an external heat removal system; and
a plurality of server blades contained in a plurality of server slots via a frontend of the electronic rack, wherein each of the server blades comprises
a server tray to contain an information technology (IT) component representing a corresponding server therein, and
a self-fitting coupling assembly having a first liquid intake connector and a first liquid outlet connector, the first liquid intake connector to receive the heat removal liquid from the heat removal liquid manifold assembly to remove heat from the IT component, the first liquid outlet connector to emit warmer liquid carrying heat exchanged from the IT component to the heat removal liquid manifold assembly, wherein the first liquid intake connector and the first liquid outlet connector are capable of moving around relative to the server tray within a predetermined tolerance space to self-align with a second liquid intake connector and a second liquid outlet connector, wherein the first liquid intake connector and the first liquid outlet connector are not fixedly attached to the server tray, and wherein the second liquid intake connector and the second liquid outlet connector are fixedly disposed on the heat removal liquid manifold assembly of the back panel respectively,
the self-fitting coupling assembly including a floating alignment plate having the first liquid intake connector and the first liquid outlet connector mounted thereon, wherein the floating alignment plate includes a stopper extended outwardly towards the back panel, the stopper including an internal screw thread.

US Pat. No. 10,227,075

SYSTEM DELAY ESTIMATION METHOD FOR AUTONOMOUS VEHICLE CONTROL

BAIDU USA LLC, Sunnyvale...

1. A computer-implemented method for operating an autonomous vehicle, the method comprising:measuring a steering control delay between time of issuing a steering control command and time of a first response from one or more wheels of an autonomous vehicle;
measuring a speed control delay between time of issuing a speed control command and time of a second response from one or more wheels of the autonomous vehicle;
estimating an overall system delay based on the steering control delay and the speed control delay using a predetermined algorithm;
perceiving a driving environment surrounding the autonomous vehicle based on sensor data obtained from a plurality of sensors mounted on the autonomous vehicle at a given point in time;
planning a route to drive the autonomous vehicle along the route based on perception data of the driving environment;
generating planning and control data for the route, the planning and control data including information specifying a plurality of control commands to be issued at different points in time along the route;
modifying the planning and control data in view of the system delay, including modifying timing of at least one of the control commands; and
controlling the autonomous vehicle based on the modified planning and control data.

US Pat. No. 10,327,359

COUPLING DESIGNS FOR A DATA CENTER BUILDING USING INDIRECT EVAPORATIVE COOLING (IDEC) UNITS

BAIDU USA LLC, Sunnyvale...

1. A data center system, comprising:a multi-floor housing comprising:
a first floor housing including a first set of electronic racks of information technology (IT) components operating therein, and
a second floor housing including a second set of electronic racks of the IT components operating therein; and
a plurality of heat-removing indirect evaporative cooling/cooler (IDEC) units, each of the IDEC units including an air-to-air heat exchanger; the plurality of IDEC units comprising:
a first set of one or more of the IDEC units, each of the IDEC units of the first set is coupled to a first portion of the multi-floor housing through a first external side of the multi-floor housing to direct a first portion of exterior air through the air-to-air heat exchanger of a corresponding one of the one or more IDEC units of the first set to cool a first portion of interior air, wherein the first portion of the interior air is directed to cool the first set of electronic racks of the first floor housing; and
a second set of one or more of the IDEC units, each of the IDEC units of the second set is coupled to a second portion of the multi-floor housing through a second external side of the multi-floor housing to direct a second portion of the exterior air through the air-to-air heat exchanger of the corresponding one of the one or more IDEC units of the of the second set to cool a second portion of the interior air, wherein the second portion of the interior air is directed to cool the second set of electronic racks of the second floor housing,
wherein the first external side of the multi-floor housing is a different side than the second external side of the multi-floor housing, the first external side of the multi-floor housing does not have any of the one or more of the IDEC units of the second set, and the second external side of the multi-floor housing does not have any of the one or more of the IDEC units of the first set.

US Pat. No. 10,235,387

METHOD FOR SELECTING IMAGES FOR MATCHING WITH CONTENT BASED ON METADATA OF IMAGES AND CONTENT IN REAL-TIME IN RESPONSE TO SEARCH QUERIES

BAIDU USA LLC, Sunnyvale...

1. A computer-implemented method of selecting images for content items in response to search queries, the method comprising:in response to a search query received at a server from a client device for searching content, searching in a content database or via a content server by a search engine hosted by the server based on one or more search terms of the search query to identify a first list of one or more content items;
identifying, by an image matching module hosted by the server, a list of one or more images by searching in an image store based on the one or more search terms;
for each of the images, calculating, by a ranking module, a matching score between the image and a first content item of the first list based on a context of the first content item and metadata associated with the image;
ranking each of the images based on a respective matching score between the image and the first content item;
selecting one of the images having a matching score greater than a predetermined threshold, wherein the selected image is to be integrated with the first content item;
integrating by the search engine each content item of the first list with one of the images, generating a second list of one or more content items having at least a portion of the images integrated therein; and
transmitting by the search engine the second list of content items to the client device, such that each content item of the first list is presented with one of the images.

US Pat. No. 10,296,535

METHOD AND SYSTEM TO RANDOMIZE IMAGE MATCHING TO FIND BEST IMAGES TO BE MATCHED WITH CONTENT ITEMS

BAIDU USA LLC, Sunnyvale...

1. A computer-implemented method for matching images with content items, the method comprising:receiving a content item that was identified in response to a search query received from a client;
determining whether the search query is within a predetermined percentage of total search queries that yield the content item;
in response to determining that the search query is within the predetermined percentage of the total search queries,
searching in an image store based on one or more keywords associated with the search query to identify a list of one or more images, and
selecting one of the identified images to be matched with the content item;
in response to determining that the search query is not within the predetermined percentage of the total search queries,
for each of predetermined preferred images associated with the content item, calculating by an image mapping server a user interactive score to generate a plurality of user interactive scores,
capturing image access statistics representing user interactions with the image,
identifying by the image mapping server a top predetermined preferred image having a top user interactive score based on the plurality of user interactive scores,
determining whether the top predetermined preferred image is the same as a current preferred image for the content item,
selecting by the image mapping server the top predetermined preferred image to be matched with the content item in response to determining that the top predetermined preferred image is not the current preferred image for the content item, and
replacing the current preferred image with the top predetermined preferred image, in response to determining that the top predetermined preferred image is not the current preferred image; and
incorporating the content item with the selected image to generate a composite content item to be transmitted to the client.

US Pat. No. 10,275,472

METHOD FOR CATEGORIZING IMAGES TO BE ASSOCIATED WITH CONTENT ITEMS BASED ON KEYWORDS OF SEARCH QUERIES

BAIDU USA LLC, Sunnyvale...

1. A computer-implemented method for categorizing images to be associated with content items based on keywords, the method comprising:providing a first keyword-to-image (keyword/image) mapping table, the first keyword mapping table including a plurality of entries, each entry mapping a keyword to one or more image identifiers (IDs) identifying one or more images;
for each of the keywords of the first keyword/image mapping table, determining, by an analysis module executed by a processor, one or more related keywords that are related to the keyword based on a similarity comparison between the one or more related keywords and the keyword;
generating, by a configuration module executed by the processor, one or more additional entries corresponding to the one or more related keywords to be incorporated into the first keyword/image mapping table to generate a second keyword/image mapping table, wherein each of the additional entries is a separate entry that maps one of the related keywords to identical images of the corresponding keyword from the first keyword/image mapping table, such that any keyword matching the related keywords and their respective original keyword leads to an identical image, wherein the second keyword/image mapping table is utilized to associate a particular image with a particular content item related to a particular keyword in response to subsequent search queries; and
for each of the additional entries in the second keyword/image mapping table, filtering, by a filtering and matching module executed by the processor, the related keywords in view of image metadata of corresponding images that correspond to the related keywords to verify whether the related keywords describe the corresponding images, wherein filtering the related keywords comprises:
for each of the related keywords,
determining whether the related keyword describes at least a portion of the image metadata of the corresponding image that corresponds to the related keyword, and
in response to determining that the related keyword does not describe at least a portion of the image metadata, removing the corresponding image that corresponds to the related keyword,
wherein the image metadata of the corresponding image is generated by performing an image recognition.

US Pat. No. 10,272,924

DETERMINING CONTROL CHARACTERISTICS FOR AN AUTONOMOUS DRIVING VEHICLE

BAIDU USA LLC, Sunnyvale...

1. A computer-implemented method of determining characteristics of an autonomous driving vehicle (ADV), comprising:initiating a first driving maneuver for the ADV, the first driving maneuver including accelerating the ADV to a predetermined velocity;
determining a first tractive force required by the ADV to achieve the acceleration during the first driving maneuver;
in response to reaching the predetermined velocity, initiating a second driving maneuver including maintaining a constant velocity for the ADV;
determining a second tractive force required by the ADV to maintain the constant velocity during the second driving maneuver;
calculating an estimated mass of the ADV based on the determined first tractive force, the determined second tractive force, and a rate of the acceleration during the first driving maneuver, wherein initiating the first and second driving maneuvers are performed during a first set of load characteristics for the ADV;
detecting a change from the first set of load characteristics to second set of load characteristics for the ADV; and
in response to detecting the change from the first set of load characteristics to the second set of load characteristics for the ADV, determining an updated estimated mass of the ADV for the second set of load characteristics different from the first set of load characteristics.

US Pat. No. 10,334,763

METHOD FOR PROVIDING COOLING TO ELECTRONIC RACKS USING LIQUID COOLING AND AIR COOLING

BAIDU USA LLC, Sunnyvale...

1. An electronic rack used in a data center, comprising:a housing to house a front panel, a back panel, and a plurality of computing nodes arranged in a stack, each computing node containing at least one processor attached to a liquid cold plate, wherein a front aisle is formed between the front panel and the stack of computing nodes, and a back aisle is formed between the back panel and the stack of computing nodes;
a plurality of cooling fans, each of the cooling fans corresponding to at least one of the computing nodes to provide air cooling to the corresponding computing node, wherein each cooling fan generates an airflow flowing through an airspace of the computing node;
a rack supply line to receive cooling liquid from a coolant distribution unit (CDU);
a rack return line coupled to each of the computing nodes to receive and return warmer liquid carrying at least a portion of heat from the computing node back to the CDU; and
an air-to-liquid (A/L) heat exchanger located underneath the stack of computing nodes at a bottom of the electronic rack and coupled to the rack supply line to receive the cooling liquid, to exchange heat carried by the airflows generated by the cooling fans using the cooling liquid, and then to distribute the cooling liquid to each of the computing nodes to provide liquid cooling to a corresponding processor via a corresponding liquid cold plate;
wherein the airflows are to flow from the A/L heat exchanger into the front aisle as cool airflows, to flow across the airspaces of the computing nodes to exchange at least a portion of the heat generated by the computing nodes, to enter the back aisle as warmer airflows carrying the exchanged heat, and to circulate back to the A/L heat exchanger.

US Pat. No. 10,289,642

METHOD AND SYSTEM FOR MATCHING IMAGES WITH CONTENT USING WHITELISTS AND BLACKLISTS IN RESPONSE TO A SEARCH QUERY

BAIDU USA LLC, Sunnyvale...

1. A computer-implemented method for matching content items with images, the method comprising:identifying, by an image selection module executed by one or more processors, a first set of one or more images as a list of image candidates to be matched with a content item that is identified in response to a search query for the content item, wherein the content item is provided by a first content provider;
performing, by the image selection module, a first lookup operation in an image content provider table to identify a second set of one or more images that are associated with a second content provider other than the first content provider, wherein the content item is not provided by the second content provider;
removing, by the image selection module, any image candidate from the list of image candidates that is included in the second set of the images; and
selecting, by the image selection module, one of the candidate images to be associated with the content item for presenting the content item with the candidate image while avoiding conflicts of the content item with the second content provider.

US Pat. No. 10,289,700

METHOD FOR DYNAMICALLY MATCHING IMAGES WITH CONTENT ITEMS BASED ON KEYWORDS IN RESPONSE TO SEARCH QUERIES

BAIDU USA LLC, Sunnyvale...

1. A computer-implemented method for matching images with content items based on keywords, the method comprising:in response to a search query received from a client device for searching web content, determining, by a query analysis module executed by a processor, one or more first keywords from the search query;
performing by a matching module a lookup operation in an image-to-keyword (image/keyword) mapping table based on the first keywords, the image/keyword mapping table including a plurality of entries, each entry mapping an image identifier (ID) identifying an image corresponding to one or more second keywords, wherein the image/keyword mapping table is indexed based on image IDs;
for each of images identified in the image/keyword mapping table, determining by a ranking module a matching degree between the first keywords and the second keywords, wherein the matching degree is determined based on (i) a first ratio of a number of keywords in the first keywords that match the second keywords to a number of keywords in the first keywords, and (ii) a second ratio of the number of matched keywords to a number of keywords in the second keywords, wherein a matching degree with at least a portion of the first keywords matching the second keywords is higher than a matching degree of which none of the first keywords matches any of the second keywords;
ranking by the ranking module a plurality of images identified in the image/keyword mapping table whose second keywords match at least a portion of the first keywords based on the determined matching degrees; and
selecting, from the ranked images, by the ranking module one or more image candidates having a matching degree higher than a predetermined threshold, wherein the selected image candidates are integrated with a list of content items.

US Pat. No. 10,262,234

AUTOMATICALLY COLLECTING TRAINING DATA FOR OBJECT RECOGNITION WITH 3D LIDAR AND LOCALIZATION

BAIDU USA LLC, Sunnyvale...

1. A computer-implemented method for training object recognition of an autonomous driving vehicle, the method comprising:receiving three-dimensional (3D) light detection and range (LIDAR) data points using a 3D LIDAR device installed within an autonomous driving vehicle (ADV), wherein the 3D LIDAR data points include respective coordinates of the 3D LIDAR data points and respective timestamps of capturing the 3D LIDAR data points;
respectively receiving global positioning system (GPS) messages from a set of objects located within a predetermined proximity surrounding the ADV, wherein each of the GPS messages includes a location of a corresponding object and a timestamp of capturing the location;
for each of the received GPS messages,
searching within the 3D LIDAR data points to match the GPS message with at least one of the 3D LIDAR data points by matching the timestamp of the GPS message and the timestamps of the at least one of the 3D LIDAR data points, and matching the location of the object included in the GPS message with the coordinates of the at least one of the 3D LIDAR data points; and
labeling the at least one of the 3D LIDAR data points based on a type of object specified within the GPS message;
wherein the labeled 3D LIDAR data points are utilized to train a machine-learning algorithm for a plurality of ADVs to recognize other objects in real-time.

US Pat. No. 10,289,110

METHOD TO DYNAMICALLY ADJUSTING STEERING RATES OF AUTONOMOUS VEHICLES

BAIDU USA LLC, Sunnyvale...

15. A data processing system, comprising:a processor; and
a memory coupled to the processor to store instructions, which when executed by the processor, cause the processor to perform operations, the operations including
determining a plurality of steering rate candidates for a steering control command of operating an autonomous vehicle,
for each of the steering rate candidates, calculating a steering rate cost for the steering rate candidate by applying a predetermined cost function, including calculating a first cost for the steering rate candidate based on a difference between a target steering position and a current steering position of the autonomous vehicle using a first predetermined cost function,
selecting one of the steering rate candidates having a lowest steering rate cost as a target steering rate, and
generating a steering control command based on the selected steering rate candidate to control a steering wheel of the autonomous vehicle.

US Pat. No. 10,272,778

METHOD AND SYSTEM FOR DETERMINING UNIT GAIN OF SPEED CONTROL FOR AUTONOMOUS DRIVING VEHICLES

BAIDU USA LLC, Sunnyvale...

1. A computer-implemented method for determining a unit gain of speed control of an autonomous driving vehicle, the method comprising:capturing speeds of the autonomous driving vehicle (ADV) in response to a plurality of speed control commands at different points in time represented by a plurality of command cycles;
for each of the command cycles,
calculating a first speed-changing rate based on the captured speeds of the ADV associated with the command cycle, and
calculating a second speed-changing rate based on a corresponding speed control command using a predetermined algorithm;
determining using a cost function a unit gain of speed control based on the first and second speed-changing rates, such that a cost representing a difference between the first and second speed-changing rates reaches minimum, wherein the unit gain is utilized to generate subsequent speed control commands; and
controlling the ADV based on the generated speed control commands.

US Pat. No. 10,485,143

COLD PLATE ASSEMBLY FOR SERVER LIQUID COOLING OF ELECTRONIC RACKS OF A DATA CENTER

BAIDU USA LLC, Sunnyvale...

1. A cooling module assembly for cooling a processor of a data processing system in an electronic rack of a data center, the cooling module assembly comprising:a cold plate to be positioned adjacent to an exterior surface of a processor to receive heat radiated from the processor;
a liquid distribution channel formed within the cold plate to distribute a cooling liquid through the cold plate to exchange the heat carried by the cold plate, wherein the liquid distribution channel includes
an ingress port to be coupled a liquid supply line to receive the cooling liquid from an external liquid source, wherein the ingress port is disposed on a first edge of the cold plate, and
an egress port to be coupled to a liquid return line to return the cooling liquid having the exchanged heat back to the external liquid source, wherein the egress port is disposed on a second edge of the cold plate opposite to the first edge; and
a cold plate mounting bracket attached to the cold plate to mount the cold plate onto the external surface of the processor, wherein the cold plate mounting bracket includes a first set of mounting slots to be aligned with a second set of mounting slots disposed on a processor mounting bracket that mounts the processor onto a motherboard of the data processing system,
wherein the first set of mounting slots is disposed on a third edge of the cold plate mounting bracket and the second set of mounting slots is disposed on a fourth edge of the cold plate mounting bracket opposite to the third edge, wherein the third edge and the fourth edge are different edges than the first and second edges,
wherein the third edge of the cold plate mounting bracket further comprises a first set of one or more aligning guide pins disposed thereon and extended downwardly, and the fourth edge of the cold plate mounting bracket further comprises a second set of one or more aligning guide pins disposed thereon and extended downwardly,
wherein when the cold plate mounting bracket together with the cold plate is mounted onto the processor mounting bracket, the aligning guide pins of the first set and second set are inserted into corresponding guide pin holes disposed on the processor mounting bracket, such that the mounting slots of the cold plate mounting bracket are aligned with the mounting slots of the processor mounting bracket,
wherein the first set of mounting slots and the second set of mounting slots allow one or more mounting pins to be inserted therethrough in a first angle and to rotate from the first angle to a second angle after the insertion to interlock the cold plate with the exterior surface of the processor;
wherein each mounting pin is configured to insert through a center of a spring in order to insert through a corresponding mounting slot of the first set such that the spring is compressed when the mounting pin is in a locked position, thereby providing a force to push the cold plate mounting bracket and the processor mounting bracket together so as to provide adequate contact between the cold plate and the processor.

US Pat. No. 10,331,138

STANDARD SCENE-BASED PLANNING CONTROL METHODS FOR OPERATING AUTONOMOUS VEHICLES

BAIDU USA LLC, Sunnyvale...

1. A computer-implemented method for operating an autonomous vehicle, the method comprising:sensing, by a data processing system within an autonomous vehicle, a driving environment surrounding the autonomous vehicle based on sensor data obtained from a plurality of sensors mounted on the autonomous vehicle;
generating, by a planning module within the autonomous vehicle based on sensing the driving environment, motion planning and control data to drive the autonomous vehicle from a first point to a second point of a path within a predetermined route, the motion planning and control data including information specifying driving parameters at different points in time along the path;
segmenting, by a path analysis module hosted within the autonomous vehicle, the path from the first point to the second point into a plurality of path segments;
for at least one of the path segments, identifying one of a plurality of predetermined driving scenes that matches motion characteristics of the corresponding path segment;
modifying the motion planning and control data associated with the path segments based on predetermined motion settings of the plurality of path segments; and
driving, by a control module within the autonomous vehicle, the autonomous vehicle through the path segments of the path based on the modified motion planning and control data.

US Pat. No. 10,268,200

METHOD AND SYSTEM TO PREDICT ONE OR MORE TRAJECTORIES OF A VEHICLE BASED ON CONTEXT SURROUNDING THE VEHICLE

BAIDU USA LLC, Sunnyvale...

1. A computer-implemented method for operating an autonomous vehicle, the method comprising:perceiving a surrounding environment of the autonomous vehicle to identify one or more vehicles nearby;
for each of the identified vehicles,
based on a current location of the identified vehicle, obtaining vehicle-independent information, including physical constraints imposed on the identified vehicle,
predicting a first set of one or more trajectories for the identified vehicle by applying a first machine learning model to a first set of features extracted from the vehicle-independent information associated with the identified vehicle,
predicting a second set of one or more trajectories by applying a second machine learning model to a second set of features extracted from sensor data captured by one or more sensors of the autonomous vehicle perceiving behaviors of the identified vehicle, and
determining a final set of predicted trajectories based on the first set of trajectories and the second set of trajectories; and
controlling the autonomous vehicle based on one or more final predicted trajectories of the one or more identified vehicles.

US Pat. No. 10,296,538

METHOD FOR MATCHING IMAGES WITH CONTENT BASED ON REPRESENTATIONS OF KEYWORDS ASSOCIATED WITH THE CONTENT IN RESPONSE TO A SEARCH QUERY

BAIDU USA LLC, Sunnyvale...

1. A computer-implemented method for matching content items with images, the method comprising:receiving content information of a content item, the content item comprising a plaintext content component and being associated with a content identifier (ID), the content information including a plurality of first keywords associated with the content item, at least some of the plurality of first keywords being specified by a provider of the plaintext content component, wherein the content item was obtained from a content database;
performing an analysis on the first keywords associated with the content item;
categorizing the first keywords into a plurality of categories based on the analysis, each category including one or more second keywords selected from the first keywords;
for each of the categories, identifying one or more images based on the corresponding second keywords; and
generating a category/image mapping table for the content item to map each of the categories to the corresponding identified images, wherein the category/image mapping table is associated with the content ID associated with the content item and identifiable by the content ID, wherein the category/image mapping table is utilized to associate the content item with one of the images based on a subsequent category determined in response to a subsequent search query associated with the content ID, the subsequent category corresponding to one of the plurality of categories, wherein a search result in response to the subsequent search query includes the content item and the of the images associated with the subsequent category integrated therein.

US Pat. No. 10,323,952

SYSTEM AND METHOD FOR PRESENTING MEDIA CONTENTS IN AUTONOMOUS VEHICLES

BAIDU USA LLC, Sunnyvale...

1. A computer-implemented method for presenting content in an autonomous vehicle, the method comprising:receiving a first image of a physical object external to the autonomous vehicle, wherein the first image is captured by an image sensor attached to the autonomous vehicle;
performing an image recognition on the first image to derive one or more keywords related to the physical object;
identifying a list of one or more content items based on the one or more keywords, including searching in a content database based on the one or more keywords as an index to identify the one or more content items;
ranking the list of content items based on a user profile of a user riding in the autonomous vehicle as a passenger, wherein the user profile includes a past behavior pattern of the user;
augmenting a first content item selected from the list of content items onto the first image to generate a second image, wherein the second image includes the actual first image, wherein the first content item is selected based on the rankings; and
displaying the second image on a display device within the autonomous vehicle,
wherein the display device is configured to replace a physical window of the autonomous vehicle to dynamically display the second image as if a passenger was looking at the physical object through the physical window, wherein the display device is molded into a shape of the physical window of the autonomous vehicle between the user and the physical object, and wherein the display device comprises a touch screen to receive user interaction with the first content item displayed within the second image.

US Pat. No. 10,373,610

SYSTEMS AND METHODS FOR AUTOMATIC UNIT SELECTION AND TARGET DECOMPOSITION FOR SEQUENCE LABELLING

Baidu USA LLC, Sunnyvale...

1. A computer-implemented method for training a transcription model for speech transcription, the method comprising:inputting a sequence of utterances into the transcription model, with the utterances and associated labels sampled from a training set;
establishing a set of grams comprising uni-grams and high-order grams;
employing striding to process the input sequence of utterances with stride size depending on the largest gram of the gram set;
predicting, from the transcription model, one or more units for each utterance of the sequence of utterances, the one or more units selected from the established set of grams; and
implementing a Gram Connectionist Temporal Classification (Gram-CTC) loss function to measure an error of the predicted one or more units for each utterance given the associated labels, the Gram-CTC loss function automatically selects useful grams and decomposes the input sequence with the selected grams, the Gram-CTC loss function implementation involving forward-backward transitions;
computing a derivative of the loss with respect to parameters of the transcription model; and
updating the transcription model using the derivative through back-propagation.

US Pat. No. 10,366,166

DEEP COMPOSITIONAL FRAMEWORKS FOR HUMAN-LIKE LANGUAGE ACQUISITION IN VIRTUAL ENVIRONMENTS

Baidu USA LLC, Sunnyvale...

1. A computer-implemented method of human-like language acquisition for a compositional framework, the method comprising:receiving, at a perception module, a vision input;
receiving, at a language module, a natural language input comprising one or more words related to the vision input;
transforming the vision input, at the perception module, into a feature map comprising a visual feature map and a spatial feature map;
obtaining, at the language module, an attention distribution over words in the natural language input;
obtaining an averaged word embedding based on the attention distribution;
using a convolution of a masked portion of the averaged word embedding with at least portion of the feature map, at the language module, to generate an attention map; and
responsive to the natural language input being a question, using a recognition module to output an answer by performing steps comprising:
obtaining a question intention from the question using an embedding mask;
obtaining extracted features using at least a portion of the feature map weighted by the attention map;
generating an embedding mask from the question intention;
masking the extracted image features with the embedding mask to obtain one or more masked features; and
classifying the one or more masked features to output the answer in response to the question.

US Pat. No. 10,343,685

PHYSICAL MODEL AND MACHINE LEARNING COMBINED METHOD TO SIMULATE AUTONOMOUS VEHICLE MOVEMENT

BAIDU USA LLC, Sunnyvale...

1. A computer-implemented method for operating an autonomous vehicle, the method comprising:identifying a driving scenario of the autonomous vehicle for a next movement, wherein the driving scenario is represented by a set of one or more predetermined parameters;
calculating a first next movement of the autonomous vehicle under the driving scenario using a physical model;
generating planning and control data based on the first next movement, the planning and control data describing a path to drive the autonomous vehicle from a first point to a second point;
applying a sideslip predictive model to the set of predetermined parameters to determine a sideslip of the autonomous vehicle given a set of driving conditions, the sideslip representing a difference between a first position of the autonomous vehicle and a second position of the autonomous vehicle, wherein the first position is an expected position of the autonomous vehicle calculated using the physical model and the set of predetermined parameters, and wherein the second position is a position where the autonomous vehicle actually ends up due to the set of driving conditions;
determining a second next movement of the autonomous vehicle based on the first next movement and the determined sideslip of the autonomous vehicle, including adjusting at least one position within the path between the first point and the second point based on the determined sideslip of the autonomous vehicle;
modifying the planning and control data based on the second next movement; and
controlling and driving the autonomous vehicle from the first point to the second point according to the path based on the modified planning and control data.

US Pat. No. 10,380,890

AUTONOMOUS VEHICLE LOCALIZATION BASED ON WALSH KERNEL PROJECTION TECHNIQUE

BAIDU USA LLC, Sunnyvale...

1. A computer-implemented method of determining a location of an autonomous driving vehicle (ADV) with respect to a high definition (HD) map and navigating the ADV, the method comprising:determining a first sub-set of a plurality of candidate cells of an ADV feature space of cells surrounding the ADV, the ADV feature space derived from a three-dimensional (3D) point cloud of sensor data obtained by sensors of the ADV;
for each candidate cell in the first sub-set of the plurality of candidate cells:
determining a similarity score between a subset of the ADV feature space that surrounds the candidate cell and an HD map feature space, by projecting the subset of the ADV feature space onto the map feature space using a first dimension projection kernel;
in response to determining that the similarity score is less than a threshold amount, marking the candidate cell as rejected, otherwise storing the similarity score in association with the candidate cell;
determining a location of the ADV with respect to the HD map feature space based at least in part on a candidate cell in the plurality of candidate cells having a highest similarity score among the plurality of candidate cells; and
navigating the ADV along a planned route, the navigating based at least in part on the determined location of the ADV with respect to the HD map.

US Pat. No. 10,372,743

SYSTEMS AND METHODS FOR HOMOGENEOUS ENTITY GROUPING

Baidu USA LLC, Sunnyvale...

1. A computer-implemented method to group synonymous entities from among a set of entities, based at least in part on their semantics, the method comprising:for each entity pair from a set of entity pairs formed from the set of entities, determining a similarity score based upon character similarity between the entities that form the entity pair;
using a prior knowledge dataset to identify one or more sets of similar entities having same or similar meanings as indicated in the prior knowledge dataset, the one or more sets of similar entities comprising entities from an entity dictionary, the entity dictionary comprising the set of entities to be grouped relative to whether they have a same or similar meaning;
using large scale text data to construct embeddings of entities in the set of entities in which embedding training using the large scale text data is constrained using at least some of the sets of similar entities so that entities from a set of similar entities are represented by a same or similar embedding;
for each entity pair from the set of entity pairs, predicting entity relations between the entities that form the entity pair using as a feature set the entity embeddings of the entities that form the entity pair and the similarity score for that entity pair, the prediction being a binary prediction of positive relation of similarity or non-positive relation of similarity;
responsive to one or more criteria not being met:
sampling a set of predicted entity relations for verification;
using verification results to update the prior knowledge dataset; and
returning to the step of using the prior knowledge dataset to identify one or more sets of similar entities having same or similar meanings to re-identify sets of similar entities, re-construct the embeddings of the entities using the re-identified sets of similar entities as constraints to the embedding training, and re-predict entity relations;
responsive to one or more criteria being met, outputting the entity pairs with positive relation predictions; and
using the positive relation predictions to group the entities from the set of entities into one or more entity groups, each entity group comprising entities that have a positive entity relationship linking the entities in the entity group to have a same or similar meaning.

US Pat. No. 10,436,595

METHOD AND SYSTEM FOR UPDATING LOCALIZATION MAPS OF AUTONOMOUS DRIVING VEHICLES

Baidu USA LLC, Sunnyvale...

1. A computer-implemented method to update a localization map of an autonomous driving vehicle (ADV), the method comprising:collecting localization data and a real-time pose of the ADV;
analyzing the collected localization data in view of a first localization map to determine a first confidence score for a plurality of candidate cells within a search space of the collected localization data based on the real-time pose, wherein analyzing the collected localization data comprises: determining a similarity score for each of a the plurality of candidate cells, ranking the plurality of candidate cells by the respective similarity score for each of the candidate cells, and dividing a similarity score of a candidate cell having a highest similarity score by a similarity score of a candidate cell having a second highest similarity score to determine the first confidence score of the plurality of candidate cells;
determining whether to select the candidate cell having the highest similarity score based on the first confidence score; and
applying a subset of the collected localization data onto the first localization map based on the selected candidate cell to generate a second localization map, wherein the second localization map is utilized to subsequently determine a location of the ADV.

US Pat. No. 10,365,649

LANE CURB ASSISTED OFF-LANE CHECKING AND LANE KEEPING SYSTEM FOR AUTONOMOUS DRIVING VEHICLES

BAIDU USA LLC, Sunnyvale...

1. A computer-implemented method for operating an autonomous driving vehicle, the method comprising:detecting, a first point in time, that a wheel of an autonomous driving vehicle (ADV) rolls onto a lane curb disposed on an edge of a lane in which the ADV is moving;
detecting, at a second point in time, that the wheel of the ADV rolls off the lane curb of the lane;
calculating an angle between a moving direction of the ADV and a lane direction of the lane based on a difference between the first point in time and the second point in time in view of a current speed of the ADV, including
determining a curb width of the lane curb based on perception data that perceives the lane curb,
calculating a distance perpendicular to the lane direction of the lane that the ADV moves from the first point in time and the second point in time based on the curb width and a wheel width of the wheel determined based on a specification of the wheel, and
determining the angle based on the distance and the current speed of the ADV in view of the difference between the first point in time and the second point in time; and
generating a control command based on the angle to adjust the moving direction of the ADV to prevent the ADV from further drifting off the lane direction of the lane.

US Pat. No. 10,366,290

SYSTEM AND METHOD FOR PROVIDING AUGMENTED VIRTUAL REALITY CONTENT IN AUTONOMOUS VEHICLES

BAIDU USA LLC, Sunnyvale...

1. A computer-implemented method for content delivery to autonomous vehicles, the method comprising:receiving at a server a request from a first autonomous vehicle for content delivery over a network, the request including an image that was captured by a camera of the first autonomous vehicle while the first autonomous vehicle was driving on a road, the image capturing an object external to the first autonomous vehicle;
in response to the request, performing a vision analysis on an image obtained from the request to determine three-dimensional (3D) positioning information of the image, including determining a position of the object within the image and a viewing angle of the object with respect to a location of the camera;
identifying a list of a plurality of content items to augment based on current vehicle information of the first autonomous vehicle in view of one or more user profiles of one or more users to be entertained while riding the first autonomous vehicle;
augmenting a first content item selected based on user preferences from the list of content items onto the image by superimposing the first content item onto the image based on the position and viewing angle of the object to generate an augmented image, wherein the first content item includes a content item previously cached by a content database maintained by the autonomous vehicle; and
transmitting the augmented image to the first autonomous vehicle over the network, wherein the augmented image is to be displayed on a display device within the autonomous vehicle in a virtual reality manner.

US Pat. No. 10,435,015

SYSTEM DELAY CORRECTED CONTROL METHOD FOR AUTONOMOUS VEHICLES

BAIDU USA LLC, Sunnyvale...

1. A computer-implemented method for operating an autonomous vehicle, the method comprising:determining a driving scenario on a planned path for a next route segment of a route for operating an autonomous vehicle, wherein the autonomous vehicle is to receive a response to a first control command associated with the driving scenario in a first period of time for an actual path of the autonomous vehicle to match the planned path of the autonomous vehicle;
performing a lookup operation in a driving scenario to system delay (scenario/delay) table to locate a matching entry matching the determined driving scenario, wherein the scenario/delay table includes a plurality of entries, each entry corresponding to one of a plurality of driving scenarios of the autonomous vehicle, wherein each entry maps one of the driving scenarios to a system delay corresponding to the driving scenario, wherein the system delay represents a difference in time between issuing a control command and a response of the autonomous vehicle in response to the control command under the corresponding driving scenario;
determining a system delay from the matching entry corresponding to the determined driving scenario; and
controlling and driving the autonomous vehicle in view of the system delay, including issuing the first control command in a second period of time, wherein the second period of time is calculated by deducting the system delay from the first period of time.

US Pat. No. 10,429,849

NON-LINEAR REFERENCE LINE OPTIMIZATION METHOD USING PIECEWISE QUINTIC POLYNOMIAL SPIRAL PATHS FOR OPERATING AUTONOMOUS DRIVING VEHICLES

BAIDU USA LLC, Sunnyvale...

19. A data processing system, comprising:a processor; and
a memory coupled to the processor to store instructions, which when executed by the processor, cause the processor to perform operations, the operations including
in response to a first reference line representing a route from a first location to a second location associated with an autonomous driving vehicle (ADV), segmenting the first reference line into a plurality of reference line segments,
for each of the reference line segments, determining a quintic polynomial function to represent the reference line segment,
determining an objective function based on the quintic polynomial functions of the reference line segments,
performing an optimization on parameters of the quintic polynomial functions in view of a set of constraints associated with the reference line segments, such that an output of the objective function reaches minimum while the set of constraints are satisfied, and
generating a second reference line based on the optimized parameters of the quintic polynomial functions of the objective function, wherein the second reference line is utilized as a reference line of the route to control the ADV.

US Pat. No. 10,421,460

EVALUATION FRAMEWORK FOR DECISION MAKING OF AUTONOMOUS DRIVING VEHICLE

BAIDU USA LLC, Sunnyvale...

1. A computer-implemented method, comprising:receiving a plurality of autonomous driving vehicle (ADV) logs;
extracting a human driving decision corresponding to an ADV driving decision using the plurality of ADV logs;
grading the human driving decision in accordance with a plurality of grading metrics; and
generating a grade for the ADV driving decision that is based at least in part on the grade of the human driving decision and a comparison of the human driving decision to the ADV driving decision, wherein the grade for the ADV driving decision is utilized to modify algorithms for subsequent ADV driving.

US Pat. No. 10,421,463

AUTOMATIC STEERING CONTROL REFERENCE ADAPTION TO RESOLVE UNDERSTEERING OF AUTONOMOUS DRIVING VEHICLES

BAIDU USA LLC, Sunnyvale...

1. A computer-implemented method for operating an autonomous driving vehicle, comprising:receiving a request to turn the autonomous driving vehicle (ADV) from a first direction to a second direction;
determining a plurality of segment masses of a plurality of segments of the ADV, wherein the segment masses are located on a plurality of predetermined locations within a vehicle platform of the ADV;
calculating a location of a mass center for an entire ADV based on the plurality of segment masses of the plurality of segments of the ADV, the mass center representing a center of an entire mass of the entire ADV, wherein the location of the mass center of the entire ADV is determined based on the plurality of segment masses in view of relative locations of the segment masses using a predetermined algorithm; and
generating a steering control command based on the location of the mass center of the entire ADV for steering control of the ADV.

US Pat. No. 10,395,118

SYSTEMS AND METHODS FOR VIDEO PARAGRAPH CAPTIONING USING HIERARCHICAL RECURRENT NEURAL NETWORKS

Baidu USA LLC, Sunnyvale...

1. A computer-implemented method for automating describing a video with a paragraph comprising multiple sentences, the method comprising:producing, using a sentence generator, multiple single sentences, each single sentence being generated sequentially and describing a specific time span and video region or regions within the video, the sentence generator comprising:
a word embedding layer word converting an input word into a word embedding;
a first recurrent layer for language modeling based at least on the word embedding, the first recurrent layer updating its hidden state when receiving a word embedding;
an attention model coupled to the first recurrent layer for selectively focusing on input video features in a video feature pool; and
a multimodal layer for integrating outputs from the first recurrent layer and the attention model to connect vision components with the language model;
using a paragraph generator to affect inter-sentence dependency for the sentence generator when producing a next sentence, the paragraph generator comprising:
a sentence embedding layer receiving word embeddings from the word embedding layer and current hidden state of the first recurrent layer to output a sentence embedding;
a second recurrent layer linked to the sentence embedding layer for inter-sentence dependency modeling, the second recurrent layer updating its hidden state when receiving a sentence embedding; and
a paragraph state component combining current hidden state of the second recurrent layer and the sentence embedding to generate of a current paragraph state as an initial hidden state when the first recurrent layer is reinitialized for next sentence generation.

US Pat. No. 10,488,205

METHOD AND SYSTEM FOR UPDATING MAPS BASED ON CONTROL FEEDBACKS OF AUTONOMOUS DRIVING VEHICLES

BAIDU USA LLC, Sunnyvale...

1. A computer-implemented method for updating maps for autonomous driving, the method comprising:recording a first driving parameter of an autonomous driving vehicle (ADV) driving on a segment of a road in response to a first control command issued autonomously at the ADV;
determining a difference between the first driving parameter and a target driving parameter corresponding to the first control command;
issuing a second control command autonomously at the ADV to compensate and cause the ADV to drive with a second driving parameter closer to the target driving parameter, in response to determining that the difference exceeds a predetermined threshold;
deriving a slope status of the segment of the road based on at least the second control command; and
updating map data of a map corresponding to the segment of the road based on the derived slope status.

US Pat. No. 10,474,157

DATA-BASED CONTROL ERROR DETECTION AND PARAMETER COMPENSATION SYSTEM

BAIDU USA LLC, Sunnyvale...

1. A computer-implemented method of operating an autonomous driving vehicle (ADV), the method comprising:measuring, for a time t0, a first state of the ADV, wherein the first state comprises at least a speed and a location of the ADV and a plurality of control inputs that control the navigation of the ADV;
accessing a state database indexed by at least the speed and the location of the ADV in the first state for the time t0, to retrieve a record of a stored first state of the ADV corresponding to the measured first state at time t0, and to retrieve a record of a stored second state of the ADV corresponding to a time t1=t0+?t, wherein ?t is a predetermined time increment;
measuring a real-time second state of the ADV corresponding to the time t1;
comparing one or more fields of the retrieved stored second state for the time t1 with the measured real-time second state of the ADV for the time t1;
in response to determining that a difference between one or more fields of the retrieved stored second state and corresponding one or more fields of the measured real-time second state is greater than a threshold value, adjusting one or more control input parameters of the ADV.

US Pat. No. 10,474,161

HIGH RESOLUTION 3D POINT CLOUDS GENERATION FROM UPSAMPLED LOW RESOLUTION LIDAR 3D POINT CLOUDS AND CAMERA IMAGES

BAIDU USA LLC, Sunnyvale...

1. A computer-implemented method for operating an autonomous driving vehicle (ADV), the method comprising:receiving a first image captured by a first camera, the first image capturing a portion of a driving environment of the ADV;
receiving a second image representing a first depth map of a first point cloud corresponding to the portion of the driving environment produced by a light detection and ranging (LiDAR) device;
upsampling the second image by a predetermined scale factor to match an image scale of the first image; and
generating a second depth map by applying a convolutional neural network (CNN) model to the first image and the upsampled second image, the second depth map having a higher resolution than the first depth map, wherein the second depth map represents a second point cloud utilized to perceive the driving environment surround the ADV.

US Pat. No. 10,457,294

NEURAL NETWORK BASED SAFETY MONITORING SYSTEM FOR AUTONOMOUS VEHICLES

BAIDU USA LLC, Sunnyvale...

1. A computer-implemented method to alert a user to intervene an autonomous driving system (ADS) of an autonomous driving vehicle (ADV), the method comprising:receiving sensor data from a plurality of sensors of the ADV;
extracting a predetermined number of features from the sensor data using a machine learning model;
determining whether to alert a user to intervene the ADS based on the extracted features using the machine learning model;
generating a safety alarm based on the determining whether to alert the user to intervene the ADS; and
alerting the user to intervene the ADS of the ADV based on the generated safety alarm.

US Pat. No. 10,423,874

INTELLIGENT IMAGE CAPTIONING

Baidu USA LLC, Sunnyvale...

1. A computer-implemented method for generating a sentence-level description of an image, the method comprising:receiving an input image;
inputting the input image into a multimodal recurrent neural network (m-RNN), the m-RNN comprising:
a convolution neural network component that generates an image representation of the input image;
at least one word embedding component that encodes syntactic and semantic meaning of a word into a word representation;
a recurrent layer component that maps a recurrent layer activation of a prior time frame into a same vector space as a word representation at a current time frame and combines them;
a multimodal component that receives a first input from the recurrent layer component related to one or more word representations and a second input from the convolution neural network component related to the image representation of the input image and combines them; and
a softmax layer component that uses an output of the multimodal component to generate a probability distribution of a next word in the sentence-level description; and
wherein the m-RNN performs the steps comprising:
determining whether an end sign has been generated as a next word based upon the probability distribution;
responsive to the next word not being the end sign:
adding the next word to the sentence-level description; and
for a next time frame:
setting the next word as the current word; and
inputting the current word into the at least one word embedding component of the m-RNN to obtain a next word; and
returning to the step of determining whether an end sign has been generated as a next word; and
responsive to the next word being the end sign, outputting the sentence-level description.

US Pat. No. 10,409,280

CONTROL DOMINATED PLANNING AND CONTROL SYSTEM FOR AUTONOMOUS DRIVING VEHICLES

BAIDU USA LLC, Sunnyvale...

1. A computer-implemented method for operating an autonomous driving vehicle, comprising:in response to a target position for an autonomous driving vehicle (ADV), estimating a control command based on the target position in view of a current speed and a current position of the ADV;
obtaining a range of confidence positions with respect to the target position that the ADV potentially reaches in response to the estimated control command, wherein the range of confidence positions was predetermined based on prior driving statistics;
determining whether the range of confidence positions is within a safety zone that is determined at a point in time based on perception data perceiving a driving environment surrounding the ADV; and
applying the control command to the ADV in response to determining that the range of confidence positions is within the safety zone.

US Pat. No. 10,407,076

METHOD AND SYSTEM FOR DETERMINING ROAD FRICTIONS OF AUTONOMOUS DRIVING VEHICLES USING LEARNING-BASED MODEL PREDICTIVE CONTROL

BAIDU USA LLC, Sunnyvale...

17. A data processing system, comprising:a processor; and
a memory coupled to the processor to store instructions, which when executed by the processor, cause the processor to perform operations, the operations including
receiving planning data to drive an autonomous driving vehicle (ADV) from a starting location to a destination location,
generating a series of control commands based on the planning data, the control commands to be applied at a plurality of points in time from the starting location to the destination location,
calculating a cost by applying a cost function to the control commands, a first road friction to be estimated in a current trip from the starting location to the destination location, and a second road friction estimated during a prior trip from the starting location to the destination location, and
estimating the first road friction of the current trip using the cost function in view of a prior termination cost of the prior trip, such that the cost reaches minimum.

US Pat. No. 10,394,245

METHOD AND SYSTEM TO PREDICT VEHICLE TRAFFIC BEHAVIOR FOR AUTONOMOUS VEHICLES TO MAKE DRIVING DECISIONS

BAIDU USA LLC, Sunnyvale...

1. A computer-implemented method for operating an autonomous vehicle, the method comprising:generating one or more predicted trajectories based on sensor data received from one or more sensors of the autonomous vehicle, each of the predicted trajectories being associated with a probability;
detecting one or more gestures from one or more vehicles other than the autonomous vehicle in accordance with a gesture detection protocol;
emitting one or more gestures from the autonomous vehicle to communicate with the one or more vehicles in accordance with a gesture emission protocol based on the detected gestures;
modifying the predicted trajectories based on the detected gestures, the emitted gestures, and the associated probabilities of the predicted trajectories; and
controlling the autonomous vehicle based on the modified predicted trajectories.

US Pat. No. 10,481,863

SYSTEMS AND METHODS FOR IMPROVED USER INTERFACE

Baidu USA LLC, Sunnyvale...

1. A non-transitory computer-readable medium or media comprising one or more sequences of instructions which, when executed by one or more processors, causes the steps to be performed comprising:presenting on a touch-sensitive display of a computing device a graphical user interface for voice input, the graphical user interface comprising:
a voice input button that, when selected by a user, causes audio input for transcription to be received via a microphone of the computing device, which transcription is displayed in a display portion that displays text of the transcription, wherein at least a portion of the audio input is submitted to a trained neural network that uses a set of spectrogram frames that represent the at least a portion of the audio input to output a set of transcribed characters corresponding to character probabilities predicted by the trained neural network;
a trackpad portion that is configured to receive via the touch-sensitive display one or more gestures from the user; and
a navigation portion that displays a plurality of buttons, each button, when selected by the user via the touch-sensitive display, activates a function associated with that button, at least one of the buttons, when activated causes an alphabet keyboard to be displayed to the user that facilitates input of text by the user.

US Pat. No. 10,474,160

HIGH RESOLUTION 3D POINT CLOUDS GENERATION FROM DOWNSAMPLED LOW RESOLUTION LIDAR 3D POINT CLOUDS AND CAMERA IMAGES

BAIDU USA LLC, Sunnyvale...

1. A computer-implemented method for operating an autonomous driving vehicle (ADV), the method comprising:receiving a first image captured by a first camera, the first image capturing a portion of a driving environment of the ADV;
receiving a second image representing a first depth map of a first point cloud corresponding to the portion of the driving environment produced by a light detection and ranging (LiDAR) device;
downsampling the second image by a predetermined scale factor until a resolution of the second image reaches a predetermined threshold; and
generating a second depth map by applying a convolutional neural network (CNN) model to the first image and the downsampled second image, the second depth map having a higher resolution than the first depth map, wherein the second depth map represents a second point cloud perceiving the driving environment surrounding the ADV.

US Pat. No. 10,474,206

RETAINING TRAY FOR RETAINING AN M.2 COMPATIBLE DEVICE USED IN AUTONOMOUS DRIVING VEHICLES

BAIDU USA LLC, Sunnyvale...

1. A tray device to retain an M.2 module used in an autonomous driving vehicle, the tray device comprising:a baseboard configured in a relatively rectangular shape having a first and second longitudinal edges and a first and second lateral edges;
a first guardrail disposed on a top surface of the baseboard along the first longitudinal edge;
a second guardrail disposed on the top surface of the baseboard along the second longitudinal edge, wherein the first and second guardrails are to keep an M.2 module in between and supported by the top surface of the baseboard, while allowing the M.2 module to extend beyond the first and second lateral edges; and
a riser member disposed on a bottom surface of the baseboard, wherein when the M.2 module is mounted together with the tray device on a motherboard of a data processing system within an autonomous driving vehicle, the riser member is configured to raise the tray device above the motherboard to prevent the baseboard from applying a pressure that may adversely affect an electronic component mounted on the motherboard and underneath the tray device.

US Pat. No. 10,459,441

METHOD AND SYSTEM FOR OPERATING AUTONOMOUS DRIVING VEHICLES BASED ON MOTION PLANS

BAIDU USA LLC, Sunnyvale...

1. A computer-implemented method of operating an autonomous vehicle, the method comprising:determining a first state of the autonomous vehicle from among a number of states of a finite state machine of the autonomous vehicle, wherein the first state of the finite state machine is determined based on current information of the autonomous vehicle and historical information of the autonomous vehicle;
determining whether one or more conditions have been satisfied based on the current information and the historical information;
determining, based on the one or more conditions that are determined to have been satisfied, a next state of the autonomous vehicle from among the number of states and a state transition of the autonomous vehicle from the first state to the next state, wherein one of the states is an abnormal state indicating no routing communication is received for a predetermined period of time, wherein the next state of the autonomous vehicle is determined using a first lookup table storing a correspondence between the one or more conditions and the next state of the autonomous vehicle, wherein the one or more conditions are represented in the first lookup table as a binary number, wherein the one or more conditions are categorized into predetermined categories of conditions and one category cannot be a subset of another category of conditions, and wherein the next state of the autonomous vehicle is determined based on a result from each category of conditions;
selecting, based on the state transition of the autonomous vehicle from the first state to the next state, one of a plurality of predetermined motion plans, each of the predetermined motion plans corresponding to one of a plurality of state transitions from a particular state to another particular state of the finite state machine; and
generating a trajectory based on the selected motion plan to control the autonomous vehicle to drive according to the trajectory.

US Pat. No. 10,459,970

METHOD AND SYSTEM FOR EVALUATING AND RANKING IMAGES WITH CONTENT BASED ON SIMILARITY SCORES IN RESPONSE TO A SEARCH QUERY

BAIDU USA LLC, Sunnyvale...

1. A computer-implemented method for evaluating matching of content items with images, the method comprising:receiving a request for evaluating matching between a content item and an image that were identified and selected from a plurality of content items and images in response to a search query;
obtaining first metadata associated with the content item, second metadata associated with the image, and third metadata associated with the search query, wherein the third metadata comprises keywords of the search query;
determining a first similarity score, “QC”, based on the first metadata associated with the content item and the third metadata associated with the search query, wherein QC represents a similarity between the search query and the content item;
determining a second similarity score, “QI”, based on the second metadata associated with the image and the third metadata associated with the search query, where QI represents a similarity between the search query and the image;
determining a third similarity score, “CI”, based on the first metadata associated with the content item and the second metadata associated with the image, wherein CI represents a similarity between the content item and the image;
calculating an evaluation score for matching of the content item and the image in view of the search query, wherein the evaluation score is an average of: a product of the first similarity score and the second similarity score, a product of the first similarity score and the third similarity score, and a product of the second similarity score and the third similarity score;
integrating the content item and the image, based on the evaluation score; and
returning the integrated content item and image in response to the search query.

US Pat. No. 10,452,074

METHOD AND SYSTEM FOR CONTROLLING AUTONOMOUS DRIVING VEHICLE REENTERING AUTONOMOUS DRIVING MODE

BAIDU USA LLC, Sunnyvale...

1. A computer-implemented method for operating an autonomous driving vehicle, the method comprising:detecting that the autonomous driving vehicle (ADV) transitions from a manual driving mode to an autonomous driving mode;
determining a first pedal value corresponding to a speed of the ADV in response to a speed control command issued at a previous command cycle during which the ADV was operating in the manual driving mode;
determining a second pedal value based on a target speed of the ADV at a current command cycle during which the ADV is operating in the autonomous driving mode; and
issuing a speed control command to the ADV based on the first pedal value and the second pedal value, wherein the speed control command results in a substantially gradual speed change of the ADV between the manual driving mode and the autonomous driving mode.

US Pat. No. 10,442,435

SPEED CONTROL PARAMETER ESTIMATION METHOD FOR AUTONOMOUS DRIVING VEHICLES

BAIDU USA LLC, Sunnyvale...

1. A computer-implemented method for operating an autonomous driving vehicle (ADV), the method comprising:issuing, by a speed controller, a speed control command in response to receiving a first target speed;
determining a first feedback parameter as a difference between an expected speed and an actual speed of the ADV in response to the speed control command, wherein the expected speed is calculated by a powertrain reference model based on the first target speed, wherein the powertrain reference model is created offline based on a plurality of design characteristics of the ADV;
determining a second feedback parameter by applying a machine learning model trained offline based on driving statistics captured from a variety of vehicles to a set of input parameters that are obtained in real-time, the set of input parameters representing a driving environment of the ADV at a point in time;
dynamically adjusting a control parameter of the speed controller based on the first feedback parameter, the second feedback parameter, and the speed control command as a third feedback parameter, wherein each of the first feedback parameter, the second feedback parameter, and the third feedback parameter is associated with an individual weight factor; and
generating a subsequent speed control command based on the adjusted control parameter of the speed controller for a subsequent command cycle in response to receiving a second target speed.

US Pat. No. 10,438,074

METHOD AND SYSTEM FOR CONTROLLING DOOR LOCKS OF AUTONOMOUS DRIVING VEHICLES BASED ON LANE INFORMATION

BAIDU USA LLC, Sunnyvale...

1. A computer-implemented method for operating an autonomous driving vehicle (ADV), the method comprising:receiving perception data that perceives a driving environment surrounding an autonomous driving vehicle (ADV);
determining, based on the perception data, a lane configuration of one or more lanes of a road in which the ADV is moving;
determining a speed of the ADV and a lane location of the ADV within a lane in which the ADV is moving;
deriving a driving scenario based on the lane configuration, the speed of the ADV, and the lane location of the ADV; and
locking or unlocking one or more door locks of the ADV based on the driving scenario, while the ADV is moving within the lane.

US Pat. No. 10,423,652

KNOWLEDGE GRAPH ENTITY RECONCILER

BAIDU USA LLC, Sunnyvale...

1. A computer-implemented method for expanding and reinforcing a knowledge graph, the method comprising:receiving source data comprising a source data topic entity and a plurality of source data edges, wherein each edge of the plurality of source data edges comprises a relationship-entity pair associated with the source data topic entity;
generating a source data graph from the source data topic entity and the plurality of source data edges;
identifying a candidate topic entity in a knowledge graph having a plurality of knowledge graph edges, wherein each edge of the plurality of knowledge graph edges comprises a knowledge graph relationship-entity pair associated with the knowledge graph candidate topic entity;
determining a similarity between the source data graph and a sub-graph of the knowledge graph having the candidate topic entity, the sub-graph having a plurality of knowledge graph edges associated with the candidate topic entity, wherein determining the similarity comprises determining a similarity between the plurality of source data edges and the plurality of knowledge sub-graph edges, and wherein determining the similarity between the plurality of source data graph edges and the plurality of knowledge sub-graph edges comprises determining a ratio of an intersection of the plurality of source data graph edges with the plurality of knowledge sub-graph edges to a union of the plurality of source data edges with the plurality of knowledge sub-graph edges;
merging the source data graph into the knowledge graph, in response to determining that the similarity is greater than a threshold value.

US Pat. No. 10,489,448

METHOD AND SYSTEM FOR DYNAMICALLY RANKING IMAGES TO BE MATCHED WITH CONTENT IN RESPONSE TO A SEARCH QUERY

BAIDU USA LLC, Sunnyvale...

1. A computer-implemented method for ranking images to be matched with content items, the method comprising:configuring a server with a search engine running thereon, the server including a query-image mapping table with mappings between keywords and image identifiers;
collecting metadata of images associated with the image identifiers, the metadata including sources of the images, contents of the images, attributes of the images, and prior user interactions with the images;
executing an image ranking model on the server, the image ranking model is created based on the metadata of the images and is configured to calculate ranking scores for the images based on a plurality of matching parameters obtaining from the metadata, each matching parameter includes a respective weight;
in response to a search query received at the search engine running on the server, identifying a plurality of image identifiers in the query-image mapping tables based on one or more keywords extracted from the search query;
identifying a plurality of images from an image server using the plurality of image identifiers to be matched with a content item that represents a text description, the content item identified by the search engine performing a search in a content database using the search query;
for each of the identified images, calculating a final matching score using the plurality of matching parameters, including
calculating a matching type score of a matching type parameter among the plurality of matching parameters based on a first set of one or more keywords associated with the search query, a second set of one or more keywords mapped with the image in the image-query mapping table, and the weight of the matching type parameter,
calculating a matching quality score of a matching quality parameter among the plurality of matching parameters based on the matching type score, a first ratio, a second ratio, and the weight of the matching quality parameter,
wherein the first ratio represents a ratio between a number of matched keywords between the first set of one or more keywords and the second set of one or more keywords and a number of keywords in the first set of one or more keywords, and wherein the second ratio represents a ratio between the number of matched keywords and a number of keywords in the second set of one or more keywords,
calculating a click-through score of a click-through parameter among the plurality of matching parameters based on prior user actions with the identified image and prior user interactions with the content item, and the weight of click-through parameter,
calculating an image quality score of an image quality parameter among the plurality of matching parameters based on one or more of a plurality of attributes of the identified image, and the weight of image quality parameter,
wherein the one or more of the plurality of attributes including a size of the identified image, a resolution of the identified image, an aspect ratio of the identified image, and a time when the identified image was last updated,
calculating an image style score of an image style parameter among the plurality of matching parameters based on whether the identified image is in a landscape format or in a portrait format and the weight of image style parameter, and
calculating a final ranking score by summing a dot product of each of the calculated scores;
ranking the plurality of images based on their final matching scores the image ranking model;
selecting one of the images with a final matching score higher than a predetermined threshold, wherein the selected image is incorporated with the content item to generate an incorporated content item; and
returning the incorporated content item and the selected image from the search engine to a client device.

US Pat. No. 10,452,065

HUMAN-MACHINE INTERFACE (HMI) ARCHITECTURE

BAIDU USA LLC, Sunnyvale...

1. A computer-implemented method of operating an autonomous driving vehicle (ADV), the method comprising:receiving, at a Web interface of a web server, an ADV command from a first client application, the ADV command being received using a Web communication protocol, wherein the first client application is one of a plurality of client applications hosted by the Web server;
in response to the ADV command, transmitting the ADV command via a remote procedure call (RPC) to a proxy server of a robotic operating system of the ADV, the proxy server configured to route the ADV command to an ADV perception and control system via a Controller Area Network (CAN) message bus; and
in response to receiving an ADV command response from the ADV perception and control system, wherein the ADV command response is stored in a results memory of the web server, multicasting the ADV command response to the first client application via a first socket connection and to a second client application of the plurality of client applications via a second socket connection.

US Pat. No. 10,496,098

ROAD SEGMENT-BASED ROUTING GUIDANCE SYSTEM FOR AUTONOMOUS DRIVING VEHICLES

BAIDU USA LLC, Sunnyvale...

1. A computer-implemented method for operating an autonomous driving vehicle, the method comprising:receiving a routing request for routing an autonomous driving vehicle (ADV) from a source lane to a target lane, the source lane being associated with a source road and the target lane being associated with the target road;
in response to the routing request, determining one or more road paths from the source road to the target road, the road paths including zero or more intermediate roads between the source road and the target road, wherein each of the intermediate roads includes one or more intermediate lanes;
for each of the road paths, determining one or more lane paths, each lane path having a plurality of lanes combined to connect the source lane of the source road to the target lane of the target road via zero or more of the intermediate lanes of the intermediate roads;
generating a plurality of lane regions based on the lane paths, each lane region having one or more lanes of the lane paths connected in sequence without having to change lane laterally, at least one lane region including lanes connected in sequence across a plurality of roads without having to change lane laterally;
planning a trajectory from the source lane of the source road to the target lane of the target road using the lane paths and the lane regions, comprising searching through the plurality of lane regions to identify a route representing the trajectory, including identifying a lane changing point between at least two of the lane regions; and
generating control signals to drive the ADV according to the trajectory.

US Pat. No. 10,496,364

SYSTEM AND METHOD FOR CONTROLLING COLORS OF SMART LIGHTS BASED ON USER INTENT USING NATURAL LANGUAGE PROCESSING

BAIDU USA LLC, Sunnyvale...

1. A computer-implemented method for controlling smart lights, the method comprising:in response to a text stream in a text format originated from a user at an electronic device, performing a natural language processing (NLP) operation on the text stream;
determining an object described by the text stream based on the NLP processing operation and a machine-learning model;
determining one or more colors associated with the object by using a color database within the electronic device and if the one or more colors associated with the object cannot be determined using the color database:
transmitting an image search request to a remote image search server to search possible images concerning the object; and
in response to an image received from the remote image search server, performing an image analysis on the image to determine the one or more colors from the image; and
transmitting a light control command from the electronic device to each of a plurality of smart lights to control a color of each smart light based on the one or more colors associated with the object, such that the smart lights are lit with colors that are associated with the object.

US Pat. No. 10,496,686

METHOD AND SYSTEM FOR SEARCHING AND IDENTIFYING CONTENT ITEMS IN RESPONSE TO A SEARCH QUERY USING A MATCHED KEYWORD WHITELIST

BAIDU USA LLC, Sunnyvale...

1. A computer-implemented method for searching content, the method comprising:providing, on a server with a search engine running thereon, a first mapping table and a second mapping table, the first mapping table including mappings between content keywords (CKs) and content identifiers, and the second mapping table including mappings between query keywords (QKs) and the CKs,
wherein the first mapping table is created by the server that performs one or more operations, including determining a plurality of content keyword elements (CKEs) based on a plurality of CKs associated with a plurality of content items, and
wherein the second mapping table is created by the server that performs one or more operations, including
determining a plurality of query keyword elements (QKEs) based on a plurality of QKs associated with a plurality of historical search queries,
identifying a first historical search query having one or more query keywords represented by a first QKE,
identifying a first content keyword represented by a first CKE that matches the first QKE, and
creating a new entry in the second mapping table to map the first search query to the first content keyword;
in response to a search query received by the search engine from a client, extracting one or more keywords from the search query;
for each of the one or more keywords extracted from the search query, determining one or more synonymous words;
performing, by the search engine, a first lookup operation in the second mapping table using the extracted keywords and the determined synonymous words to obtain one or more mapping content keywords;
performing, by the search engine, a second lookup operation in the first mapping table using the one or more content keywords to obtain one or more mapping content identifiers;
performing, by the search engine based on the one or more content identifiers, a third lookup operation in a content database on the server to retrieve one or more content items;
for each of the one or more content items, selecting an image to be associated with the content item to create an incorporated content item, the incorporated content item including the image as a background image to the content item; and
returning, by the search engine, the one or more incorporated content items as part of a search result of the search query to the client.

US Pat. No. 10,496,698

METHOD AND SYSTEM FOR DETERMINING IMAGE-BASED CONTENT STYLES

BAIDU USA LLC, Sunnyvale...

1. A computer-implemented method for determining image-based content styles, the method comprising:in response to a content description describing a content item, performing an analysis on the content description;
identifying an image having image content that is related to the content description based on the analysis;
deriving one or more phrases from the content description based on the analysis, the one or more phrases summarizing the content description;
determining a list of a plurality of content styles based on a relationship between the image and the one or more phrases, each of the plurality of content styles corresponding to a style template, each style template comprising style metadata specifying relative positions of the image and the one or more phrases within a composite content item;
determining a selected content style from the list of content styles based on a content style selection received from a user, the selected content style corresponding to a first style template;
storing the first style template; and
generating a composite content item having the image and the one or more phrases therein based on the stored first style template, the image and the one or more phrases being arranged according to a layout corresponding to the first style template.