WO2025112451A1 - 能够遵循指令实现自主脱困的自动驾驶方法、装置和车辆 - Google Patents

能够遵循指令实现自主脱困的自动驾驶方法、装置和车辆 Download PDF

Info

Publication number
WO2025112451A1
WO2025112451A1 PCT/CN2024/099367 CN2024099367W WO2025112451A1 WO 2025112451 A1 WO2025112451 A1 WO 2025112451A1 CN 2024099367 W CN2024099367 W CN 2024099367W WO 2025112451 A1 WO2025112451 A1 WO 2025112451A1
Authority
WO
WIPO (PCT)
Prior art keywords
information
moment
tensor
autonomous driving
input
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
PCT/CN2024/099367
Other languages
English (en)
French (fr)
Inventor
黄际洲
王凡
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beijing Baidu Netcom Science and Technology Co Ltd
Original Assignee
Beijing Baidu Netcom Science and Technology Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Beijing Baidu Netcom Science and Technology Co Ltd filed Critical Beijing Baidu Netcom Science and Technology Co Ltd
Publication of WO2025112451A1 publication Critical patent/WO2025112451A1/zh
Anticipated expiration legal-status Critical
Pending legal-status Critical Current

Links

Classifications

    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B60—VEHICLES IN GENERAL
    • B60W—CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION; CONTROL SYSTEMS SPECIALLY ADAPTED FOR HYBRID VEHICLES; ROAD VEHICLE DRIVE CONTROL SYSTEMS FOR PURPOSES NOT RELATED TO THE CONTROL OF A PARTICULAR SUB-UNIT
    • B60W60/00—Drive control systems specially adapted for autonomous road vehicles
    • B60W60/001—Planning or execution of driving tasks

Definitions

  • the present disclosure relates to the field of computer technology, in particular to the field of autonomous driving and artificial intelligence technology, and specifically to an autonomous driving method, device and vehicle capable of assisting escape from distress.
  • Artificial intelligence is a discipline that studies how to use computers to simulate certain human thought processes and intelligent behaviors (such as learning, reasoning, thinking, planning, etc.). It includes both hardware-level and software-level technologies. Artificial intelligence hardware technologies generally include technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, and big data processing; artificial intelligence software technologies mainly include computer vision technology, speech recognition technology, natural language processing technology, as well as machine learning/deep learning, big data processing technology, knowledge graph technology, and other major directions.
  • Autonomous driving technology integrates many technologies such as recognition, decision-making, positioning, communication security, and human-computer interaction. Artificial intelligence learning can assist in generating autonomous driving strategies.
  • the present disclosure provides an automatic driving method, device and vehicle that can follow instructions to achieve assisted autonomous escape.
  • an autonomous driving method comprising: obtaining input information, the input information including historical decision information, perception information, traffic information, and interaction information at a current moment;
  • the method comprises the steps of: encoding the historical decision information, the perception information, the traffic information and the interaction information to obtain a first tensor, a second tensor, a third tensor and a fourth tensor respectively used to represent the historical decision information, the perception information, the traffic information and the interaction information; performing autoregressive inference on an input tensor formed by the first tensor, the second tensor, the third tensor and the fourth tensor to obtain a hidden state for a first moment after the current moment; and decoding based on the hidden state at the first moment to obtain the interaction information for the first moment and the autonomous driving decision information for the first moment; wherein the interaction information at the first moment includes a signal for indicating that the autonomous driving process requires assistance.
  • an automatic driving device comprising: an acquisition unit, configured to acquire input information, the input information including historical decision information, perception information, traffic information and interaction information at a current moment; an encoding unit, configured to encode the historical decision information, the perception information, the traffic information and the interaction information to obtain a first tensor, a second tensor, a third tensor and a fourth tensor respectively used to represent the historical decision information, the perception information, the traffic information and the interaction information; an inference unit, configured to infer the input tensor formed by the first tensor, the second tensor, the third tensor and the fourth tensor to obtain a hidden state for a first moment after the current moment; a decoding unit, configured to decode based on the hidden state at the first moment to obtain interaction information for the next moment and automatic driving decision information for the first moment, wherein the interaction information at the first moment includes a signal for indicating that the automatic driving process requires assistance.
  • an electronic device comprising: at least one processor; and a memory communicatively connected to the at least one processor; the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor so that the at least one processor can execute the above method.
  • a non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to enable the computer to execute the above method.
  • a computer program product comprising a computer program, wherein the computer program implements the above method when executed by a processor.
  • an autonomous driving vehicle comprising: an autonomous driving device or one of electronic devices according to an embodiment of the present disclosure.
  • the prompt information of natural language interaction can be used to guide the autonomous driving model to control the vehicle, so as to realize the rapid generation of the vehicle escape plan, and the vehicle can escape autonomously according to the escape plan instructions.
  • the autonomous driving model can actively initiate an interaction request to request remote assistance when appropriate, and can realize cloud guidance for multiple autonomous driving vehicles.
  • FIG1 shows a schematic diagram of an exemplary system in which various methods described herein may be implemented according to an embodiment of the present disclosure
  • FIG2 shows an exemplary block diagram of an autonomous driving model according to an embodiment of the present disclosure
  • FIG3A shows an exemplary flow chart of an autonomous driving method implemented by using an autonomous driving model according to an embodiment of the present disclosure
  • FIG. 3B shows an exemplary process for assisting autonomous driving according to an embodiment of the present disclosure.
  • FIG4 shows a structural block diagram of an automatic driving device according to an embodiment of the present disclosure.
  • FIG. 5 shows a block diagram of an exemplary electronic device that can be used to implement the embodiments of the present disclosure.
  • first, second, etc. to describe various elements is not intended to limit the positional relationship, temporal relationship, or importance relationship of these elements, and such terms are only used to distinguish one element from another element.
  • first element and the second element may refer to the same instance of the element, and in some cases, based on the description of the context, they may also refer to different instances.
  • FIG1 shows a schematic diagram of an exemplary system 100 in which various methods and apparatuses described herein may be implemented according to an embodiment of the present disclosure.
  • the system 100 includes a motor vehicle 110, a server 120, and one or more communication networks 130 coupling the motor vehicle 110 to the server 120.
  • the motor vehicle 110 may include a computing device according to an embodiment of the present disclosure and/or be configured to perform a method according to an embodiment of the present disclosure.
  • the server 120 may run one or more services or software applications that enable autonomous driving. In some embodiments, the server 120 may also provide other services or software applications, which may include non-virtual environments and virtual environments. In the configuration shown in FIG1 , the server 120 may include one or more components that implement the functions performed by the server 120. These components may include software components, hardware components, or a combination thereof that may be executed by one or more processors. A user of the motor vehicle 110 may in turn utilize one or more client applications to interact with the server 120 to utilize the services provided by these components. It should be understood that a variety of different system configurations are possible, which may be different from the system 100. Therefore, FIG1 is an example of a system for implementing the various methods described herein and is not intended to be limiting.
  • Server 120 may include one or more general purpose computers, dedicated server computers (e.g., PC (personal computer) servers, UNIX servers, mid-range servers), blade servers, mainframe computers, server clusters, or any other suitable arrangement and/or combination.
  • Server 120 may include one or more virtual machines running virtual operating systems, or other computing architectures involving virtualization (e.g., one or more flexible pools of logical storage devices that may be virtualized to maintain a server's virtual storage device).
  • server 120 may run one or more services or software applications that provide the functionality described below.
  • the computing units in the server 120 may run one or more operating systems including any of the above operating systems and any commercially available server operating systems.
  • the server 120 may also run any of a variety of additional server applications and/or middle-tier applications, including HTTP servers, FTP servers, CGI servers, JAVA servers, database servers, etc.
  • server 120 may include one or more applications to analyze and consolidate data feeds and/or event updates received from motor vehicle 110.
  • Server 120 may also include one or more applications to display data feeds and/or real-time events via one or more display devices of motor vehicle 110.
  • the network 130 may be any type of network known to those skilled in the art that may support data communications using any of a variety of available protocols, including, but not limited to, TCP/IP, SNA, IPX, etc.
  • one or more networks 130 may be a satellite communications network, a local area network (LAN), an Ethernet-based network, a token ring, a wide area network (WAN), the Internet, a virtual network, a virtual private network (VPN), an intranet, an extranet, a blockchain network, a public switched telephone network (PSTN), an infrared network, a wireless network (including, for example, Bluetooth, WiFi), and/or any combination of these and other networks.
  • LAN local area network
  • Ethernet-based network a token ring
  • WAN wide area network
  • VPN virtual private network
  • PSTN public switched telephone network
  • PSTN public switched telephone network
  • WiFi wireless network
  • the system 100 may also include one or more databases 150.
  • these databases may be used to store data and other information.
  • one or more of the databases 150 may be used to store information such as audio files and video files.
  • the data repository 150 may reside in various locations.
  • the data repository used by the server 120 may be local to the server 120, or may be remote from the server 120 and may communicate with the server 120 via a network-based or dedicated connection.
  • the data repository 150 may be of different types.
  • the data repository used by the server 120 may be a database, such as a relational database.
  • One or more of these databases may store, update, and retrieve data to and from the database in response to a command.
  • one or more of the databases 150 may also be used by an application to store application data.
  • the databases used by the application may be different types of databases, such as a key-value store, an object store, or a conventional store backed by a file system.
  • the motor vehicle 110 may include a sensor 111 for sensing the surrounding environment.
  • the sensor 111 may include one or more of the following sensors: a visual camera, an infrared camera, an ultrasonic sensor, a millimeter wave radar, and a laser radar (LiDAR). Different sensors may provide different detection accuracy and range.
  • the camera may be installed at the front, rear, or other location of the vehicle.
  • the visual camera may capture the situation inside and outside the vehicle in real time and present it to the driver and/or passengers.
  • information such as traffic light indications, intersection conditions, and the operating status of other vehicles may be obtained.
  • the infrared camera may capture objects under night vision conditions.
  • Ultrasonic sensors may be installed around the vehicle to measure the distance of objects outside the vehicle from the vehicle using the characteristics of strong ultrasonic directionality.
  • Millimeter wave radars may be installed at the front, rear, or other location of the vehicle to measure the distance of objects outside the vehicle from the vehicle using the characteristics of electromagnetic waves.
  • Laser radars may be installed at the front, rear, or other location of the vehicle to detect object edges and shape information, thereby identifying and tracking objects. Due to the Doppler effect, the radar device may also measure the speed change of the vehicle and the moving object.
  • the motor vehicle 110 may also include a communication device 112.
  • the communication device 112 may include a device capable of receiving satellite positioning signals (e.g., Beidou, GPS, GLONASS, and GALILEO) from satellites 141 and generating a signal based on these signals.
  • the communication device 112 may also include a module for communicating with a mobile communication base station 142, and the mobile communication network may implement any suitable communication technology, such as GSM/GPRS, CDMA, LTE and other current or developing wireless communication technologies (such as 5G technology).
  • the communication device 112 may also have a vehicle network or vehicle-to-everything (V2X) module, which is configured to realize, for example, vehicle-to-vehicle (V2V) communication with other vehicles 143 and vehicle-to-infrastructure (V2I) communication with the infrastructure 144.
  • V2X vehicle network or vehicle-to-everything
  • the communication device 112 may also have a module configured to communicate with a user terminal 145 (including but not limited to a smart phone, a tablet computer or a wearable device such as a watch), for example, by using a wireless local area network or Bluetooth of the IEEE802.11 standard.
  • the motor vehicle 110 can also access the server 120 via the network 130.
  • the motor vehicle 110 may also include a control device 113.
  • the control device 113 may include a processor that communicates with various types of computer-readable storage devices or media, such as a central processing unit (CPU) or a graphics processing unit (GPU), or other dedicated processors.
  • the control device 113 may include an automatic driving system for automatically controlling various actuators in the vehicle.
  • the automatic driving system is configured to control the powertrain, steering system, and braking system of the motor vehicle 110 (not shown) via multiple actuators in response to inputs from multiple sensors 111 or other input devices to control acceleration, steering, and braking, respectively, without human intervention or limited human intervention.
  • Some processing functions of the control device 113 can be implemented through cloud computing.
  • control device 113 can be configured to perform a method according to the present disclosure.
  • control device 113 can be implemented as an example of a computing device on the motor vehicle side (client) according to the present disclosure.
  • the system 100 of FIG. 1 may be configured and operated in various ways to enable application of various methods and apparatuses described according to the present disclosure.
  • End-to-end autonomous driving models can continuously achieve better performance based on massive data, but explainability and controllability are bottlenecks in the application of end-to-end autonomous driving models.
  • the present disclosure provides a new autonomous driving model.
  • FIG. 2 shows an exemplary block diagram of an autonomous driving model according to an embodiment of the present disclosure.
  • the autonomous driving model 200 includes an input layer 210 , an encoding layer 220 , an autoregressive inference layer 230 , and a decoding layer 240 .
  • the input layer 210 is configured to receive historical decision information 201 , perception information 202 , traffic information 203 , and interaction information 204 at a current moment.
  • the encoding layer 220 is configured to encode the historical decision information, the perception information, the traffic information and the interaction information to obtain a first tensor, a second tensor, a third tensor and a fourth tensor respectively used to represent the historical decision information, the perception information, the traffic information and the interaction information.
  • the autoregressive inference layer 230 is configured to perform inference on an input tensor formed by the first tensor, the second tensor, the third tensor, and the fourth tensor to obtain a hidden state for the next moment;
  • the decoding layer 240 is configured to decode based on the hidden state at the next moment to obtain the interaction information for the next moment and the autonomous driving decision information for the next moment.
  • the autonomous driving model can understand the current driving environment by reasoning about perception information, traffic information, and interaction information, and can further combine reasoning about historical decision data to better understand the impact of historical operations on the autonomous driving process, thereby making the output results of the autonomous driving model more interpretable and controllable.
  • the input layer 210 is configured to receive historical decision information 201 , perception information 202 , traffic information 203 , and interaction information 204 at a current moment.
  • the historical decision information 201 may include the autonomous driving decision information output by the decoding layer at least one previous moment before the current moment t.
  • the autonomous driving decision information may include information such as a planned trajectory or a control signal for the vehicle (e.g., a signal for controlling the throttle, brake, steering amplitude, etc.). That is, the historical decision information may include a sequence of historical trajectories and/or historical control signals output by the autonomous driving model before the current moment.
  • the historical decision information may include historical decision information at all moments after the start of the autonomous driving process (e.g., for moment t, the historical decision information includes autonomous driving decision information from moment 0 to moment t-1), and may also include historical decision information within a predetermined time period before the current moment t (e.g., autonomous driving decision information from moment t-k to moment t, where k represents a predetermined time range).
  • Perception information 202 may include sensor input collected by at least one sensor installed on the autonomous driving vehicle. Perception information for the vehicle's surroundings may include at least one of the following: perception information from one or more cameras, perception information from one or more lidars, and perception information from one or more millimeter-wave radars. Perception information 202 may include sensor input collected by the sensor at the current moment t, and may also include historical perception information of sensor input collected by the sensor at at least one previous moment before the current moment t.
  • the historical perception information may include historical perception information of all moments since the start of the autonomous driving process (e.g., from moment 0 to moment t-1), and may also include historical perception information within a predetermined time period before the current moment t (e.g., from moment tk to moment t, where k represents a predetermined time range).
  • Traffic information 203 may include at least one of speed limit information, map information, and navigation information of the current route.
  • map information may include lane information, stop line information, traffic light information, etc.
  • traffic information may include lane-level or road-level traffic information.
  • Traffic information 203 may include traffic information at the current time t, and may also include historical traffic information at at least one previous time before the current time t.
  • historical traffic information may include historical traffic information at all times since the start of the autonomous driving process (e.g., from time 0 to time t-1), and may also include historical traffic information within a predetermined time period before the current time t (e.g., from time t-k to time t, where k represents a predetermined time range).
  • the interactive information 204 may include at least one of traffic command information, interactive information from passengers, and interactive information from safety officers.
  • the traffic command information may include actions and/or language from outside the vehicle for traffic command purposes.
  • the interactive information from passengers and the interactive information from safety officers may include actions and/or language collected in the vehicle when the passenger or safety officer is in the vehicle, and may also include actions and/or language used to communicate with the vehicle when the passenger or safety officer is outside the vehicle, and may also include instruction information sent by the passenger or safety officer to the vehicle end through a remote communication device.
  • the interactive information 204 may be information collected by sensors such as cameras and microphones, or may be information received remotely via a communication device.
  • the interactive information 204 may include interactive information obtained at the current time t, and may also include historical interactive information obtained before the current time t.
  • the historical interactive information may include historical interactive information at all times after the start of the automatic driving process (for example, from 0 to t-1), and may also include historical interactive information within a predetermined time period before the current time t (for example, from t-k to t, where k represents a predetermined time range).
  • the encoding layer 220 can be configured to encode historical decision information, perception information, traffic information and interaction information to obtain a first tensor, a second tensor, a third tensor and a fourth tensor for representing the historical decision information, perception information, traffic information and interaction information, respectively, wherein the first tensor, the second tensor, the third tensor and the fourth tensor have the same spatial representation.
  • the encoding layer may include a recurrent neural network or a Transformer network, and is configured to encode historical decision information using the recurrent neural network or the Transformer network.
  • the historical trajectory of the vehicle can be determined based on the historical decision information, and the coordinates of the trajectory points at each historical moment in the vehicle coordinate system at the current time t can be determined.
  • Each trajectory coordinate in the historical trajectory can be input into a recurrent neural network or a Transformer network to obtain a one-dimensional vector or a two-dimensional tensor for representing the coordinates of the trajectory point.
  • the one-dimensional vectors or two-dimensional tensors corresponding to the trajectory point coordinates at each historical moment are stacked in chronological order to obtain a two-dimensional or three-dimensional vector with an added time dimension.
  • the first tensor used to represent historical decision information.
  • the coding layer may also include a coding network for mapping information to a bird's-eye view BEV space, such as a BEVFormer, and is configured to map the perception information to the BEV space to obtain a BEV representation of the perception information.
  • the perception information collected by the sensor at each moment can be input into the BEVFormer network, and a BEV representation of the perception information at that moment can be obtained.
  • the BEV representation can be a three-dimensional vector.
  • the BEV representations of the perception information at each moment can be stacked in chronological order to obtain a four-dimensional tensor with an added time dimension. as the second tensor used to represent the perceptual information.
  • the encoding layer can also be configured to map traffic information to BEV space to obtain a bird's-eye view BEV representation of traffic information.
  • the traffic information at each moment can be vectorized and the vectorized traffic information can be encoded using BEVFormer to obtain the BEV representation of the traffic information at that moment.
  • the BEV representations of the traffic information at each moment can be stacked in chronological order to obtain a four-dimensional tensor with an added time dimension. As the third tensor used to represent traffic information.
  • the BEVFormer for processing perception information and the BEVFormer for processing traffic information can be configured separately according to actual conditions.
  • the encoding layer may further include a pre-trained language model (PLM).
  • PLM pre-trained language model
  • the pre-trained language model can be used to vectorize natural language to convert natural language information into information that can be processed by a machine.
  • the pre-trained language model can be any model that can process the input natural language information.
  • LLM large language model
  • a suitable image recognition algorithm can be used to convert the information in the image into natural language information or vectorized information that can be processed by a machine.
  • the interaction information can be encoded using a pre-trained language model to convert the natural language information into a fourth tensor containing multiple dimensions (such as 2 dimensions)
  • the natural language information may also include timestamp information.
  • the first tensor, the second tensor, the third tensor, and the fourth tensor are the results of output through different encoding methods
  • the above tensors can have the same length and width.
  • the first tensor used to represent historical decision information and the fourth tensor used to represent interaction information are not BEV representations, they can have the same spatial representation as the second tensor and the third tensor represented in the BEV space.
  • the first tensor, the second tensor, the third tensor, and the fourth tensor can be processed uniformly in subsequent model processing, so that the model can uniformly process and reason about different input information when making reasoning decisions, so as to make autonomous driving decisions while considering input information of different modalities.
  • the autoregressive inference layer 230 is configured to perform inference on an input tensor formed of the first tensor, the second tensor, the third tensor, and the fourth tensor to obtain a hidden state for the next moment.
  • the autoregressive reasoning layer 230 can be implemented by a world model.
  • the autoregressive reasoning layer can be implemented by a recurrent neural network (such as a long short-term memory network LSTM, a gated recurrent unit GRU), a Transformer, or a recursive structure with a memory mechanism (such as a recursive memory Transformer network (Recurrent Memory Transformer)) or a diffusion model.
  • a recurrent neural network such as a long short-term memory network LSTM, a gated recurrent unit GRU
  • a Transformer or a recursive structure with a memory mechanism (such as a recursive memory Transformer network (Recurrent Memory Transformer)) or a diffusion model.
  • the combination of the first tensor, the second tensor, the third tensor, and the fourth tensor output by the encoding layer can be used as the input of the autoregressive reasoning layer 230.
  • the first tensor, the second tensor, the third tensor, and the fourth tensor can be flattened into a two-dimensional tensor sequence, and the above two-dimensional tensor sequence can be used as the input of the autoregressive reasoning layer.
  • the memory tensor M 0 at the initial moment can be pre-set.
  • the parameters in M 0 can be randomly initialized parameters.
  • the memory tensor M t at time t can be input into the Transformer layer together with the two-dimensional tensor sequence, so that each vector in the memory tensor is used to process the two-dimensional tensor sequence based on the attention mechanism to obtain the memory tensor M t+1 at the next moment as the hidden state at the next moment.
  • the memory tensor M t is equivalent to the query parameter (Q) of the input Transformer layer
  • the two-dimensional tensor sequence is equivalent to the key (K) and value (V) parameters of the input Transformer layer.
  • the decoding layer 240 is configured to decode based on the hidden state at the next moment to obtain the interaction information for the next moment and the autonomous driving decision information for the next moment.
  • the decoding layer may include a Transformer network.
  • the Transformer may be used to decode the hidden state M t+1 at time t+1 to obtain the autonomous driving decision information at time t+1, such as the control throttle, brake, steering amplitude and other signals at time t+1.
  • the Transformer can also be used to process the hidden state M t+1 at time t+1 to obtain the interactive information output at time t+1.
  • the interactive information output at time t+1 can include a natural language description, which can be used to respond to the interactive information at time t in natural language.
  • the hidden state M t+1 at time t +1 can also be used to generate future prediction information.
  • the future prediction information may include a future prediction image at time t+1, which is used to indicate the obstacle position at time t+1 or the perception information at future times.
  • the hidden state M t+1 represented by the BEV can be spatially mapped to be converted to the sensor (such as a camera) coordinate system. And the M t+1 in the sensor coordinate can be further processed using image diffusion and inverse convolution to obtain the future prediction image at time t+1.
  • the future prediction image can be used to train the autonomous driving model in a self-supervised manner so that the autonomous driving model has accurate future prediction capabilities, thereby improving the accuracy of the decision information output by the autonomous driving model.
  • a method for autonomous driving is also provided.
  • FIG3A shows an exemplary flow chart of an automatic driving method 300 according to an embodiment of the present disclosure.
  • the automatic driving method 300 may be implemented using the automatic driving model 200 described in conjunction with FIG2 .
  • the advantages of the automatic driving model 200 described in conjunction with FIG2 are also applicable to the automatic driving method 300 and will not be described in detail herein.
  • step S302 input information may be obtained, wherein the input information includes historical decision information, perception information, traffic information, and interaction information at the current moment.
  • step S304 historical decision information, perception information, traffic information and interaction information may be encoded to obtain a first tensor, a second tensor, a third tensor and a fourth tensor respectively used to represent the historical decision information, perception information, traffic information and interaction information.
  • autoregressive inference may be performed on the input tensor formed by the first tensor, the second tensor, the third tensor, and the fourth tensor to obtain a hidden state for a first moment after the current moment.
  • step S308 decoding can be performed based on the hidden state at the first moment to obtain the interaction information for the first moment and the automatic driving decision information for the first moment.
  • the interaction information at the first moment may include a signal indicating that the autonomous driving process requires assistance.
  • the autonomous driving decision information output by the autonomous driving model is invalid.
  • the autonomous driving decision information output by the autonomous driving model may cause the autonomous driving vehicle to be unable to continue along the predetermined navigation route.
  • the interactive information output in the autonomous driving method 300 may include a signal for indicating that the autonomous driving process requires assistance.
  • the signal may be a request signal for requesting manual or other models to intervene in the autonomous driving, and the signal may be in the form of natural language, so that the request signal can accurately and completely describe the current situation of the autonomous driving vehicle and is easy to understand.
  • the autonomous driving method 300 may also include sending the interaction information (i.e., assistance request) at the first moment and the current driving state of the autonomous driving vehicle to a remote server, and receiving response information from the remote server, wherein the response information may include prompt interaction information for prompting the autonomous driving process.
  • the response information may include prompt interaction information for prompting the autonomous driving process.
  • a signal indicating that the autonomous driving process requires assistance, the perception information at the first moment, and the natural language description information of the current driving scene of the autonomous driving vehicle at the first moment may be sent to the remote server.
  • the remote server may be a cloud server.
  • the current driving state of the autonomous vehicle may include a natural language description and/or picture of the driving scene at the first moment.
  • the natural language description of the current driving scene may be output by an autonomous driving model deployed on the autonomous vehicle.
  • a large language model may be deployed on the remote server.
  • the current driving state of the autonomous vehicle may be processed by the large language model to obtain response information for the assistance request.
  • the large language model may provide at least one response information for the current driving state of the autonomous vehicle to assist the autonomous driving model. Type. At least one of the above response information can be evaluated, and the response information with the best evaluation result can be sent to the autonomous driving vehicle.
  • a cloud-based autonomous driving model with a larger number of parameters can be deployed on the remote server.
  • the larger number of parameters mentioned here refers to the autonomous driving model deployed at the autonomous driving vehicle.
  • the cloud-based autonomous driving model can handle more complex autonomous driving processes.
  • the cloud-based autonomous driving model can be an autonomous driving model that can process multimodal input information (such as images, videos, languages, etc.).
  • the current driving state of the autonomous driving vehicle may include sensor input collected by the sensor of the autonomous driving vehicle at the second moment, traffic information at the second moment, and other information.
  • the current driving state and/or response information of the autonomous driving vehicle can be input into the cloud-based autonomous driving model with a larger number of parameters to obtain the response information.
  • the response information can be natural language information rather than a control signal for directly controlling the autonomous driving vehicle.
  • the prompt interaction information in the natural language form in the response information can include instructions for guiding the driving of the vehicle.
  • the response information can also include any other form of information that can be processed by the autonomous driving model for guiding the driving of the vehicle.
  • the autonomous driving vehicle can obtain remote driving assistance for the current driving scene.
  • the autonomous driving model can generate autonomous driving decisions according to the instructions given by the remote server, thereby realizing the autonomous driving vehicle's self-driving escape.
  • the autonomous driving vehicle can rely on the intelligence of the autonomous driving model to escape without manual control of the vehicle.
  • the autonomous driving method 300 may further include, at a second moment after receiving the response information, encoding the perception information and response information acquired at the second moment to obtain a fifth tensor and a sixth tensor for representing the perception information and the response information at the second moment; performing autoregressive inference on the input tensor formed by the fifth tensor and the sixth tensor to obtain a hidden state at a third moment after the second moment, and decoding based on the hidden state at the third moment to obtain autonomous driving decision information for the third moment.
  • the autonomous driving model when using the input information at the second moment to infer the autonomous driving decision information at the third moment, can also simultaneously receive the historical decision information at the second moment and the traffic information at the second moment, and use the autonomous driving model described in combination with Figure 2 to obtain the autonomous driving decision information at the third moment.
  • the autonomous driving vehicle uses the response information fed back remotely as part of the interactive information received by the autonomous driving model for reasoning about autonomous driving decisions.
  • the response information output by the model located on the remote server does not need to be directly applied to the autonomous driving vehicle, so the model deployed at the remote server can be a model with general problem-solving capabilities.
  • the autonomous driving model can use the response information as part of the interactive information input during model reasoning to make decisions. Generate appropriate autonomous driving decision information based on the current actual driving situation, the historical trajectory of the current autonomous driving process, and the content of the response information.
  • the historical decision information includes autonomous driving decision information output by the decoding layer at least one previous moment before the current moment.
  • the perception information includes sensor input collected by at least one sensor mounted on the autonomous vehicle.
  • the traffic information includes at least one of speed limit information, map information, and navigation information of the current travel route.
  • the interactive information includes at least one of traffic control information, interactive information from passengers, and interactive information from security personnel.
  • a recurrent neural network or a Transformer network may be used to encode historical decision information.
  • the perceptual information may be mapped to a bird's eye view (BEV) space to obtain a bird's eye view (BEV) representation of the perceptual information.
  • BEV bird's eye view
  • traffic information may be mapped to a bird's eye view BEV space to obtain a bird's eye view BEV representation of the traffic information.
  • the interaction information may be encoded using a pre-trained language model.
  • the autoregressive inference layer can be implemented by a recurrent neural network (such as a long short-term memory network LSTM, a gated recurrent unit GRU), a Transformer, or a recursive structure with a memory mechanism (such as a recursive memory Transformer network (Recurrent Memory Transformer)) or a diffusion model.
  • a recurrent neural network such as a long short-term memory network LSTM, a gated recurrent unit GRU
  • GRU gated recurrent unit
  • Transformer or a recursive structure with a memory mechanism (such as a recursive memory Transformer network (Recurrent Memory Transformer)) or a diffusion model.
  • the prompt information of natural language interaction can be used to guide the autonomous driving model to control the vehicle, so as to realize the rapid generation of escape solutions for the autonomous driving vehicle's escape, intervention and other control.
  • the autonomous driving model can actively initiate an interaction request to request remote assistance when appropriate. Cloud guidance for a certain number (tens of vehicles) of autonomous driving vehicles can be achieved using a remotely deployed model with general capabilities. As a result, users of autonomous driving vehicles only need to enable the autonomous driving model on the vehicle side. Even if difficulties arise during the autonomous driving process, the autonomous driving model can autonomously generate an assistance signal and send it to a remote server to obtain an escape solution without the need for manual control of the vehicle to resolve difficulties during driving.
  • FIG. 3B shows an exemplary process for assisting autonomous driving according to an embodiment of the present disclosure.
  • the autonomous driving model of the autonomous driving vehicle has a problem and outputs an assistance request.
  • the model deployed in the cloud receives the assistance request and/or the current autonomous driving state of the autonomous driving vehicle.
  • three escape plan instructions in natural language are generated according to the assistance request and/or the current autonomous driving state of the autonomous driving vehicle.
  • escape plan one includes: 1) backing up one meter while ensuring safety; 2) trying to turn left to the adjacent lane and quickly leaving when there is no car on the opposite side; 3) returning to the lane.
  • escape plan two and escape plan three include escape instructions with different contents in natural language.
  • escape plan one, escape plan two and escape plan three (such as escape plan one) selected from escape plan one, escape plan two and escape plan three can be included in the response information and sent to the autonomous driving vehicle.
  • the autonomous driving vehicle uses the autonomous driving model described in conjunction with FIG. 2 to input the response information as interactive information into the autonomous driving model, and generates subsequent autonomous driving decisions to achieve autonomous escape.
  • an automatic driving device based on an automatic driving model is provided.
  • FIG4 shows a block diagram of an automatic driving device 400 according to an embodiment of the present disclosure.
  • the automatic driving device 400 includes an acquisition unit 410, an encoding unit 420, an inference unit 430, and a decoding unit 440.
  • the automatic driving device 400 can be implemented based on the automatic driving model 200 described in conjunction with FIG2 .
  • the acquisition unit 410 may be configured to acquire input information, where the input information includes historical decision information, perception information, traffic information, and interaction information at the current moment.
  • the encoding unit 420 can be configured to encode the historical decision information, perception information, traffic information and interaction information to obtain a first tensor, a second tensor, a third tensor and a fourth tensor respectively used to represent the historical decision information, perception information, traffic information and interaction information.
  • the inference unit 430 may be configured to perform inference on an input tensor formed of a first tensor, a second tensor, a third tensor, and a fourth tensor to obtain a hidden state for a first time period after a current time period.
  • the decoding unit 440 may be configured to perform decoding based on the hidden state at the first moment to obtain the interaction information for the first moment and the autonomous driving decision information for the first moment, wherein the interaction information at the first moment includes a signal for indicating that the autonomous driving process requires assistance.
  • the autonomous driving device 400 may further include a communication unit configured to: send the interaction information at the first moment and the current driving state of the autonomous driving vehicle to a remote server; and receive response information from the remote server. For example, a signal indicating that the autonomous driving process requires assistance, the perception information at the first moment, and the natural language description information of the current driving scene of the autonomous driving vehicle at the first moment may be sent to the remote server.
  • the encoding unit is further configured to encode the perception information and the response information acquired at the second moment after receiving the response information to obtain a fifth tensor and a sixth tensor for representing the perception information and the response information at the second moment.
  • the reasoning unit is further configured to perform autoregressive reasoning on the input tensor formed by the fifth tensor and the sixth tensor to obtain a hidden state at a third moment after the second moment.
  • the decoding unit is further configured to decode based on the hidden state at the third moment to obtain autonomous driving decision information for the third moment.
  • the response information is natural language information.
  • the response information is the result of processing the current driving state of the autonomous driving vehicle using a large language model.
  • various techniques may be described herein in the general context of software hardware elements or program modules.
  • the various units described above with respect to FIG. 4 may be implemented in hardware or in hardware in combination with software and/or firmware.
  • these units may be implemented as computer program codes/instructions configured to be executed in one or more processors and stored in a computer-readable storage medium.
  • these units may be implemented as hardware logic/circuits.
  • one or more of units 410 to 440 may be implemented together in a system on chip (SoC).
  • SoC system on chip
  • SoC may include an integrated circuit chip (which includes a processor (e.g., a central processing unit (CPU), a microcontroller, a microprocessor, a digital signal processor (DSP), etc.), a memory, one or more communication interfaces, and/or one or more components in other circuits), and may optionally execute the received program code and/or include embedded firmware to perform functions.
  • a processor e.g., a central processing unit (CPU), a microcontroller, a microprocessor, a digital signal processor (DSP), etc.
  • DSP digital signal processor
  • an electronic device including: at least one processor; and a memory communicatively connected to the at least one processor; the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the autonomous driving method according to an embodiment of the present disclosure.
  • a non-transitory computer-readable storage medium storing computer instructions
  • the computer instructions are used to enable the computer to execute the automatic driving method according to an embodiment of the present disclosure.
  • a computer program product including a computer program, wherein the computer program implements the automatic driving method according to an embodiment of the present disclosure when executed by a processor.
  • an autonomous driving vehicle comprising an autonomous driving device according to an embodiment of the present disclosure and one of the above-mentioned electronic devices.
  • FIG. 5 a block diagram of an electronic device 500 that can be used as a server or client of the present disclosure will now be described, which is an example of a hardware device that can be applied to various aspects of the present disclosure.
  • the electronic device is intended to represent various forms of digital Electronic computer devices, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers.
  • Electronic devices may also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices.
  • the components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and/or claimed herein.
  • the electronic device 500 includes a computing unit 501, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 502 or a computer program loaded from a storage unit 508 into a random access memory (RAM) 503.
  • ROM read-only memory
  • RAM random access memory
  • various programs and data required for the operation of the electronic device 500 can also be stored.
  • the computing unit 501, the ROM 502, and the RAM 503 are connected to each other via a bus 504.
  • An input/output (I/O) interface 505 is also connected to the bus 504.
  • the input unit 506 can be any type of device that can input information to the electronic device 500.
  • the input unit 506 can receive input digital or character information and generate key signal input related to user settings and/or function control of the electronic device, and can include but is not limited to a mouse, a keyboard, a touch screen, a track pad, a track ball, a joystick, a microphone, and/or a remote controller.
  • the output unit 507 can be any type of device that can present information, and can include but is not limited to a display, a speaker, a video/audio output terminal, a vibrator, and/or a printer.
  • the storage unit 508 can include but is not limited to a disk, an optical disk.
  • the communication unit 509 allows the electronic device 500 to exchange information/data with other devices through a computer network such as the Internet and/or various telecommunication networks, and can include but is not limited to a modem, a network card, an infrared communication device, a wireless communication transceiver, and/or a chipset, such as a Bluetooth device, an 802.11 device, a WiFi device, a WiMax device, a cellular communication device, and/or the like.
  • the computing unit 501 may be a variety of general and/or special processing components with processing and computing capabilities. Some examples of the computing unit 501 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc.
  • the computing unit 501 performs the various methods and processes described above, such as the method (or process) 300.
  • the method (or process) 300 may be implemented as a computer software program that is tangibly contained in a machine-readable medium, such as a storage unit 508.
  • part or all of the computer program may be loaded and/or installed on the electronic device 500 via the ROM 502 and/or the communication unit 509.
  • the computer program When the computer program is loaded into the RAM 503 and executed by the computing unit 501, one or more steps of the method (or process) 300 described above may be performed.
  • the computing unit 501 may be configured to execute the method (or process) 300 in any other appropriate manner (eg, by means of firmware).
  • Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and/or combinations thereof.
  • FPGAs field programmable gate arrays
  • ASICs application specific integrated circuits
  • ASSPs application specific standard products
  • SOCs systems on chips
  • CPLDs load programmable logic devices
  • Various implementations can include: being implemented in one or more computer programs that can be executed and/or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.
  • a programmable processor which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.
  • the program code for implementing the method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions/operations specified in the flow chart and/or block diagram to be implemented.
  • the program code may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.
  • a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment.
  • a machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium.
  • a machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing.
  • a more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.
  • RAM random access memory
  • ROM read-only memory
  • EPROM or flash memory erasable programmable read-only memory
  • CD-ROM portable compact disk read-only memory
  • CD-ROM compact disk read-only memory
  • magnetic storage device or any suitable combination of the foregoing.
  • the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer.
  • a display device e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor
  • a keyboard and pointing device e.g., a mouse or trackball
  • Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).
  • the systems and techniques described herein can be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components.
  • the components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), the Internet, and a blockchain network.
  • a computer system may include a client and a server.
  • the client and the server are generally remote from each other and usually interact through a communication network.
  • the relationship of client and server is generated by computer programs running on respective computers and having a client-server relationship with each other.
  • the server may be a cloud server, a server of a distributed system, or a server combined with a blockchain.

Landscapes

  • Engineering & Computer Science (AREA)
  • Automation & Control Theory (AREA)
  • Human Computer Interaction (AREA)
  • Transportation (AREA)
  • Mechanical Engineering (AREA)
  • Traffic Control Systems (AREA)

Abstract

本公开提供了一种能够遵循指令进行自主脱困的自动驾驶方法、装置和车辆，涉及计算机技术领域，尤其涉及自动驾驶和人工智能技术领域。自动驾驶方法，包括：获取输入信息；对输入信息进行编码，以得到与输入信息对应的输入张量；对输入张量进行自回归推理以得到用于当前时刻之后的第一时刻的隐状态；以及基于第一时刻的隐状态进行解码，以得到用于第一时刻的交互信息和用于第一时刻的自动驾驶决策信息。第一时刻的交互信息包括用于指示自动驾驶过程需要协助的信号。由此，可以使用自然语言交互的提示信息指导自动驾驶模型对车辆进行控制，实现车辆脱困方案的快速产生，同时车辆能根据脱困方案指令进行自主脱困。

Description

能够遵循指令实现自主脱困的自动驾驶方法、装置和车辆
相关申请的交叉引用
本申请要求于2023年11月29日提交的中国专利申请202311612484.0的优先权，其全部内容通过引用整体结合在本申请中。
技术领域
本公开涉及计算机技术领域，尤其涉及自动驾驶和人工智能技术领域，具体涉及一种能够实现辅助脱困的自动驾驶方法、装置和车辆。
背景技术
人工智能是研究使计算机来模拟人的某些思维过程和智能行为(如学习、推理、思考、规划等)的学科，既有硬件层面的技术也有软件层面的技术。人工智能硬件技术一般包括如传感器、专用人工智能芯片、云计算、分布式存储、大数据处理等技术；人工智能软件技术主要包括计算机视觉技术、语音识别技术、自然语言处理技术以及机器学习/深度学习、大数据处理技术、知识图谱技术等几大方向。
自动驾驶技术融合了识别、决策、定位、通信安全和人机交互等诸多方面的技术。通过人工智能学习能够辅助生成自动驾驶策略。
在此部分中描述的方法不一定是之前已经设想到或采用的方法。除非另有指明，否则不应假定此部分中描述的任何方法仅因其包括在此部分中就被认为是现有技术。类似地，除非另有指明，否则此部分中提及的问题不应认为在任何现有技术中已被公认。
在此部分中描述的方法不一定是之前已经设想到或采用的方法。除非另有指明，否则不应假定此部分中描述的任何方法仅因其包括在此部分中就被认为是现有技术。类似地，除非另有指明，否则此部分中提及的问题不应认为在任何现有技术中已被公认。
发明内容
本公开提供了一种能够遵循指令实现辅助自主脱困的自动驾驶方法、装置和车辆。
根据本公开的一方面，提供了一种自动驾驶方法，包括：获取输入信息，所述输入信息包括当前时刻的历史决策信息、感知信息、交通信息以及交互信息；对所述历史决 策信息、所述感知信息、所述交通信息以及所述交互信息进行编码，以得到分别用于表示所述历史决策信息、所述感知信息、所述交通信息以及所述交互信息的第一张量、第二张量、第三张量和第四张量；对由所述第一张量、所述第二张量、所述第三张量和所述第四张量形成的输入张量进行自回归推理，以得到用于所述当前时刻之后的第一时刻的隐状态；以及基于所述第一时刻的隐状态进行解码，以得到用于所述第一时刻的交互信息和用于所述第一时刻的自动驾驶决策信息；其中，所述第一时刻的交互信息包括用于指示自动驾驶过程需要协助的信号。
根据本公开的另一方面，提供了一种自动驾驶装置，包括：获取单元，被配置成获取输入信息，所述输入信息包括当前时刻的历史决策信息、感知信息、交通信息以及交互信息；编码单元，被配置成对所述历史决策信息、所述感知信息、所述交通信息以及所述交互信息进行编码，以得到分别用于表示所述历史决策信息、所述感知信息、所述交通信息以及所述交互信息的第一张量、第二张量、第三张量和第四张量；推理单元，被配置成对由所述第一张量、所述第二张量、所述第三张量和所述第四张量形成的输入张量进行推理，以得到用于当前时刻之后的第一时刻的隐状态；解码单元，被配置成基于所述第一时刻的隐状态进行解码，以得到用于所述下一时刻的交互信息和用于所述第一时刻的自动驾驶决策信息，其中，所述第一时刻的交互信息包括用于指示自动驾驶过程需要协助的信号。
根据本公开的另一方面，提供了一种电子设备，包括：至少一个处理器；以及与所述至少一个处理器通信连接的存储器；存储器存储有可被所述至少一个处理器执行的指令，所述指令被所述至少一个处理器执行，以使所述至少一个处理器能够执行上述方法。
根据本公开的另一方面，提供了一种存储有计算机指令的非瞬时计算机可读存储介质，其中，所述计算机指令用于使所述计算机执行上述方法。
根据本公开的另一方面，提供了一种计算机程序产品，包括计算机程序，其中，所述计算机程序在被处理器执行时实现上述方法。
根据本公开的另一方面，提供了一种自动驾驶车辆，包括：根据本公开实施例的自动驾驶装置、电子设备中的一者。
利用本公开的实施例，可以使用自然语言交互的提示信息指导自动驾驶模型对车辆进行控制，实现车辆脱困方案的快速生成，同时车辆能根据脱困方案指令进行自主脱困。自动驾驶模型能够在适当的时候主动发起交互请求以请求远程协助，并可以实现针对多辆自动驾驶车辆的云指导。
应当理解，本部分所描述的内容并非旨在标识本公开的实施例的关键或重要特征，也不用于限制本公开的范围。本公开的其它特征将通过以下的说明书而变得容易理解。
附图说明
附图示例性地示出了实施例并且构成说明书的一部分，与说明书的文字描述一起用于讲解实施例的示例性实施方式。所示出的实施例仅出于例示的目的，并不限制权利要求的范围。在所有附图中，相同的附图标记指代类似但不一定相同的要素。
图1示出了根据本公开的实施例的可以在其中实施本文描述的各种方法的示例性系统的示意图；
图2示出了根据本公开的实施例的自动驾驶模型的示例性框图；
图3A示出了根据本公开的实施例的利用自动驾驶模型实现的自动驾驶方法的示例性流程图；
图3B示出了根据本公开的实施例的用于辅助自动驾驶的示例性过程
图4示出了根据本公开的实施例的自动驾驶装置的结构框图；以及
图5示出了能够用于实现本公开的实施例的示例性电子设备的结构框图。
具体实施方式
以下结合附图对本公开的示范性实施例做出说明，其中包括本公开实施例的各种细节以助于理解，应当将它们认为仅仅是示范性的。因此，本领域普通技术人员应当认识到，可以对这里描述的实施例做出各种改变和修改，而不会背离本公开的范围。同样，为了清楚和简明，以下的描述中省略了对公知功能和结构的描述。
在本公开中，除非另有说明，否则使用术语“第一”、“第二”等来描述各种要素不意图限定这些要素的位置关系、时序关系或重要性关系，这种术语只是用于将一个要素与另一要素区分开。在一些示例中，第一要素和第二要素可以指向该要素的同一实例，而在某些情况下，基于上下文的描述，它们也可以指代不同实例。
在本公开中对各种所述示例的描述中所使用的术语只是为了描述特定示例的目的，而并非旨在进行限制。除非上下文另外明确地表明，如果不特意限定要素的数量，则该要素可以是一个也可以是多个。此外，本公开中所使用的术语“和/或”涵盖所列出的项目中的任何一个以及全部可能的组合方式。
本公开的技术方案中，所涉及的用户个人信息的获取，存储和应用等，均符合相关法律法规的规定，且不违背公序良俗。
下面将结合附图详细描述本公开的实施例。
图1示出了根据本公开的实施例可以将本文描述的各种方法和装置在其中实施的示例性系统100的示意图。参考图1，该系统100包括机动车辆110、服务器120以及将机动车辆110耦接到服务器120的一个或多个通信网络130。
在本公开的实施例中，机动车辆110可以包括根据本公开实施例的计算设备和/或被配置以用于执行根据本公开实施例的方法。
服务器120可以运行使得能够实现自动驾驶的一个或多个服务或软件应用。在某些实施例中，服务器120还可以提供其他服务或软件应用，这些服务或软件应用可以包括非虚拟环境和虚拟环境。在图1所示的配置中，服务器120可以包括实现由服务器120执行的功能的一个或多个组件。这些组件可以包括可由一个或多个处理器执行的软件组件、硬件组件或其组合。机动车辆110的用户可以依次利用一个或多个客户端应用程序来与服务器120进行交互以利用这些组件提供的服务。应当理解，各种不同的系统配置是可能的，其可以与系统100不同。因此，图1是用于实施本文所描述的各种方法的系统的一个示例，并且不旨在进行限制。
服务器120可以包括一个或多个通用计算机、专用服务器计算机(例如PC(个人计算机)服务器、UNIX服务器、中端服务器)、刀片式服务器、大型计算机、服务器群集或任何其他适当的布置和/或组合。服务器120可以包括运行虚拟操作系统的一个或多个虚拟机，或者涉及虚拟化的其他计算架构(例如可以被虚拟化以维护服务器的虚拟存储设备的逻辑存储设备的一个或多个灵活池)。在各种实施例中，服务器120可以运行提供下文所描述的功能的一个或多个服务或软件应用。
服务器120中的计算单元可以运行包括上述任何操作系统以及任何商业上可用的服务器操作系统的一个或多个操作系统。服务器120还可以运行各种附加服务器应用程序和/或中间层应用程序中的任何一个，包括HTTP服务器、FTP服务器、CGI服务器、JAVA服务器、数据库服务器等。
在一些实施方式中，服务器120可以包括一个或多个应用程序，以分析和合并从机动车辆110接收的数据馈送和/或事件更新。服务器120还可以包括一个或多个应用程序，以经由机动车辆110的一个或多个显示设备来显示数据馈送和/或实时事件。
网络130可以是本领域技术人员熟知的任何类型的网络，其可以使用多种可用协议中的任何一种(包括但不限于TCP/IP、SNA、IPX等)来支持数据通信。仅作为示例，一个或多个网络130可以是卫星通信网络、局域网(LAN)、基于以太网的网络、令牌环、广域网(WAN)、因特网、虚拟网络、虚拟专用网络(VPN)、内部网、外部网、区块链网络、公共交换电话网(PSTN)、红外网络、无线网络(包括例如蓝牙、WiFi)和/或这些与其他网络的任意组合。
系统100还可以包括一个或多个数据库150。在某些实施例中，这些数据库可以用于存储数据和其他信息。例如，数据库150中的一个或多个可用于存储诸如音频文件和视频文件的信息。数据存储库150可以驻留在各种位置。例如，由服务器120使用的数据存储库可以在服务器120本地，或者可以远离服务器120且可以经由基于网络或专用的连接与服务器120通信。数据存储库150可以是不同的类型。在某些实施例中，由服务器120使用的数据存储库可以是数据库，例如关系数据库。这些数据库中的一个或多个可以响应于命令而存储、更新和检索到数据库以及来自数据库的数据。
在某些实施例中，数据库150中的一个或多个还可以由应用程序使用来存储应用程序数据。由应用程序使用的数据库可以是不同类型的数据库，例如键值存储库，对象存储库或由文件系统支持的常规存储库。
机动车辆110可以包括传感器111用于感知周围环境。传感器111可以包括下列传感器中的一个或多个：视觉摄像头、红外摄像头、超声波传感器、毫米波雷达以及激光雷达(LiDAR)。不同的传感器可以提供不同的检测精度和范围。摄像头可以安装在车辆的前方、后方或其他位置。视觉摄像头可以实时捕获车辆内外的情况并呈现给驾驶员和/或乘客。此外，通过对视觉摄像头捕获的画面进行分析，可以获取诸如交通信号灯指示、交叉路口情况、其他车辆运行状态等信息。红外摄像头可以在夜视情况下捕捉物体。超声波传感器可以安装在车辆的四周，用于利用超声波方向性强等特点来测量车外物体距车辆的距离。毫米波雷达可以安装在车辆的前方、后方或其他位置，用于利用电磁波的特性测量车外物体距车辆的距离。激光雷达可以安装在车辆的前方、后方或其他位置，用于检测物体边缘、形状信息，从而进行物体识别和追踪。由于多普勒效应，雷达装置还可以测量车辆与移动物体的速度变化。
机动车辆110还可以包括通信装置112。通信装置112可以包括能够从卫星141接收卫星定位信号(例如，北斗、GPS、GLONASS以及GALILEO)并且基于这些信号产生 坐标的卫星定位模块。通信装置112还可以包括与移动通信基站142进行通信的模块，移动通信网络可以实施任何适合的通信技术，例如GSM/GPRS、CDMA、LTE等当前或正在不断发展的无线通信技术(例如5G技术)。通信装置112还可以具有车联网或车联万物(Vehicle-to-Everything，V2X)模块，被配置用于实现例如与其它车辆143进行车对车(Vehicle-to-Vehicle，V2V)通信和与基础设施144进行车辆到基础设施(Vehicle-to-Infrastructure，V2I)通信的车与外界的通信。此外，通信装置112还可以具有被配置为例如通过使用IEEE802.11标准的无线局域网或蓝牙与用户终端145(包括但不限于智能手机、平板电脑或诸如手表等可佩戴装置)进行通信的模块。利用通信装置112，机动车辆110还可以经由网络130接入服务器120。
机动车辆110还可以包括控制装置113。控制装置113可以包括与各种类型的计算机可读存储装置或介质通信的处理器，例如中央处理单元(CPU)或图形处理单元(GPU)，或者其他的专用处理器等。控制装置113可以包括用于自动控制车辆中的各种致动器的自动驾驶系统。自动驾驶系统被配置为经由多个致动器响应来自多个传感器111或者其他输入设备的输入而控制机动车辆110(未示出的)动力总成、转向系统以及制动系统等以分别控制加速、转向和制动，而无需人为干预或者有限的人为干预。控制装置113的部分处理功能可以通过云计算实现。例如，可以使用车载处理器执行某一些处理，而同时可以利用云端的计算资源执行其他一些处理。控制装置113可以被配置以执行根据本公开的方法。此外，控制装置113可以被实现为根据本公开的机动车辆侧(客户端)的计算设备的一个示例。
图1的系统100可以以各种方式配置和操作，以使得能够应用根据本公开所描述的各种方法和装置。
端到端无人驾驶模型能够基于海量数据驱动不断达到更好的表现，但是可解释性和可控性是端到端无人驾驶模型应用的瓶颈。
为了提高自动驾驶模型的效果，本公开提供了一种新的自动驾驶模型。
图2示出了根据本公开的实施例的自动驾驶模型的示例性框图。
如图2所示，自动驾驶模型200包括输入层210、编码层220、自回归推理层230以及解码层240。
输入层210被配置成接收当前时刻的历史决策信息201、感知信息202、交通信息203以及交互信息204。
编码层220被配置成对历史决策信息、感知信息、交通信息以及交互信息进行编码，以得到分别用于表示历史决策信息、感知信息、交通信息以及交互信息的第一张量、第二张量、第三张量和第四张量。
自回归推理层230被配置成对由第一张量、第二张量、第三张量和第四张量形成的输入张量进行推理，以得到用于下一时刻的隐状态；
解码层240被配置成基于下一时刻的隐状态进行解码，以得到用于下一时刻的交互信息和用于下一时刻的自动驾驶决策信息。
利用本公开的实施例提供的自动驾驶模型，自动驾驶模型可以通过对感知信息、交通信息和交互信息的推理对当前驾驶环境进行理解，进一步还可以结合对历史决策数据进行推理来更好地理解历史操作对自动驾驶过程的影响，从而使得自动驾驶模型的输出结果具有更好的可解释性和可控性。
以下将详细描述本公开的原理。
输入层210被配置成接收当前时刻的历史决策信息201、感知信息202、交通信息203以及交互信息204。
历史决策信息201可以包括解码层在当前时刻t之前的至少一个先前时刻输出的自动驾驶决策信息。在一些实现方式中，自动驾驶决策信息可以包括规划轨迹或针对车辆的控制信号(例如控制油门、刹车、转向幅度等的信号)等信息。也就是说，历史决策信息可以包括在当前时刻之前的自动驾驶模型输出的历史轨迹和/或历史控制信号的序列。在一些示例中，历史决策信息可以包括从自动驾驶过程开始之后的所有时刻的历史决策信息(例如对于t时刻来说，历史决策信息包括从0至t-1时刻的自动驾驶决策信息)，也可以包括当前时刻t之前预定时间段内的历史决策信息(例如从t-k至t时刻的自动驾驶决策信息，其中k表示预定时间范围)。
感知信息202可以包括由安装在自动驾驶车辆上的至少一个传感器采集的传感器输入。针对车辆周围环境的感知信息可以包括以下项中的至少一项：一个或多个摄像机的感知信息、一个或多个激光雷达的感知信息、以及一个或多个毫米波雷达的感知信息。感知信息202可以包括当前时刻t传感器采集的传感器输入，也可以包括在当前时刻t之前的至少一个先前时刻传感器采集的传感器输入的历史感知信息。在一些示例中，历史感知信息可以包括从自动驾驶过程开始之后的所有时刻的历史感知信息(例如从0至t-1时刻)，也可以包括当前时刻t之前预定时间段内的历史感知信息(例如从t-k至t时刻，其中k表示预定时间范围)。
交通信息203可以包括当前行驶的路线的限速信息、地图信息和导航信息中的至少一种。例如，地图信息可以包括车道信息、停止线信息、红绿灯信息等。在示例中，交通信息可以包括车道级或道路级的交通信息。交通信息203可以包括当前时刻t的交通信息，也可以包括在当前时刻t之前的至少一个先前时刻的历史交通信息。在一些示例中，历史交通信息可以包括从自动驾驶过程开始之后的所有时刻的历史交通信息(例如从0至t-1时刻)，也可以包括当前时刻t之前预定时间段内的历史交通信息(例如从t-k至t时刻，其中k表示预定时间范围)。
交互信息204可以包括交通指挥信息、来自乘客的交互信息和来自安全员的交互信息中的至少一种。其中，交通指挥信息可以包括来自车外的用于交通指挥目的的动作和/或语言。来自乘客的交互信息和来自安全员的交互信息可以包括当乘客或安全员位于车内时在车内采集到的动作和/或语言，也可以包括当乘客或安全员位于车外时与车辆交流使用的动作和/或语言，还可以包括乘客或安全员通过远程通讯设备发送给车辆端的指令信息。交互信息204可以是通过摄像头、麦克风等传感器采集的信息，也可以是经由通讯设备从远程接收的信息。在一些实现方式中，交互信息204可以包括当前时刻t获取的交互信息，也可以包括在当前时刻t之前获取的历史交互信息。历史交互信息可以包括从自动驾驶过程开始之后的所有时刻的历史交互信息(例如从0至t-1时刻)，也可以包括当前时刻t之前预定时间段内的历史交互信息(例如从t-k至t时刻，其中k表示预定时间范围)。
编码层220可以被配置成对历史决策信息、感知信息、交通信息以及交互信息进行编码，以得到分别用于表示历史决策信息、感知信息、交通信息以及交互信息的第一张量、第二张量、第三张量和第四张量，其中，第一张量、第二张量、第三张量和第四张量具有相同的空间表示。
在一些实施例中，编码层可以包括循环神经网络或Transformer网络，并被配置成利用循环神经网络或Transformer网络对历史决策信息进行编码。可以基于历史决策信息确定车辆的历史轨迹，并确定在当前t时刻的车辆坐标系中各个历史时刻的轨迹点坐标。可以将历史轨迹中的各轨迹坐标输入循环神经网络或Transformer网络以得到用于表示轨迹点坐标的一维向量或二维张量。进一步地，将历史各个时刻的轨迹点坐标对应的一维向量或二维张量按照时间顺序堆叠起来，可以得到增加了时间维度的2维或者3维向量作为用于表示历史决策信息的第一张量。
在一些实施例中，编码层还可以包括用于将信息映射到鸟瞰图BEV空间的编码网络，如BEVFormer，并被配置成将感知信息映射到BEV空间，以得到感知信息的BEV表示。可以将每一时刻传感器采集到的感知信息输入BEVFormer网络，并得到该时刻感知信息的BEV表示。在示例中，BEV表示可以是三维向量。在输入信息包括多个时刻的感知信息的情况下，可以将各个时刻的感知信息的BEV表示按照时间顺序堆叠起来，可以得到增加了时间维度的四维张量作为用于表示感知信息的第二张量。
编码层还可以被配置成将交通信息映射到BEV空间，以得到交通信息的鸟瞰图BEV表示。例如，可以对每一时刻的交通信息进行矢量化，并利用BEVFormer对矢量化的交通信息进行编码，以得到该时刻交通信息的BEV表示。在输入信息包括多个时刻的交通信息的情况下，可以将各个时刻的交通信息的BEV表示按照时间顺序堆叠起来，可以得到增加了时间维度的四维张量作为用于表示交通信息的第三张量。
可以理解的是，可以根据实际情况分别配置用于对感知信息进行处理的BEVFormer和用于对交通信息进行处理的BEVFormer。
在一些实施例中，编码层还可以包括预训练语言模型(PLM)。其中，预训练语言模型可以用于对自然语言进行向量化，以将自然语言信息转化为机器能够处理的信息。预训练语言模型可以是任何能够对输入的自然语言信息进行处理的模型。在一些示例中，也可以使用大语言模型(LLM)来实现对自然语言进行向量化。当输入的交互信息包括由图像表示的动作信息时，可以利用合适的图像识别算法将图像中的信息转换为自然语言信息或机器能够处理的向量化信息。可以利用预训练语言模型对交互信息进行编码，以将自然语言信息转换为包含多维(如2维)的第四张量其中，表示自然语言信息的还可以包括时间戳信息。
尽管第一张量、第二张量、第三张量和第四张量是通过不同的编码方式输出的结果，但上述各个张量可以具有相同的长、宽尺寸。例如，用于表示历史决策信息的第一张量和用于表示交互信息的第四张量尽管不是BEV表示，但可以与BEV空间中表示的第二张量和第三张量具有相同的空间表示。利用这种方式，可以在后续的模型处理中对第一张量、第二张量、第三张量和第四张量进行统一处理，从而使得模型再进行推理决策时可以对不同的输入信息进行统一处理和推理，以在同时考虑不同模态的输入信息的情况下进行自动驾驶决策。
自回归推理层230被配置成对由第一张量、第二张量、第三张量和第四张量形成的输入张量进行推理，以得到用于下一时刻的隐状态。
在一些实施例中，自回归推理层230可以由世界模型来实现。在一些实现方式中，自回归推理层可以由循环神经网络(如长短期记忆网络LSTM、门控循环单元GRU)、Transformer,或者带有记忆机制的递归结构(如递归记忆Transformer网络(Recurrent Memory Transformer))或者扩散模型实现。利用自回归的方式，可以利用t时刻的输入信号来预测t+1时刻的输出结果。可以将编码层输出的第一张量、第二张量、第三张量和第四张量的组合作为自回归推理层230的输入。例如，可以将第一张量、第二张量、第三张量和第四张量平铺成二维张量序列，并可以将上述二维张量序列作为自回归推理层的输入。
以递归记忆Transformer网络为例，可以预先设置初始时刻的记忆张量M0。其中，M0中的参数可以是随机初始化的参数。可以将t时刻的记忆张量Mt与二维张量序列一起输入Transformer层，使得利用记忆张量中的每个向量对二维张量序列进行基于注意力机制的处理，以得到下一时刻的记忆张量Mt+1作为下一时刻的隐状态。其中，记忆张量Mt相当于输入Transformer层查询参数(Q)，二维张量序列相当于输入Transformer层的键(K)、值(V)参数。
解码层240被配置成基于下一时刻的隐状态进行解码，以得到用于下一时刻的交互信息和用于下一时刻的自动驾驶决策信息。
在一些实施例中，解码层可以包括Transformer网络。可以利用Transformer对t+1时刻的隐状态Mt+1进行解码，以得到t+1时刻的自动驾驶决策信息，例如t+1时刻的控制油门、刹车、转向幅度等信号。
还可以利用Transformer对t+1时刻的隐状态Mt+1，以得到t+1时刻的交互信息输出。其中t+1时刻的交互信息输出可以包括自然语言描述，可以用于对t时刻的交互信息进行自然语言形式的响应。
还可以利用t+1时刻的隐状态Mt+1生成未来预测信息。未来预测信息可以包括t+1时刻的未来预测图像，用于指示t+1时刻障碍物位置或未来时刻的感知信息。可以对BEV表示的隐状态Mt+1进行空间映射，以转换到传感器(如摄像头)坐标系下。并可以进一步利用图像扩散以及逆卷积对传感器坐标下的Mt+1进行处理，以得到t+1时刻的未来预测图像。未来预测图像可以用于以自监督的方式对自动驾驶模型进行训练，以使得自动驾驶模型的具有准确的未来预测能力，从而提高自动驾驶模型输出的决策信息的准确性。
根据本公开的一个方面，还提供了一种用于自动驾驶方法。
图3A示出了根据本公开的实施例的自动驾驶方法300的示例性流程图。可以利用结合图2描述的自动驾驶模型200来实现自动驾驶方法300。结合图2描述的自动驾驶模型200的优点也同样是用于自动驾驶方法300，在此不再加以赘述。
在步骤S302中，可以获取输入信息。其中，输入信息包括当前时刻的历史决策信息、感知信息、交通信息以及交互信息。
在步骤S304中，可以对历史决策信息、感知信息、交通信息以及交互信息进行编码，以得到分别用于表示历史决策信息、感知信息、交通信息以及交互信息的第一张量、第二张量、第三张量和第四张量。
在步骤S306中，可以对由第一张量、第二张量、第三张量和第四张量形成的输入张量进行自回归推理，以得到用于当前时刻之后的第一时刻的隐状态。
在步骤S308中，可以基于第一时刻的隐状态进行解码，以得到用于第一时刻的交互信息和用于第一时刻的自动驾驶决策信息。
其中，第一时刻的交互信息可以包括用于指示自动驾驶过程需要协助的信号。
在一些情况下，自动驾驶模型输出的自动驾驶决策信息是无效的。例如，自动驾驶模型输出的自动驾驶决策信息可能导致自动驾驶车辆无法按照预定的导航路线继续前进。在这种情况下，自动驾驶方法300中输出的交互信息可以包括用于指示自动驾驶过程需要协助的信号。该信号可以是请求人工或其他模型介入自动驾驶的请求信号，并且该信号可以是自然语言形式的，从而使得请求信号能够准确完整地描述自动驾驶车辆目前的处境，并且是易于被理解的。
在一些实施例中，自动驾驶方法300还可以包括将第一时刻的交互信息(即协助请求)以及自动驾驶车辆的当前驾驶状态发送至远程服务器，并从远程服务器接收响应信息，其中响应信息可以包括用于提示自动驾驶过程的提示交互信息。例如，可以将指示自动驾驶过程需要协助的信号、第一时刻的感知信息以及第一时刻的自动驾驶车辆的当前驾驶场景的自然语言描述信息发送至远程服务器。利用这种方法，可以在自动驾驶模型出现问题的情况下从远程服务器获取解决方案。其中，远程服务器可以是云端服务器。
在一些实现方式中，其中自动驾驶车辆的当前驾驶状态可以包括针对第一时刻的驾驶场景的自然语言描述和/或图片。可以利用自动驾驶车辆上部署的自动驾驶模型输出当前驾驶场景的自然语言描述。远程服务器上可以部署有大语言模型。可以利用大语言模型对自动驾驶车辆的当前驾驶状态进行处理，来得到用于协助请求的响应信息。大语言模型可以针对自动驾驶车辆的当前驾驶状态提供至少一种响应信息用于辅助自动驾驶模 型。可以对上述至少一种响应信息进行评估，并将具有最优评估结果的响应信息发送给自动驾驶车辆。在另一些实现方式中，远程服务器上可以部署有具有更大参数量的云端自动驾驶模型。这里所说的更大参数量是相对于部署在自动驾驶车辆处的自动驾驶模型而言。由此，云端自动驾驶模型能够处理更复杂的自动驾驶过程。云端自动驾驶模型可以是能够对多模态的输入信息(如图像、视频、语言等)进行处理的自动驾驶模型。自动驾驶车辆的当前驾驶状态可以包括自动驾驶车辆的传感器在第二时刻采集的传感器输入、第二时刻的交通信息等信息。可以将自动驾驶车辆的当前驾驶状态和/或响应信息输入具有更大参数量的云端自动驾驶模型来得到响应信息。在一些示例中，响应信息可以是自然语言信息而非是用于直接控制自动驾驶车辆的控制信号。例如，响应信息中的自然语言形式的提示交互信息可以包括指导车辆驾驶的指令。根据实际情况，响应信息还可以包括用于指导车辆驾驶的、能够由自动驾驶模型处理的其他任何形式的信息。利用这种方式，自动驾驶车辆可以获取针对当前驾驶场景的远程驾驶辅助。通过将所接收的响应信息中的指导车辆驾驶的指令和当前时刻的历史决策信息、感知信息、交通信息一起输入自动驾驶模型，自动驾驶模型能够遵循远程服务器给出的指令生成自动驾驶决策，从而实现自动驾驶车辆的自主脱困。在这一过程中，自动驾驶车辆可以依赖于自动驾驶模型的智能而无需人工接管车辆来实现脱困。
在一些实施例中，自动驾驶方法300还可以包括在接收到响应信息之后的第二时刻，对第二时刻获取的感知信息和响应信息进行编码，以得到用于表示第二时刻的感知信息和响应信息的第五张量和第六张量；对第五张量和第六张量形成的输入张量进行自回归推理，以得到第二时刻之后的第三时刻的隐状态，以及基于第三时刻的隐状态进行解码，以得到用于第三时刻的自动驾驶决策信息。
可以理解的是，在利用第二时刻的输入信息对第三时刻的自动驾驶决策信息进行推理时，自动驾驶模型还可以同时接收第二时刻的历史决策信息和第二时刻的交通信息，并利用结合图2描述的自动驾驶模型来得到第三时刻的自动驾驶决策信息。
区别于从远程直接发送控制信号来接管自动驾驶过程，利用本公开的实现方式，自动驾驶车辆将远程反馈的响应信息作为自动驾驶模型接收的交互信息的一部分用于自动驾驶决策的推理。利用这种方式，位于远程服务器的模型输出的响应信息不需要被直接应用于自动驾驶车辆，因此远程服务器处部署的模型可以是具有通用问题解决能力的模型。通过使用具有通用问题解决能力的模型，可以在生成响应信息时考虑更全面的参考信息。而自动驾驶模型通过在模型推理时将响应信息作为交互信息输入的一部分，可以 结合当前的实际驾驶情况、当前自动驾驶过程的历史轨迹以及响应信息的内容生成合适的自动驾驶决策信息。
在一些实施例中，历史决策信息包括解码层在当前时刻之前的至少一个先前时刻输出的自动驾驶决策信息。
在一些实施例中，感知信息包括由安装在自动驾驶车辆上的至少一个传感器采集的传感器输入。
在一些实施例中，交通信息包括当前行驶的路线的限速信息、地图信息和导航信息中的至少一种。
在一些实施例中，交互信息包括交通指挥信息、来自乘客的交互信息和来自安全员的交互信息中的至少一种。
在一些实施例中，可以利用循环神经网络或Transformer网络对历史决策信息进行编码。
在一些实施例中，可以将感知信息映射到鸟瞰图BEV空间，以得到感知信息的鸟瞰图BEV表示。
在一些实施例中，可以将交通信息映射到鸟瞰图BEV空间，以得到交通信息的鸟瞰图BEV表示。
在一些实施例中，可以利用预训练语言模型对交互信息进行编码。
在一些实施例中，自回归推理层可以由循环神经网络(如长短期记忆网络LSTM、门控循环单元GRU)、Transformer,或者带有记忆机制的递归结构(如递归记忆Transformer网络(Recurrent Memory Transformer))或者扩散模型实现。
利用本公开的实施例提供的自动驾驶方法，可以使用自然语言交互的提示信息指导自动驾驶模型对车辆进行控制，实现针对自动驾驶车辆的脱困、干预等控制的脱困方案的快速生成。自动驾驶模型能够在适当的时候主动发起交互请求以请求远程协助。利用具有通用能力的远程部署的模型可以实现针对一定数量(数十辆)自动驾驶车辆的云指导。由此，自动驾驶车辆的用户仅需要在车辆端启用自动驾驶模型即可。即使在自动驾驶过程中出现困难，自动驾驶模型也能够自主产生协助信号，并发送至远程服务器来获取脱困方案，而无需人为控制操作车辆来解决驾驶过程中的困难。
图3B示出了根据本公开的实施例的用于辅助自动驾驶的示例性过程。
如图3B所示，在框301中，自动驾驶车辆的自动驾驶模型出现问题，并输出协助请求。在框303中，云端部署的模型接收协助请求和/或自动驾驶车辆的当前自动驾驶状态。 在框305中，并根据协助请求和/或自动驾驶车辆的当前自动驾驶状态生成三个自然语言形式的脱困方案的指令。其中，脱困方案一包括：1)在保证安全的情况下往后倒车一米；2)尝试左拐至临近车道，并在对向无车的情况下快速驶离；3)并回车道。类似地，脱困方案二、脱困方案三包括自然语言形式的不同内容的脱困指令。可以在响应信息中包括从脱困方案一、脱困方案二以及脱困方案三中选择的一个(如脱困方案一)发送给自动驾驶车辆。在框307中，自动驾驶车辆利用结合图2描述的自动驾驶模型，将响应信息作为交互信息输入自动驾驶模型，并生成后续的自动驾驶决策来实现自主脱困。
根据本公开的另一方面，提供了一种基于自动驾驶模型的自动驾驶装置。
图4示出了根据本公开的实施例的自动驾驶装置400的结构框图。如图4所示，自动驾驶装置400包括获取单元410、编码单元420、推理单元430以及解码单元440。可以基于结合图2描述的自动驾驶模型200实现自动驾驶装置400。
获取单元410可以被配置成获取输入信息，输入信息包括当前时刻的历史决策信息、感知信息、交通信息以及交互信息。
编码单元420可以被配置成对历史决策信息、感知信息、交通信息以及交互信息进行编码，以得到分别用于表示历史决策信息、感知信息、交通信息以及交互信息的第一张量、第二张量、第三张量和第四张量。
推理单元430可以被配置成对由第一张量、第二张量、第三张量和第四张量形成的输入张量进行推理，以得到用于当前时刻之后的第一时刻的隐状态。
解码单元440可以被配置成基于第一时刻的隐状态进行解码，以得到用于第一时刻的交互信息和用于第一时刻的自动驾驶决策信息。其中，第一时刻的交互信息包括用于指示自动驾驶过程需要协助的信号。
在一些实施例中，自动驾驶装置400还可以包括通讯单元，被配置成：将第一时刻的交互信息以及自动驾驶车辆的当前驾驶状态发送至远程服务器；以及从远程服务器接收响应信息。例如，可以将指示自动驾驶过程需要协助的信号、第一时刻的感知信息以及第一时刻的自动驾驶车辆的当前驾驶场景的自然语言描述信息发送至远程服务器。
在一些实施例中，编码单元还配置成在接收到响应信息之后的第二时刻，对第二时刻获取的感知信息和所述响应信息进行编码，以得到用于表示第二时刻的感知信息和所述响应信息的第五张量和第六张量。推理单元还被配置成对第五张量和第六张量形成的输入张量进行自回归推理，以得到第二时刻之后的第三时刻的隐状态。解码单元还被配置成基于第三时刻的隐状态进行解码，以得到用于第三时刻的自动驾驶决策信息。
在一些实施例中，响应信息是自然语言信息。
在一些实施例中，响应信息是利用大语言模型对自动驾驶车辆的当前驾驶状态进行处理得到的结果。
应当理解，图4中所示装置400的各个模块或单元可以与参考图3A描述的方法300中的各个步骤相对应。由此，上面针对方法300描述的操作、特征和优点同样适用于装置400及其包括的模块以及单元。为了简洁起见，某些操作、特征和优点在此不再赘述。
虽然上面参考特定模块讨论了特定功能，但是应当注意，本文讨论的各个单元的功能可以分为多个单元，和/或多个单元的至少一些功能可以组合成单个单元。
还应当理解，本文可以在软件硬件元件或程序模块的一般上下文中描述各种技术。上面关于图4描述的各个单元可以在硬件中或在结合软件和/或固件的硬件中实现。例如，这些单元可以被实现为计算机程序代码/指令，该计算机程序代码/指令被配置为在一个或多个处理器中执行并存储在计算机可读存储介质中。可替换地，这些单元可以被实现为硬件逻辑/电路。例如，在一些实施例中，单元410至440中的一个或多个可以一起被实现在片上系统(System on Chip,SoC)中。SoC可以包括集成电路芯片(其包括处理器(例如，中央处理单元(Central Processing Unit,CPU)、微控制器、微处理器、数字信号处理器(Digital Signal Processor,DSP)等)、存储器、一个或多个通信接口、和/或其他电路中的一个或多个部件)，并且可以可选地执行所接收的程序代码和/或包括嵌入式固件以执行功能。
根据本公开的另一方面，还提供了一种电子设备，包括：至少一个处理器；以及与所述至少一个处理器通信连接的存储器；所述存储器存储有可被所述至少一个处理器执行的指令，所述指令被所述至少一个处理器执行，以使所述至少一个处理器能够执行根据本公开实施例的自动驾驶方法。
根据本公开的另一方面，还提供了一种存储有计算机指令的非瞬时计算机可读存储介质，计算机指令用于使所述计算机执行根据本公开实施例的自动驾驶方法。
根据本公开的另一方面，还提供了一种计算机程序产品，包括计算机程序，其中，计算机程序在被处理器执行时实现根据本公开实施例的自动驾驶方法。
根据本公开的另一方面，还提供了一种自动驾驶车辆，包括根据本公开实施例的自动驾驶装置以及上述的电子设备中的一者。
参考图5，现将描述可以作为本公开的服务器或客户端的电子设备500的结构框图，其是可以应用于本公开的各方面的硬件设备的示例。电子设备旨在表示各种形式的数字 电子的计算机设备，诸如，膝上型计算机、台式计算机、工作台、个人数字助理、服务器、刀片式服务器、大型计算机、和其它适合的计算机。电子设备还可以表示各种形式的移动装置，诸如，个人数字处理、蜂窝电话、智能电话、可穿戴设备和其它类似的计算装置。本文所示的部件、它们的连接和关系、以及它们的功能仅仅作为示例，并且不意在限制本文中描述的和/或者要求的本公开的实现。
如图5所示，电子设备500包括计算单元501，其可以根据存储在只读存储器(ROM)502中的计算机程序或者从存储单元508加载到随机访问存储器(RAM)503中的计算机程序，来执行各种适当的动作和处理。在RAM 503中，还可存储电子设备500操作所需的各种程序和数据。计算单元501、ROM 502以及RAM 503通过总线504彼此相连。输入/输出(I/O)接口505也连接至总线504。
电子设备500中的多个部件连接至I/O接口505，包括：输入单元506、输出单元507、存储单元508以及通信单元509。输入单元506可以是能向电子设备500输入信息的任何类型的设备，输入单元506可以接收输入的数字或字符信息，以及产生与电子设备的用户设置和/或功能控制有关的键信号输入，并且可以包括但不限于鼠标、键盘、触摸屏、轨迹板、轨迹球、操作杆、麦克风和/或遥控器。输出单元507可以是能呈现信息的任何类型的设备，并且可以包括但不限于显示器、扬声器、视频/音频输出终端、振动器和/或打印机。存储单元508可以包括但不限于磁盘、光盘。通信单元509允许电子设备500通过诸如因特网的计算机网络和/或各种电信网络与其他设备交换信息/数据，并且可以包括但不限于调制解调器、网卡、红外通信设备、无线通信收发机和/或芯片组，例如蓝牙设备、802.11设备、WiFi设备、WiMax设备、蜂窝通信设备和/或类似物。
计算单元501可以是各种具有处理和计算能力的通用和/或专用处理组件。计算单元501的一些示例包括但不限于中央处理单元(CPU)、图形处理单元(GPU)、各种专用的人工智能(AI)计算芯片、各种运行机器学习模型算法的计算单元、数字信号处理器(DSP)、以及任何适当的处理器、控制器、微控制器等。计算单元501执行上文所描述的各个方法和处理，例如方法(或过程)300。例如，在一些实施例中，方法(或过程)300可被实现为计算机软件程序，其被有形地包含于机器可读介质，例如存储单元508。在一些实施例中，计算机程序的部分或者全部可以经由ROM 502和/或通信单元509而被载入和/或安装到电子设备500上。当计算机程序加载到RAM 503并由计算单元501执行时，可以执行上文描述的方法(或过程)300的一个或多个步骤。备选地，在其他实 施例中，计算单元501可以通过其他任何适当的方式(例如，借助于固件)而被配置为执行方法(或过程)300。
本文中以上描述的系统和技术的各种实施方式可以在数字电子电路系统、集成电路系统、场可编程门阵列(FPGA)、专用集成电路(ASIC)、专用标准产品(ASSP)、芯片上系统的系统(SOC)、负载可编程逻辑设备(CPLD)、计算机硬件、固件、软件、和/或它们的组合中实现。这些各种实施方式可以包括：实施在一个或者多个计算机程序中，该一个或者多个计算机程序可在包括至少一个可编程处理器的可编程系统上执行和/或解释，该可编程处理器可以是专用或者通用可编程处理器，可以从存储系统、至少一个输入装置、和至少一个输出装置接收数据和指令，并且将数据和指令传输至该存储系统、该至少一个输入装置、和该至少一个输出装置。
用于实施本公开的方法的程序代码可以采用一个或多个编程语言的任何组合来编写。这些程序代码可以提供给通用计算机、专用计算机或其他可编程数据处理装置的处理器或控制器，使得程序代码当由处理器或控制器执行时使流程图和/或框图中所规定的功能/操作被实施。程序代码可以完全在机器上执行、部分地在机器上执行，作为独立软件包部分地在机器上执行且部分地在远程机器上执行或完全在远程机器或服务器上执行。
在本公开的上下文中，机器可读介质可以是有形的介质，其可以包含或存储以供指令执行系统、装置或设备使用或与指令执行系统、装置或设备结合地使用的程序。机器可读介质可以是机器可读信号介质或机器可读储存介质。机器可读介质可以包括但不限于电子的、磁性的、光学的、电磁的、红外的、或半导体系统、装置或设备，或者上述内容的任何合适组合。机器可读存储介质的更具体示例会包括基于一个或多个线的电气连接、便携式计算机盘、硬盘、随机存取存储器(RAM)、只读存储器(ROM)、可擦除可编程只读存储器(EPROM或快闪存储器)、光纤、便捷式紧凑盘只读存储器(CD-ROM)、光学储存设备、磁储存设备、或上述内容的任何合适组合。
为了提供与用户的交互，可以在计算机上实施此处描述的系统和技术，该计算机具有：用于向用户显示信息的显示装置(例如，CRT(阴极射线管)或者LCD(液晶显示器)监视器)；以及键盘和指向装置(例如，鼠标或者轨迹球)，用户可以通过该键盘和该指向装置来将输入提供给计算机。其它种类的装置还可以用于提供与用户的交互；例如，提供给用户的反馈可以是任何形式的传感反馈(例如，视觉反馈、听觉反馈、或者触觉反馈)；并且可以用任何形式(包括声输入、语音输入或者、触觉输入)来接收来自用户的输入。
可以将此处描述的系统和技术实施在包括后台部件的计算系统(例如，作为数据服务器)、或者包括中间件部件的计算系统(例如，应用服务器)、或者包括前端部件的计算系统(例如，具有图形用户界面或者网络浏览器的用户计算机，用户可以通过该图形用户界面或者该网络浏览器来与此处描述的系统和技术的实施方式交互)、或者包括这种后台部件、中间件部件、或者前端部件的任何组合的计算系统中。可以通过任何形式或者介质的数字数据通信(例如，通信网络)来将系统的部件相互连接。通信网络的示例包括：局域网(LAN)、广域网(WAN)、互联网和区块链网络。
计算机系统可以包括客户端和服务器。客户端和服务器一般远离彼此并且通常通过通信网络进行交互。通过在相应的计算机上运行并且彼此具有客户端-服务器关系的计算机程序来产生客户端和服务器的关系。服务器可以是云服务器，也可以为分布式系统的服务器，或者是结合了区块链的服务器。
应该理解，可以使用上面所示的各种形式的流程，重新排序、增加或删除步骤。例如，本公开中记载的各步骤可以并行地执行、也可以顺序地或以不同的次序执行，只要能够实现本公开公开的技术方案所期望的结果，本文在此不进行限制。
虽然已经参照附图描述了本公开的实施例或示例，但应理解，上述的方法、系统和设备仅仅是示例性的实施例或示例，本发明的范围并不由这些实施例或示例限制，而是仅由授权后的权利要求书及其等同范围来限定。实施例或示例中的各种要素可以被省略或者可由其等同要素替代。此外，可以通过不同于本公开中描述的次序来执行各步骤。进一步地，可以以各种方式组合实施例或示例中的各种要素。重要的是随着技术的演进，在此描述的很多要素可以由本公开之后出现的等同要素进行替换。

Claims (19)

  1. 一种自主脱困的自动驾驶方法，包括：
    获取输入信息，所述输入信息包括当前时刻的历史决策信息、感知信息、交通信息以及交互信息；
    对所述历史决策信息、所述感知信息、所述交通信息以及所述交互信息进行编码，以得到分别用于表示所述历史决策信息、所述感知信息、所述交通信息以及所述交互信息的第一张量、第二张量、第三张量和第四张量；
    对由所述第一张量、所述第二张量、所述第三张量和所述第四张量形成的输入张量进行自回归推理，以得到用于所述当前时刻之后的第一时刻的隐状态；以及
    基于所述第一时刻的隐状态进行解码，以得到用于所述第一时刻的交互信息和用于所述第一时刻的自动驾驶决策信息；
    其中，所述第一时刻的交互信息包括用于指示自动驾驶过程需要协助的信号。
  2. 如权利要求1所述的自动驾驶方法，还包括：
    将所述指示自动驾驶过程需要协助的信号、第一时刻的感知信息以及第一时刻的自动驾驶车辆的当前驾驶场景的自然语言描述信息发送至远程服务器；以及
    从所述远程服务器接收响应信息。
  3. 如权利要求2所述的自动驾驶方法，还包括：
    在接收到所述响应信息之后的第二时刻，对第二时刻获取的感知信息和所述响应信息进行编码，以得到用于表示第二时刻的感知信息和所述响应信息的第五张量和第六张量；
    对所述第五张量和所述第六张量形成的输入张量进行自回归推理，以得到所述第二时刻之后的第三时刻的隐状态；
    基于所述第三时刻的隐状态进行解码，以得到用于所述第三时刻的自动驾驶决策信息。
  4. 如权利要求2所述的自动驾驶方法，其中所述响应信息是自然语言信息。
  5. 如权利要求4所述的自动驾驶方法，其中，所述响应信息是利用大语言模型对自动驾驶车辆的当前驾驶状态进行处理得到的结果。
  6. 如权利要求1-5中任一项所述的自动驾驶方法，其中，对所述历史决策信息进行编码包括：利用循环神经网络或Transformer网络对所述历史决策信息进行编码。
  7. 如权利要求6所述的自动驾驶方法，其中，所述历史决策信息包括所述解码层在所述当前时刻之前的至少一个先前时刻输出的自动驾驶决策信息。
  8. 如权利要求1-7中任一项所述的自动驾驶方法，其中，对所述感知信息进行编码包括：将所述感知信息映射到鸟瞰图BEV空间，以得到所述感知信息的鸟瞰图BEV表示。
  9. 如权利要求8所述的自动驾驶方法，其中所述感知信息包括由安装在自动驾驶车辆上的至少一个传感器采集的传感器输入。
  10. 如权利要求1-9中任一项所述的自动驾驶方法，其中，对所述交通信息进行编码包括：将所述交通信息映射到鸟瞰图BEV空间，以得到所述交通信息的鸟瞰图BEV表示。
  11. 如权利要求10所述的自动驾驶方法，其中，所述交通信息包括当前行驶的路线的限速信息、地图信息和导航信息中的至少一种。
  12. 如权利要求1-11中任一项所述的自动驾驶方法，其中，对所述交互信息进行编码包括：利用预训练语言模型对所述交互信息进行编码。
  13. 如权利要求12所述的自动驾驶方法，其中，所述交互信息包括交通指挥信息、来自乘客的交互信息和来自安全员的交互信息中的至少一种。
  14. 如权利要求1至13中任一项所述的自动驾驶方法，其中，对所述输入张量进行自回归推理包括：利用循环神经网络、Transformer、递归记忆Transformer、以及扩散模型中的一种实现对所述输入张量进行自回归推理。
  15. 一种自动驾驶装置，包括：
    获取单元，被配置成获取输入信息，所述输入信息包括当前时刻的历史决策信息、感知信息、交通信息以及交互信息；
    编码单元，被配置成对所述历史决策信息、所述感知信息、所述交通信息以及所述交互信息进行编码，以得到分别用于表示所述历史决策信息、所述感知信息、所述交通信息以及所述交互信息的第一张量、第二张量、第三张量和第四张量；
    推理单元，被配置成对由所述第一张量、所述第二张量、所述第三张量和所述第四张量形成的输入张量进行推理，以得到用于当前时刻之后的第一时刻的隐状态；以及
    解码单元，被配置成基于所述第一时刻的隐状态进行解码，以得到用于所述下一时刻的交互信息和用于所述第一时刻的自动驾驶决策信息，
    其中，所述第一时刻的交互信息包括用于指示自动驾驶过程需要协助的信号。
  16. 一种电子设备，包括：
    至少一个处理器；以及
    与所述至少一个处理器通信连接的存储器；其中
    所述存储器存储有可被所述至少一个处理器执行的指令，所述指令被所述至少一个处理器执行，以使所述至少一个处理器能够执行权利要求1-14中任一项所述的方法。
  17. 一种存储有计算机指令的非瞬时计算机可读存储介质，其中，所述计算机指令用于使所述计算机执行根据权利要求1-14中任一项所述的方法。
  18. 一种计算机程序产品，包括计算机程序，其中，所述计算机程序在被处理器执行时实现权利要求1-14中任一项所述的方法。
  19. 一种自动驾驶车辆，包括：
    根据权利要求15所述的自动驾驶装置、根据权利要求16所述的电子设备中的一者。
PCT/CN2024/099367 2023-11-29 2024-06-14 能够遵循指令实现自主脱困的自动驾驶方法、装置和车辆 Pending WO2025112451A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202311612484.0A CN117539253B (zh) 2023-11-29 2023-11-29 能够遵循指令实现自主脱困的自动驾驶方法、装置和车辆
CN202311612484.0 2023-11-29

Publications (1)

Publication Number Publication Date
WO2025112451A1 true WO2025112451A1 (zh) 2025-06-05

Family

ID=89782133

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2024/099367 Pending WO2025112451A1 (zh) 2023-11-29 2024-06-14 能够遵循指令实现自主脱困的自动驾驶方法、装置和车辆

Country Status (2)

Country Link
CN (1) CN117539253B (zh)
WO (1) WO2025112451A1 (zh)

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117539253B (zh) * 2023-11-29 2025-03-14 北京百度网讯科技有限公司 能够遵循指令实现自主脱困的自动驾驶方法、装置和车辆
CN117601898A (zh) * 2023-11-29 2024-02-27 北京百度网讯科技有限公司 能够实现多模态交互的自动驾驶模型、方法、装置和车辆
CN118850117B (zh) * 2024-09-25 2025-01-14 浙江吉利控股集团有限公司 一种自动驾驶的脱困方法、装置、电子设备、介质及产品

Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20210150225A1 (en) * 2019-11-15 2021-05-20 Honda Motor Co., Ltd. System and method for providing an interpretable and unified representation for trajectory prediction
US20210302992A1 (en) * 2020-03-25 2021-09-30 Mitsubishi Electric Research Laboratories, Inc. Bird's Eye View Map Based Recognition and Motion Prediction for Autonomous Systems
CN114194211A (zh) * 2021-11-30 2022-03-18 浪潮(北京)电子信息产业有限公司 一种自动驾驶方法、装置及电子设备和存储介质
WO2022214414A1 (de) * 2021-04-07 2022-10-13 Zf Friedrichshafen Ag Computerimplementiertes verfahren, computerprogramm und anordnung zum vorhersagen und planen von trajektorien
CN115563989A (zh) * 2022-08-31 2023-01-03 西安工程大学 基于自适应上下文推理机制的内容倾向性评判及预测方法
JP2023079022A (ja) * 2021-11-26 2023-06-07 株式会社Preferred Networks 情報処理装置及び情報生成方法
CN116859724A (zh) * 2023-06-21 2023-10-10 北京百度网讯科技有限公司 时序自回归同时决策和预测的自动驾驶模型及其训练方法
CN117010265A (zh) * 2023-04-14 2023-11-07 北京百度网讯科技有限公司 能够进行自然语言交互的自动驾驶模型及其训练方法
CN117539253A (zh) * 2023-11-29 2024-02-09 北京百度网讯科技有限公司 能够遵循指令实现自主脱困的自动驾驶方法、装置和车辆
CN117601898A (zh) * 2023-11-29 2024-02-27 北京百度网讯科技有限公司 能够实现多模态交互的自动驾驶模型、方法、装置和车辆

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113261035B (zh) * 2019-12-30 2022-09-16 华为技术有限公司 一种轨迹预测方法及相关设备
CN116991157B (zh) * 2023-04-14 2024-09-10 北京百度网讯科技有限公司 具备人类专家驾驶能力的自动驾驶模型、训练方法和车辆
CN116776151B (zh) * 2023-06-21 2025-12-26 北京百度网讯科技有限公司 可与车外人员进行自主交互的自动驾驶模型和训练方法

Patent Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20210150225A1 (en) * 2019-11-15 2021-05-20 Honda Motor Co., Ltd. System and method for providing an interpretable and unified representation for trajectory prediction
US20210302992A1 (en) * 2020-03-25 2021-09-30 Mitsubishi Electric Research Laboratories, Inc. Bird's Eye View Map Based Recognition and Motion Prediction for Autonomous Systems
WO2022214414A1 (de) * 2021-04-07 2022-10-13 Zf Friedrichshafen Ag Computerimplementiertes verfahren, computerprogramm und anordnung zum vorhersagen und planen von trajektorien
JP2023079022A (ja) * 2021-11-26 2023-06-07 株式会社Preferred Networks 情報処理装置及び情報生成方法
CN114194211A (zh) * 2021-11-30 2022-03-18 浪潮(北京)电子信息产业有限公司 一种自动驾驶方法、装置及电子设备和存储介质
CN115563989A (zh) * 2022-08-31 2023-01-03 西安工程大学 基于自适应上下文推理机制的内容倾向性评判及预测方法
CN117010265A (zh) * 2023-04-14 2023-11-07 北京百度网讯科技有限公司 能够进行自然语言交互的自动驾驶模型及其训练方法
CN116859724A (zh) * 2023-06-21 2023-10-10 北京百度网讯科技有限公司 时序自回归同时决策和预测的自动驾驶模型及其训练方法
CN117539253A (zh) * 2023-11-29 2024-02-09 北京百度网讯科技有限公司 能够遵循指令实现自主脱困的自动驾驶方法、装置和车辆
CN117601898A (zh) * 2023-11-29 2024-02-27 北京百度网讯科技有限公司 能够实现多模态交互的自动驾驶模型、方法、装置和车辆

Also Published As

Publication number Publication date
CN117539253B (zh) 2025-03-14
CN117539253A (zh) 2024-02-09

Similar Documents

Publication Publication Date Title
CN116880462B (zh) 自动驾驶模型、训练方法和自动驾驶方法和车辆
KR102745062B1 (ko) 앵커 궤적들을 이용하는 에이전트 궤적 예측
CN117519206B (zh) 基于生成式扩散模型的自动驾驶模型、方法、装置和车辆
CN117539253B (zh) 能够遵循指令实现自主脱困的自动驾驶方法、装置和车辆
WO2025112453A1 (zh) 能够实现多模态交互的自动驾驶模型、方法、装置和车辆
CN116859724B (zh) 时序自回归同时决策和预测的自动驾驶模型及其训练方法
CN115273002A (zh) 一种图像处理方法、装置、存储介质及计算机程序产品
CN114758502B (zh) 双车联合轨迹预测方法及装置、电子设备和自动驾驶车辆
CN118163808B (zh) 世界知识增强的自动驾驶模型、训练方法和自动驾驶方法
WO2025118534A1 (zh) 基于时序递归自回归推理的自动驾驶模型、方法和车辆
CN116991157B (zh) 具备人类专家驾驶能力的自动驾驶模型、训练方法和车辆
CN115366920A (zh) 用于自动驾驶车辆的决策方法及装置、设备和介质
CN116776151A (zh) 可与车外人员进行自主交互的自动驾驶模型和训练方法
CN117010265B (zh) 能够进行自然语言交互的自动驾驶模型及其训练方法
CN115675528A (zh) 基于相似场景挖掘的自动驾驶方法和车辆
CN117035032A (zh) 融合文本数据和自动驾驶数据进行模型训练的方法和车辆
CN115713738A (zh) 使用神经网络模型的凝视和意识预测
CN116861230A (zh) 输出解释信息的自动驾驶模型、训练方法、装置和车辆
US20260116406A1 (en) Autonomous driving method and electronic device
CN115082690B (zh) 目标识别方法、目标识别模型训练方法及装置
CN115235487A (zh) 数据处理方法及装置、设备和介质
US20240425085A1 (en) Method for content generation
CN118657044A (zh) 对自动驾驶模型进行训练的方法、装置和电子设备
CN118551806A (zh) 基于状态节点预测的自动驾驶模型、自动驾驶方法和装置
CN116469069A (zh) 用于自动驾驶的场景编码模型训练方法、装置及介质

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 24895554

Country of ref document: EP

Kind code of ref document: A1