Method, device and electronic device for training autonomous driving model

Abstract

The present disclosure provides a method, device, and electronic device for training an autonomous driving model, in the field of computer technology, and particularly in the field of autonomous driving and artificial intelligence technology. The implementation scheme comprises: determining a training data set, the training data set including a plurality of sample environmental data; determining a plurality of coaching models, wherein the coaching models output coaching driving behaviors as target driving behaviors of the trained autonomous driving model based on the sample environmental data, wherein each coaching model is capable of processing at least a portion of the sample environmental data in the training data set; processing the sample environmental data in the training data set using the coaching model to obtain coaching driving behaviors; processing the sample environmental data in the training data set using the autonomous driving model to obtain predicted driving behaviors; and training the autonomous driving model by adjusting parameters of the autonomous driving model, wherein the parameters are adjusted to reduce the difference between the predicted driving behaviors and the coaching driving behaviors.

Description

Technical Field

[0001] The present disclosure relates to the fields of computer technology, in particular to the fields of autonomous driving and artificial intelligence technology, and specifically to a method, device, electronic device, computer-readable storage medium, and computer program product for training an autonomous driving model. Background Art

[0002] Artificial intelligence (AI) is the study of how computers can simulate certain human thought processes and intelligent behaviors (such as learning, reasoning, thinking, and planning). It encompasses both hardware and software technologies. AI hardware technologies generally include sensors, specialized AI chips, cloud computing, distributed storage, and big data processing. AI software technologies primarily encompass computer vision, speech recognition, natural language processing, machine learning / deep learning, big data processing, and knowledge graphs.

[0003] When training the autonomous driving model, a combination of imitation learning and reinforcement learning is used to improve the effect of model training.

[0004] The approaches described in this section are not necessarily approaches that have been previously conceived or employed. Unless otherwise indicated, it should not be assumed that any approach described in this section is prior art simply by virtue of its inclusion in this section. Similarly, unless otherwise indicated, the issues raised in this section should not be considered as having been recognized in any prior art. Summary of the Invention

[0005] The present disclosure provides a method, apparatus, electronic device, computer-readable storage medium, and computer program product for training an autonomous driving model.

[0006] According to one aspect of the present disclosure, a method for training an autonomous driving model is provided, comprising: determining a training data set, the training data set comprising a plurality of sample environmental data; determining a plurality of coach models, wherein the coach models output coach driving behaviors as target driving behaviors of the trained autonomous driving model based on the sample environmental data, wherein each coach model is capable of processing at least a portion of the sample environmental data in the training data set; processing the sample environmental data in the training data set using the coach models to obtain coach driving behaviors; processing the sample environmental data in the training data set using the autonomous driving model to obtain predicted driving behaviors; and training the autonomous driving model by adjusting parameters of the autonomous driving model, wherein the parameters are adjusted to reduce the difference between the predicted driving behaviors and the coach driving behaviors.

[0007] According to another aspect of the present disclosure, a device for training an autonomous driving model is provided, comprising: a training data determination unit configured to determine a training data set, the training data set comprising a plurality of sample environmental data; a coach model determination unit configured to determine a plurality of coach models, wherein the coach model outputs a coach driving behavior as a target driving behavior of the trained autonomous driving model based on the sample environmental data, wherein each coach model is capable of processing at least a portion of the sample environmental data in the training data set; a coach behavior acquisition unit configured to process the sample environmental data in the training data set using the coach model to obtain a coach driving behavior; a prediction unit configured to process the sample environmental data in the training data set using the autonomous driving model to obtain a predicted driving behavior; and a training unit configured to train the autonomous driving model by adjusting parameters of the autonomous driving model, wherein the parameters are adjusted to reduce the difference between the predicted driving behavior and the coach driving behavior.

[0008] According to another aspect of the present disclosure, an electronic device is also provided, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the method provided according to an embodiment of the present disclosure.

[0009] According to another aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is further provided, wherein the computer instructions are used to enable the computer to execute the method provided according to the embodiment of the present disclosure.

[0010] According to another aspect of the present disclosure, a computer program product is provided, including a computer program, wherein when the computer program is executed by a processor, the method provided according to the embodiment of the present disclosure is implemented.

[0011] According to another aspect of the present disclosure, an autonomous driving vehicle is provided, on which a trained autonomous driving model is deployed, wherein the autonomous driving model is trained using a method provided according to an embodiment of the present disclosure.

[0012] According to one or more embodiments of the present disclosure, by utilizing multiple coaching models with different capabilities to provide imitation targets for the autonomous driving model to be trained, the performance ceiling of the trained autonomous driving model will not be constrained by any single coach, and it can integrate the strengths of various models to achieve better than the original, with comprehensive performance exceeding that of any single coach.

[0013] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present disclosure, nor are they intended to limit the scope of the present disclosure. Other features of the present disclosure will become readily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] The accompanying drawings illustrate exemplary embodiments and constitute a part of the specification. Together with the description of the specification, they serve to explain exemplary implementation of the embodiments. The illustrated embodiments are for illustrative purposes only and do not limit the scope of the claims. Throughout the drawings, the same reference numerals designate similar, but not necessarily identical, elements.

[0015] Figure 1 A schematic diagram illustrating an exemplary system in which the various methods described herein may be implemented according to an embodiment of the present disclosure;

[0016] Figure 2 An exemplary flowchart of a method for training an autonomous driving model according to an embodiment of the present disclosure is shown;

[0017] Figure 3 An exemplary process of training a coaching model according to an embodiment of the present disclosure is shown;

[0018] Figure 4 An exemplary block diagram of an apparatus for training an autonomous driving model according to an embodiment of the present disclosure is shown;

[0019] Figure 5 A structural block diagram of an exemplary electronic device that can be used to implement the embodiments of the present disclosure is shown. DETAILED DESCRIPTION

[0020] The following description of exemplary embodiments of the present disclosure is made in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be appreciated by those skilled in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0021] In this disclosure, unless otherwise specified, the use of terms such as "first" and "second" to describe various elements is not intended to limit the positional relationship, temporal relationship, or importance relationship of these elements. Such terms are only used to distinguish one element from another. In some examples, the first element and the second element may refer to the same instance of the element, while in some cases, based on the context of the description, they may also refer to different instances.

[0022] The terms used in the descriptions of the various examples described in this disclosure are for the purpose of describing specific examples only and are not intended to be limiting. Unless the context clearly indicates otherwise, if the number of elements is not specifically limited, the element may be one or more. In addition, the term "and / or" used in this disclosure encompasses any one and all possible combinations of the listed items.

[0023] The embodiments of the present disclosure will be described in detail below with reference to the accompanying drawings.

[0024] Figure 1 FIG2 is a schematic diagram of an exemplary system 100 in which the various methods and apparatuses described herein may be implemented according to an embodiment of the present disclosure. Figure 1 , 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 .

[0025] In an embodiment of the present disclosure, 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.

[0026] The server 120 may run one or more services or software applications that enable the method of training autonomous driving according to an embodiment of the present disclosure. In some embodiments, the server 120 may also provide other services or software applications, which may include non-virtual environments and virtual environments. Figure 1 In the configuration shown, 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 can be executed by one or more processors. The 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 differ from the system 100. Therefore, Figure 1 is one example of a system for implementing the various methods described herein and is not intended to be limiting.

[0027] Server 120 may include one or more general-purpose computers, specialized server computers (e.g., PC (personal computer) servers, UNIX servers, midrange 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). In various embodiments, server 120 may run one or more services or software applications that provide the functionality described below.

[0028] The computing units in the server 120 may run one or more operating systems including any of the operating systems described above as well as 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, and the like.

[0029] In some embodiments, 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.

[0030] The network 130 may be any type of network known to those skilled in the art that can support data communications using any of a variety of available protocols, including, but not limited to, TCP / IP, SNA, IPX, etc. By way of example only, the one or more networks 130 may be a satellite communication 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.

[0031] The system 100 may also include one or more databases 150. In some embodiments, these databases may be used to store data and other information. For example, one or more of the databases 150 may be used to store information such as audio files and video files. The databases 150 may reside in a variety of locations. For example, the database used by the server 120 may be local to the server 120, or may be remote from the server 120 and communicate with the server 120 via a network-based or dedicated connection. The databases 150 may be of different types. In some embodiments, the databases used by the server 120 may be relational databases. One or more of these databases may store, update, and retrieve data to and from the databases in response to commands.

[0032] In some embodiments, one or more of the databases 150 may also be used by applications to store application data. The databases used by the applications may be different types of databases, such as a key-value store, an object store, or a conventional store backed by a file system.

[0033] Motor vehicle 110 may include sensors 111 for sensing its surroundings. Sensors 111 may include one or more of the following: visual cameras, infrared cameras, ultrasonic sensors, millimeter-wave radar, and laser radar (LiDAR). Different sensors offer varying detection accuracy and range. Cameras may be mounted on the front, rear, or other locations of the vehicle. Visual cameras can capture real-time images of the vehicle's interior and exterior and present them to the driver and / or passengers. Furthermore, by analyzing the images captured by the visual cameras, information such as traffic light indications, intersection conditions, and the operating status of other vehicles can be obtained. Infrared cameras can detect objects in night vision conditions. Ultrasonic sensors can be mounted on all sides of the vehicle, utilizing the strong directionality of ultrasonic waves to measure the distance of external objects from the vehicle. Millimeter-wave radars can be mounted on the front, rear, or other locations of the vehicle, utilizing the properties of electromagnetic waves to measure the distance of external objects from the vehicle. LiDARs can be mounted on the front, rear, or other locations of the vehicle, detecting object edges and shapes for object recognition and tracking. Radar devices can also measure changes in the speed of the vehicle and moving objects due to the Doppler effect.

[0034] The motor vehicle 110 may also include a communication device 112. The communication device 112 may include a satellite positioning module capable of receiving satellite positioning signals (e.g., Beidou, GPS, GLONASS, and GALILEO) from satellites 141 and generating coordinates based on these signals. The communication device 112 may also include a module for communicating with a mobile communication base station 142. The mobile communication network may implement any suitable communication technology, such as GSM / GPRS, CDMA, LTE, or other current or evolving wireless communication technologies (e.g., 5G technology). The communication device 112 may also include a vehicle-to-everything (V2X) module configured to facilitate vehicle-to-everything (V2V) communications with other vehicles 143 and vehicle-to-infrastructure (V2I) communications with infrastructure 144. Furthermore, the communication device 112 may also include a module configured to communicate with a user terminal 145 (including but not limited to a smartphone, tablet computer, or wearable device such as a watch) via a wireless local area network (WLAN) or Bluetooth using the IEEE 802.11 standard, for example. The motor vehicle 110 may also access the server 120 via the network 130 using the communication device 112 .

[0035] The motor vehicle 110 may also include a control device 113. The control device 113 may include a processor, such as a central processing unit (CPU) or a graphics processing unit (GPU), or other specialized processor, in communication with various types of computer-readable storage devices or media. The control device 113 may include an autonomous driving system for automatically controlling various actuators in the vehicle. The autonomous 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 input from multiple sensors 111 or other input devices, thereby controlling acceleration, steering, and braking, respectively, without requiring or with limited human intervention. Some processing functions of the control device 113 may be implemented through cloud computing. For example, some processing may be performed using an onboard processor while other processing may be performed using computing resources in the cloud. The control device 113 may be configured to execute the methods according to the present disclosure. Furthermore, the control device 113 may be implemented as an example of a computing device on the motor vehicle side (client) according to the present disclosure.

[0036] Figure 1 The system 100 may be configured and operated in various ways to enable application of the various methods and apparatuses described in accordance with the present disclosure.

[0037] In most model-based systems for autonomous driving, models primarily acquire driving capabilities through imitation learning (or behavioral cloning) based on manually collected offline data. This learning approach inherently leads to compounding errors in closed-loop deployment. For example, a small initial deviation from the dataset can lead to new situations that the model has never seen, further exacerbating the deviation in simulation results.

[0038] To address the compound error problem in the imitation learning process of modeled autonomous driving systems, the closest existing technology primarily introduces online reinforcement learning to replace imitation learning for greater scalability. This technology uses the policy model obtained through reinforcement learning as a coach, and environmental data such as driving routes as a carrier, distilling a more generalizable student model through imitation learning.

[0039] First, imitation learning only learns static labels of offline data. In a fixed driving dataset, the states that the model can see are very limited. However, online reinforcement learning does not directly learn static labels. Instead, it continuously enters new states based on its own exploration of driving strategies, allowing the model to more fully and comprehensively understand the driving environment.

[0040] Secondly, imitation learning is very hungry for supervised data (human driving data), and usually requires learning from massive amounts of data to reach a good state. Reinforcement learning, on the other hand, only requires environmental data (such as road and traffic information) to explore on its own, generate massive learning signals, and ultimately obtain a policy model that is generalizable in the corresponding environment.

[0041] Finally, since reinforcement learning exploration cannot be performed directly in the real environment (such as exploring the action of hitting an obstacle), reinforcement learning is generally performed in a simulation environment. After obtaining the coach model, imitation learning is used to guide the student model in the real environment in order to obtain a more generalized imitation learning model.

[0042] Related technologies use a single trainer model to learn from a training dataset. However, it's extremely difficult to train a single trainer to fully master all scenarios within the same training dataset. Furthermore, there are constraints during the learning process: achieving good performance in one scenario may compromise performance in another. Therefore, training a single trainer model to achieve good driving performance across the entire training dataset incurs significant training costs.

[0043] In order to solve the above problems, the present disclosure provides a new method for training an autonomous driving model.

[0044] Figure 2 An exemplary flowchart of a method for training an autonomous driving model according to an embodiment of the present disclosure is shown.

[0045] In step S202, a training data set is determined, which includes a plurality of sample environmental data.

[0046] In step S204, a plurality of coaching models are determined, wherein each coaching model outputs a coaching driving behavior based on the sample environment data as a target driving behavior for the trained autonomous driving model, wherein each coaching model is capable of processing at least a portion of the sample environment data in the training dataset.

[0047] In step S206 , the sample environment data in the training data set is processed using the coaching model to obtain coaching driving behavior.

[0048] In step S208, the sample environment data in the training data set is processed using the autonomous driving model to obtain a predicted driving behavior.

[0049] In step S210 , the autonomous driving model is trained by adjusting parameters of the autonomous driving model, wherein the parameters are adjusted to reduce the difference between the predicted driving behavior and the coached driving behavior.

[0050] The method for training an autonomous driving model provided by the embodiments of the present disclosure utilizes multiple coaching models with different capabilities to provide imitation targets for the autonomous driving model to be trained, so that the performance ceiling of the trained autonomous driving model will not be constrained by any single coach, and it can integrate the strengths of multiple models to achieve better than the original, with comprehensive performance exceeding that of any single coach.

[0051] The principles of the present disclosure will be described in detail below.

[0052] In step S202, a training data set may be determined. The training data set includes a plurality of sample environmental data. In some embodiments, the sample environmental data may include data about the environment of at least one route to be traveled, including the route's geographic coordinates, road conditions, traffic light information, speed limit information, and other environmental data that may affect vehicle driving conditions. The environmental data used for training in the embodiments of the present disclosure may be data collected in a real-world scenario or data generated by a simulation system.

[0053] In step S204, a plurality of coaching models may be determined, wherein each coaching model outputs a coaching driving behavior based on the sample environment data as a target driving behavior for the trained autonomous driving model, wherein each coaching model is capable of processing at least a portion of the sample environment data in the training dataset.

[0054] The coaching model can be implemented using various models that can be used for autonomous driving, such as autonomous driving models based on convolutional neural network models (CNNs) or Transformer-based autonomous driving models. These autonomous driving models can perceive environmental data surrounding the vehicle and make decisions about future driving behavior based on at least the perceived environmental data, the vehicle's current state, and its historical state.

[0055] The coaching model's ability to process sample environmental data means that the coaching model can output a route corresponding to the sample environmental data in which the driving behavior successfully completes. Successful route completion means that the length of the road segments traveled while meeting safety requirements accounts for a proportion of the corresponding route length greater than a predetermined threshold, such as 95%.

[0056] Each of the multiple training models is capable of processing at least a portion of the sample environment data in the training dataset. In some examples, the sample environment data in the training dataset that each training model can process is different. In other examples, a single sample environment data set can be successfully processed by at least two training models. Those skilled in the art can determine the scope of sample environment data that a training model can process based on actual circumstances, and this is not limited herein, as long as the union of the sample environment data that can be processed by the multiple training models constitutes the training dataset.

[0057] The coaching model's output of the coaching behavior can be evaluated using pre-set rules or a pre-trained evaluation model to determine the coaching model's success rate. In some implementations, the evaluation metrics may include at least one of the following: the probability of a traffic violation occurring during the coaching behavior; the probability of a dangerous behavior occurring during the coaching behavior; and the ratio of the length of the road section safely traveled due to the coaching behavior to the length of the route corresponding to the sample environmental data. These evaluation metrics can effectively assess the coaching model's driving capabilities.

[0058] In some embodiments, the above-mentioned multiple coaching models can be trained by the following method: using a first coaching model to learn the training data set by reinforcement learning, and adjusting the parameters of the first coaching model until the first coaching model can at least correctly process the first sample environment data set in the training data set; determining a second sample environment data set, wherein the second sample environment data set includes the remaining sample environment data in the training data set except the first sample environment data set; using a second coaching model to learn the second sample environment data set by reinforcement learning, and adjusting the parameters of the second coaching model until the second coaching model can at least correctly process at least a part of the second sample environment data set.

[0059] During reinforcement learning, the model interacts with the environment and uses evaluation to determine whether its behavior is optimal. This process eliminates the need for manual learning goals for the model; instead, the model adjusts itself through feedback. By training the coach model using reinforcement learning, the coach model can freely explore the behaviors allowed by the environment based on data.

[0060] In some examples, adjusting the parameters of the first trainer model until the first trainer model can correctly process at least a first sample environment data set in the training dataset can include adjusting the parameters of the first trainer model using reinforcement learning until the first trainer model can successfully process at least a first number of sample environment data in the training dataset. In this example, the trained trainer model can be configured to learn a predetermined number of sample data in the training dataset. In this manner, the data scope covered by a single trainer model can be limited, simplifying the learning difficulty of the single trainer model.

[0061] In other examples, adjusting the parameters of the first coaching model until the first coaching model can correctly process at least the first sample environmental data set in the training data set may include adjusting the parameters of the first coaching model using reinforcement learning until the number of sample environmental data in the training data set that the first coaching model can successfully process reaches a maximum. In this example, the trained coaching model can be configured to learn as much as possible in the data set until its driving performance cannot be further improved. In this way, the coaching model can determine its own driving style and learn environmental data for scenarios in which it is good until it cannot be further improved. In this case, there is no need to change the model's driving style by adjusting parameters to adapt to other environmental data, but rather to learn again by training another coaching model. Each coaching model trained in this way has the ability to handle different scenarios.

[0062] Similarly, adjusting the parameters of the second coaching model until the second coaching model can correctly process at least a portion of the second sample environment data set may include: adjusting the parameters of the second coaching model using reinforcement learning until the second coaching model can successfully process at least a first number of sample environment data in the second sample environment data set; or adjusting the parameters of the second coaching model using reinforcement learning until the number of sample environment data in the training data set that can be successfully processed by the second coaching model reaches a maximum value.

[0063] Furthermore, it may be determined whether there is sample environment data in the training dataset that cannot be successfully processed by the first trainer model or the second trainer model. In response to determining that at least one remaining sample environment data cannot be successfully processed by the first trainer model or the second trainer model, a third trainer model is trained based on the at least one remaining sample environment data until all sample environment data are successfully learned.

[0064] Figure 3 An exemplary process of training a coaching model according to an embodiment of the present disclosure is shown.

[0065] like Figure 3 As shown in , D0 represents the entire training data set. Through reinforcement learning (RL), the Coach0 model with initial parameters is used to learn the data in D0, and the parameters of Coach0 are adjusted so that Coach0 can correctly process part of the data in D0. Figure 3 In the dataset D0, the dataset that can be processed by the Coach0 model after training is denoted as Succ0. Correspondingly, the dataset that cannot be processed by the Coach0 model in the dataset D0 is denoted as D1.

[0066] Similarly, the Coach1 model with initial parameters is used to learn the data in D1 through reinforcement learning. The dataset that Coach1 can handle after training is denoted as Succ1. The dataset that cannot be processed by either Coach0 or Coach1 is denoted as D2.

[0067] Furthermore, the Coach2 model with initial parameters is used to learn the data in D2 through reinforcement learning. Figure 3 In the example shown in , the trained Coach2 model is able to process all the data in D2.

[0068] The trained models Coach0, Coach1, and Coach2 can be used to provide imitation learning targets for the autonomous driving model to be trained.

[0069] although Figure 3 Only three coaching models are shown in the figure. Without departing from the principles of the present disclosure, those skilled in the art can train more or fewer coaching models according to actual conditions.

[0070] Return Reference Figure 2 In step S206, the sample environment data in the training data set can be processed using the coaching model to obtain the coaching driving behavior.

[0071] For each sample environment data, the sample environment data can be input into a coaching model that can process the sample environment data to obtain a successful coaching driving behavior as an imitation target for the autonomous driving model to be trained.

[0072] In some embodiments, it may be determined that the first sample environment data can be processed by a first coaching model. The first sample environment data may then be input into the first coaching model to obtain a first coaching driving behavior for the first sample environment data. Similarly, it may be determined that the second sample environment data can be processed by a second coaching model different from the first coaching model. The second sample environment data may then be input into a second coaching model to obtain a second coaching driving behavior for the second sample environment data, where the second sample environment data is different from the first sample environment data.

[0073] Using the above method, the coaching model can be used to process the sample environment data that it can process, thereby obtaining high-quality coaching driving behavior as the imitation target of the autonomous driving model.

[0074] The correspondence between the training model and the sample environment data it can process can be pre-stored. Based on the pre-stored correspondence, it can be queried whether the training model can process a sample environment data, or which (or which) training model can process a sample environment data.

[0075] In some embodiments, it can be determined that the sample environment data can be processed by the coaching model in the following manner: the sample environment data is input into the coaching model to obtain the coaching driving behavior for the sample environment data, the success rate of the coaching driving behavior is evaluated based on predetermined evaluation indicators, and when it is determined that the success rate of the coaching driving behavior is higher than a predetermined threshold, it is determined that the sample environment data can be processed by the coaching model.

[0076] In step S208, the sample environment data in the training data set may be processed using the autonomous driving model to obtain a predicted driving behavior. The autonomous driving model may have the same model structure as the coaching model.

[0077] In step S210 , the autonomous driving model may be trained by adjusting parameters of the autonomous driving model, wherein the parameters are adjusted to reduce the difference between the predicted driving behavior and the coached driving behavior.

[0078] In some embodiments, step S210 may include: inputting the first sample environment data into the autonomous driving model to obtain a first predicted driving behavior, and adjusting parameters of the autonomous driving model to reduce the difference between the first predicted driving behavior and the first coached driving behavior. Step S210 may also include: inputting the second sample environment data into the autonomous driving model to obtain a second predicted driving behavior, and adjusting parameters of the autonomous driving model to reduce the difference between the second predicted driving behavior and the second coached driving behavior.

[0079] Using this method, during the training process of the autonomous driving model, a coaching model capable of handling each scenario in the dataset is selected to provide guidance. Each scenario can be assigned a suitable coaching model to produce appropriate results. Because the coaching models are trained to excel at handling different scenarios, a single model guided by multiple coaches can potentially surpass the capabilities of any single coach.

[0080] Figure 4 An exemplary block diagram of an apparatus for training an autonomous driving model according to an embodiment of the present disclosure is shown.

[0081] like Figure 4 As shown, the apparatus 400 may include a training data determination unit 410 , a coaching model determination unit 420 , a coaching behavior acquisition unit 430 , a prediction unit 440 , and a training unit 450 .

[0082] The training data determination unit 410 can be configured to determine a training data set, which includes a plurality of sample environmental data. The coach model determination unit 420 can be configured to determine a plurality of coach models, wherein the coach model outputs a coach driving behavior as a target driving behavior of the trained autonomous driving model based on the sample environmental data, wherein each coach model is capable of processing at least a portion of the sample environmental data in the training data set. The coach behavior acquisition unit 430 can be configured to process the sample environmental data in the training data set using the coach model to obtain the coach driving behavior. The prediction unit 440 is configured to process the sample environmental data in the training data set using the autonomous driving model to obtain the predicted driving behavior. The training unit 450 is configured to train the autonomous driving model by adjusting the parameters of the autonomous driving model, wherein the parameters are adjusted to reduce the difference between the predicted driving behavior and the coach driving behavior.

[0083] By utilizing the apparatus for training an autonomous driving model provided by the embodiments of the present disclosure, a plurality of coaching models with different capabilities are utilized to provide imitation targets for the autonomous driving model to be trained. This ensures that the performance ceiling of the trained autonomous driving model is not constrained by any single coach, and that the model can integrate the strengths of various models, thereby surpassing the original and achieving comprehensive performance exceeding that of any single coach.

[0084] In some embodiments, the coaching behavior acquisition unit is configured to: determine that the first sample environment data can be processed by the first coaching model; input the first sample environment data into the first coaching model to obtain a first coaching driving behavior for the first sample environment data; determine that the second sample environment data can be processed by the second coaching model; input the second sample environment data into the second coaching model to obtain a second coaching driving behavior for the second sample environment data, wherein the second sample environment data is different from the first sample environment data.

[0085] In some embodiments, the training is configured to: input first sample environmental data into the autonomous driving model to obtain a first predicted driving behavior; adjust the parameters of the autonomous driving model to reduce the difference between the first predicted driving behavior and the first coached driving behavior; input second sample environmental data into the autonomous driving model to obtain a second predicted driving behavior; adjust the parameters of the autonomous driving model to reduce the difference between the second predicted driving behavior and the second coached driving behavior.

[0086] In some embodiments, the sample environment data in the training dataset that each training model can process is different from each other.

[0087] In some embodiments, multiple coaching models are trained by the following method: using a first coaching model to learn a training data set by reinforcement learning, and adjusting the parameters of the first coaching model until the first coaching model can at least correctly process a first sample environment data set in the training data set; determining a second sample environment data set, wherein the second sample environment data set includes the remaining sample environment data in the training data set except the first sample environment data set; using a second coaching model to learn the second sample environment data set by reinforcement learning, and adjusting the parameters of the second coaching model until the second coaching model can at least correctly process at least a portion of the second sample environment data set.

[0088] In some embodiments, adjusting the parameters of the first coaching model until the first coaching model can correctly process at least a first sample environment data set in the training data set includes: adjusting the parameters of the first coaching model using reinforcement learning until the first coaching model can successfully process at least a first number of sample environment data in the training data set; or adjusting the parameters of the first coaching model using reinforcement learning until the number of sample environment data in the training data set that can be successfully processed by the first coaching model reaches a maximum value.

[0089] In some embodiments, adjusting the parameters of the second trainer model until the second trainer model can correctly process at least a portion of the second sample environment data set in the training data set includes: adjusting the parameters of the second trainer model using reinforcement learning until the second trainer model can successfully process at least a first number of sample environment data in the second sample environment data set; or adjusting the parameters of the second trainer model using reinforcement learning until the number of sample environment data in the training data set that can be successfully processed by the second trainer model reaches a maximum value; and determining whether there is sample environment data in the training data set that cannot be successfully processed by the first trainer model or the second trainer model; in response to determining that at least one remaining sample environment data cannot be successfully processed by the first trainer model or the second trainer model, training a third trainer model based on the at least one remaining sample environment data.

[0090] In some embodiments, the sample environment data includes data of at least one environment of a route to be traveled.

[0091] In some embodiments, it is determined that the sample environment data can be processed by the coaching model in the following manner: the sample environment data is input into the coaching model to obtain the coaching driving behavior for the sample environment data; the success rate of the coaching driving behavior is evaluated based on predetermined evaluation indicators; when it is determined that the success rate of the coaching driving behavior is higher than a predetermined threshold, it is determined that the sample environment data can be processed by the coaching model.

[0092] In some embodiments, the evaluation indicators include at least one of the following: the probability of illegal behavior occurring in the coach's driving behavior; the probability of dangerous behavior occurring in the coach's driving behavior; the proportion of the length of the road section that is safely passed due to the coach's driving behavior to the length of the route corresponding to the sample environment data.

[0093] It should be understood that Figure 4 The modules or units of the apparatus 400 shown in FIG. 4 can be used in conjunction with the reference Figure 2 The steps in the method 200 described above correspond to each other. Therefore, the operations, features and advantages described above for the method 200 are also applicable to the apparatus 400 and the modules and units included therein. For the sake of brevity, some operations, features and advantages are not repeated here.

[0094] Although specific functionality is discussed above with reference to specific modules, it should be noted that the functionality of the various units discussed herein may be separated into multiple units, and / or at least some functionality of multiple units may be combined into a single unit.

[0095] It should also be understood that various techniques may be described herein in the general context of software hardware elements or program modules. Figure 4The various units described can be implemented in hardware or in hardware combined with software and / or firmware. For example, these units can be implemented as computer program code / instructions that are configured to be executed in one or more processors and stored in a computer-readable storage medium. Alternatively, these units can be implemented as hardware logic / circuits. For example, in some embodiments, one or more of units 410 to 450 can be implemented together in a system on chip (SoC). The SoC can 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.), memory, one or more communication interfaces, and / or one or more components in other circuits), and can optionally execute received program code and / or include embedded firmware to perform functions.

[0096] In the technical solutions disclosed herein, the acquisition, storage, and application of user personal information involved comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0097] According to an embodiment of the present disclosure, an electronic device, a readable storage medium, a computer program product and an autonomous driving vehicle are also provided.

[0098] According to an embodiment of the present disclosure, an electronic device is also provided, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the method provided according to an embodiment of the present disclosure.

[0099] According to an embodiment of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is further provided, wherein the computer instructions are used to enable the computer to execute the method provided according to the embodiment of the present disclosure.

[0100] According to an embodiment of the present disclosure, a computer program product is further provided, including a computer program, wherein the computer program implements the method provided according to the embodiment of the present disclosure when executed by a processor.

[0101] According to an embodiment of the present disclosure, an autonomous driving vehicle is also provided, on which a trained autonomous driving model is deployed, wherein the autonomous driving model is trained using a method provided according to an embodiment of the present disclosure.

[0102] refer to Figure 5, a block diagram of an electronic device 500 that can serve 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. The electronic device can 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 required herein.

[0103] like Figure 5 As shown, 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. RAM 503 may also store various programs and data required for the operation of electronic device 500. Computing unit 501, ROM 502, and RAM 503 are interconnected via a bus 504. An input / output (I / O) interface 505 is also connected to bus 504.

[0104] Multiple components within electronic device 500 are connected to I / O interface 505, including an input unit 506, an output unit 507, a storage unit 508, and a communication unit 509. Input unit 506 can be any type of device capable of inputting information into electronic device 500. Input unit 506 can receive input numeric or character information and generate key signal input related to user settings and / or function control of the electronic device. It may include, but is not limited to, a mouse, keyboard, touch screen, trackpad, trackball, joystick, microphone, and / or remote control. Output unit 507 can be any type of device capable of presenting information, and may include, but is not limited to, a display, a speaker, a video / audio output terminal, a vibrator, and / or a printer. Storage unit 508 may include, but is not limited to, a magnetic disk or an optical disk. Communication unit 509 allows electronic device 500 to exchange information / data with other devices via computer networks such as the Internet and / or various telecommunication networks. It may include, but is not limited to, a modem, a network card, an infrared communication device, a wireless communication transceiver and / or chipset, such as a Bluetooth™ device, an 802.11 device, a WiFi device, a WiMax device, a cellular communication device, and / or the like.

[0105] Computing unit 501 can be any general-purpose and / or specialized processing component with processing and computing capabilities. Some examples of computing unit 501 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various specialized artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. Computing unit 501 performs the various methods and processes described above, such as method 200. For example, in some embodiments, method 200 can be implemented as a computer software program tangibly embodied in a machine-readable medium, such as storage unit 508. In some embodiments, part or all of the computer program can be loaded and / or installed onto electronic device 500 via ROM 502 and / or communication unit 509. When the computer program is loaded into RAM 503 and executed by computing unit 501, one or more steps of method 200 described above can be performed. Alternatively, in other embodiments, computing unit 501 can be configured to perform method 200 in any other suitable manner (e.g., via firmware).

[0106] Various embodiments of the systems and techniques described above 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), system-on-chip systems (SOCs), programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include being implemented in one or more computer programs that are executable and / or interpreted on a programmable system that includes 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.

[0107] The program code for implementing the method of the present disclosure can be written in any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device so that when the program code is executed by the processor or controller, the functions / operations specified in the flow chart and / or block diagram are implemented. The program code can be executed entirely on the machine, partially on the machine, as a stand-alone software package, partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0108] In the context of the present disclosure, 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, apparatus, or device. 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, apparatus, or device, or any suitable combination of the foregoing. More specific examples of machine-readable storage media may include an electrical connection based on one or more wires, 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), optical fibers, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0109] To provide interaction with a user, 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. 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).

[0110] The systems and techniques described herein can be implemented in a computing system that includes back-end 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 front-end components (e.g., a user computer with a graphical user interface or 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 back-end components, middleware components, or front-end 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.

[0111] A computer system may include a client and a server. The client and server are generally remote from each other and typically interact through a communication network. The client-server relationship arises through computer programs running on the respective computers and having a client-server relationship with each other. The server may be a cloud server, a server in a distributed system, or a server integrated with a blockchain.

[0112] It should be understood that the various forms of the processes shown above can be used to reorder, add, or delete steps. For example, the steps described in this disclosure can be performed in parallel, sequentially, or in a different order, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved. This is not limited herein.

[0113] Although the embodiments or examples of the present disclosure have been described with reference to the accompanying drawings, it should be understood that the above-mentioned methods, systems and devices are merely exemplary embodiments or examples, and the scope of the present invention is not limited by these embodiments or examples, but is only limited by the claims after authorization and their equivalents. Various elements in the embodiments or examples may be omitted or replaced by their equivalents. In addition, the steps may be performed in an order different from that described in this disclosure. Further, the various elements in the embodiments or examples may be combined in various ways. It is important that as technology evolves, many of the elements described herein may be replaced by equivalent elements that appear after this disclosure.

Claims

1. A method for training an autonomous driving model, comprising: Determine a training data set, wherein the training data set includes a plurality of sample environmental data; determining a plurality of coaching models, wherein the coaching models output coaching driving behaviors as target driving behaviors of the trained autonomous driving model based on the sample environment data, wherein each coaching model is capable of processing at least a portion of the sample environment data in the training dataset; Processing the sample environment data in the training data set using the coaching model to obtain coaching driving behavior; Processing the sample environment data in the training dataset using the autonomous driving model to obtain a predicted driving behavior; training the autonomous driving model by adjusting parameters of the autonomous driving model, wherein the parameters are adjusted to reduce the difference between the predicted driving behavior and the coached driving behavior, The multiple coaching models are trained by the following method: Using a first coaching model to learn the training dataset through reinforcement learning, and adjusting parameters of the first coaching model until the first coaching model can correctly process at least a first sample environment dataset in the training dataset; Determining a second sample environment data set, wherein the second sample environment data set includes remaining sample environment data in the training data set except the first sample environment data set; Using a second trainer model to learn the second sample environment dataset through reinforcement learning, and adjusting parameters of the second trainer model until the second trainer model can correctly process at least a portion of the second sample environment dataset; determining whether there is sample environment data in the training dataset that cannot be successfully processed by the first coaching model or the second coaching model; In response to determining that at least one remaining sample environment data cannot be successfully processed by the first trainer model or the second trainer model, at least one third trainer model is trained based on the at least one remaining sample environment data until the sample environment data in the training dataset is successfully learned.

2. The method according to claim 1, wherein Processing the sample environment data in the training data set using the coaching model to obtain coaching driving behavior includes: determining that the first sample environment data can be processed by the first coaching model; inputting first sample environment data into the first coaching model to obtain a first coaching driving behavior for the first sample environment data; determining that second sample environment data can be processed by a second coaching model different from the first coaching model; Second sample environment data is input into the second coaching model to obtain a second coaching driving behavior for the second sample environment data, wherein the second sample environment data is different from the first sample environment data.

3. The method according to claim 2, wherein: Training the autonomous driving model by adjusting parameters of the autonomous driving model includes: Inputting the first sample environment data into the autonomous driving model to obtain a first predicted driving behavior; adjusting parameters of the autonomous driving model to reduce a difference between the first predicted driving behavior and the first coached driving behavior; Inputting the second sample environment data into the autonomous driving model to obtain a second predicted driving behavior; Adjust parameters of the autonomous driving model to reduce the difference between the second predicted driving behavior and the second coached driving behavior.

4. The method according to claim 1, wherein The sample environment data in the training dataset that can be processed by each training model are different from each other.

5. The method according to claim 1, wherein Adjusting the parameters of the first coaching model until the first coaching model can at least correctly process the first sample environment dataset in the training dataset includes: Adjusting parameters of the first coaching model using reinforcement learning until the first coaching model can successfully process at least a first amount of sample environment data in the training dataset; or The parameters of the first coaching model are adjusted using reinforcement learning until the number of sample environment data in the training dataset that can be successfully processed by the first coaching model reaches a maximum value.

6. The method according to claim 5, wherein: Adjusting the parameters of the second coaching model until the second coaching model can correctly process at least a portion of the second sample environment dataset in the training dataset includes: Adjusting parameters of the second coaching model using reinforcement learning until the second coaching model can successfully process at least a first amount of sample environment data in the second sample environment dataset; or The parameters of the second coaching model are adjusted by using reinforcement learning until the number of sample environment data in the training data set that can be successfully processed by the second coaching model reaches a maximum value.

7. The method according to any one of claims 1 to 6, wherein The sample environment data includes data of at least one route environment to be traveled.

8. The method according to any one of claims 1 to 6, wherein Determining that the sample environment data can be processed by the coaching model is done by: inputting the sample environment data into the coaching model to obtain a coaching driving behavior for the sample environment data; Evaluating the success rate of the coach's driving behavior based on predetermined evaluation indicators; When it is determined that the success rate of the coaching driving behavior is higher than a predetermined threshold, it is determined that the sample environment data can be processed by the coaching model.

9. The method of claim 8, wherein: The evaluation indicators include at least one of the following: The probability of the instructor's driving behavior violating traffic rules; The probability of dangerous behavior occurring during the instructor's driving behavior; The ratio of the length of the road section safely traveled as a result of the instructor's driving behavior to the length of the route corresponding to the sample environment data.

10. A device for training an autonomous driving model, comprising: a training data determination unit configured to determine a training data set, wherein the training data set includes a plurality of sample environmental data; a coaching model determining unit configured to determine a plurality of coaching models, wherein the coaching models output coaching driving behaviors as target driving behaviors of the trained autonomous driving model based on the sample environment data, wherein each coaching model is capable of processing at least a portion of the sample environment data in the training dataset; a coaching behavior acquisition unit configured to process the sample environment data in the training data set using the coaching model to obtain a coaching driving behavior; a prediction unit configured to process the sample environment data in the training dataset using the autonomous driving model to obtain a predicted driving behavior; a training unit configured to train the autonomous driving model by adjusting parameters of the autonomous driving model, wherein the parameters are adjusted to reduce the difference between the predicted driving behavior and the coached driving behavior, The multiple coaching models are trained by the following method: Using a first coaching model to learn the training dataset through reinforcement learning, and adjusting parameters of the first coaching model until the first coaching model can correctly process at least a first sample environment dataset in the training dataset; Determining a second sample environment data set, wherein the second sample environment data set includes remaining sample environment data in the training data set except the first sample environment data set; Using a second trainer model to learn the second sample environment dataset through reinforcement learning, and adjusting parameters of the second trainer model until the second trainer model can correctly process at least a portion of the second sample environment dataset; determining whether there is sample environment data in the training dataset that cannot be successfully processed by the first coaching model or the second coaching model; In response to determining that at least one remaining sample environment data cannot be successfully processed by the first trainer model or the second trainer model, at least one third trainer model is trained based on the at least one remaining sample environment data until the sample environment data in the training dataset is successfully learned.

11. The device according to claim 10, wherein The coaching behavior acquisition unit is configured to: determining that first sample environmental data can be processed by the first coaching model; inputting first sample environment data into a first coaching model to obtain a first coaching driving behavior for the first sample environment data; determining that the second sample environment data can be processed by the second coaching model; Second sample environment data is input into the second coaching model to obtain a second coaching driving behavior for the second sample environment data, wherein the second sample environment data is different from the first sample environment data.

12. The device according to claim 11, wherein The training unit is configured to: Inputting the first sample environment data into the autonomous driving model to obtain a first predicted driving behavior; adjusting parameters of the autonomous driving model to reduce a difference between the first predicted driving behavior and the first coached driving behavior; Inputting the second sample environment data into the autonomous driving model to obtain a second predicted driving behavior; Adjust parameters of the autonomous driving model to reduce the difference between the second predicted driving behavior and the second coached driving behavior.

13. The device according to claim 10, wherein The sample environment data in the training dataset that can be processed by each training model are different from each other.

14. The apparatus of claim 13, wherein: Adjusting the parameters of the first coaching model until the first coaching model can at least correctly process the first sample environment dataset in the training dataset includes: Adjusting parameters of the first coaching model using reinforcement learning until the first coaching model can successfully process at least a first amount of sample environment data in the training dataset; or The parameters of the first coaching model are adjusted using reinforcement learning until the number of sample environment data in the training dataset that can be successfully processed by the first coaching model reaches a maximum value.

15. The apparatus of claim 14, wherein: Adjusting the parameters of the second coaching model until the second coaching model can correctly process at least a portion of the second sample environment dataset in the training dataset includes: Adjusting parameters of the second coaching model using reinforcement learning until the second coaching model can successfully process at least a first amount of sample environment data in the second sample environment dataset; or The parameters of the second coaching model are adjusted by using reinforcement learning until the number of sample environment data in the training data set that can be successfully processed by the second coaching model reaches a maximum value.

16. The device according to any one of claims 10 to 15, wherein The sample environment data includes data of at least one route environment to be traveled.

17. The device according to any one of claims 10 to 15, wherein Determining that the sample environment data can be processed by the coaching model is done by: inputting the sample environment data into the coaching model to obtain a coaching driving behavior for the sample environment data; Evaluating the success rate of the coach's driving behavior based on predetermined evaluation indicators; When it is determined that the success rate of the coaching driving behavior is higher than a predetermined threshold, it is determined that the sample environment data can be processed by the coaching model.

18. The apparatus of claim 17, wherein: The evaluation indicators include at least one of the following: The probability of the instructor's driving behavior violating traffic rules; The probability of dangerous behavior occurring during the instructor's driving behavior; The ratio of the length of the road section safely traveled as a result of the instructor's driving behavior to the length of the route corresponding to the sample environment data.

19. An electronic device comprising: at least one processor; as well as a memory communicatively coupled to the at least one processor; in The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 9.

20. A non-transitory computer-readable storage medium storing computer instructions, wherein: The computer instructions are used to cause the computer to execute the method according to any one of claims 1 to 9.

21. A computer program product comprising a computer program, wherein When the computer program is executed by a processor, the method according to any one of claims 1 to 9 is implemented.

22. An autonomous driving vehicle having a trained autonomous driving model deployed thereon, wherein the autonomous driving model is trained by the method according to any one of claims 1-9.

Images