A vehicle action prediction model training method and a vehicle control method

Abstract

This disclosure relates to a training method for a vehicle motion prediction model and a vehicle control method. The vehicle motion prediction model includes a current motion prediction network and a subsequent motion prediction network. The training method includes: acquiring training data, which includes motion data and first state data at a first time step, and second state data at a second time step; inputting the motion data and first state data at the first time step into the current motion prediction network, and inputting the second state data into the subsequent motion prediction network; performing performance prediction based on the prediction results of the first and second networks for the current and subsequent motion prediction networks, respectively; and adjusting the parameters of each network according to the performance prediction results. The training data is determined based on vehicle motion data generated by the driver driving the vehicle, the autonomous driving simulation platform, and the autonomous driving trajectory planning, as well as a high-precision map. This disclosure can improve the quality of training data and avoid getting trapped in local optima.

Description

Technical Field

[0001] This disclosure relates to the field of autonomous driving technology, and in particular to a training method for a vehicle motion prediction model and a vehicle control method. Background Technology

[0002] In recent years, autonomous driving technology has become a research hotspot for major internet service providers. The proposed methods related to trajectory planning and vehicle control can be broadly categorized into two types: non-data-driven methods and data-driven methods based on machine learning. Currently, most existing machine learning-based trajectory planning and vehicle control methods rely on online algorithms. The agent interacts with the simulation environment, maximizing the reward function through trial and error to find the optimal trajectory strategy. This method requires additional work in addition to algorithm development and simulation environment setup, increasing the workload for technical personnel. Secondly, the agent in this method can only learn based on its own experience and cannot learn from the experience of other optimal or suboptimal strategies, making it prone to getting trapped in local optima. Furthermore, most of these methods are applied to high-speed scenarios, and considering information such as traffic lights and signs may lead to failure in trajectory planning and vehicle control tasks. Summary of the Invention

[0003] To address at least one of the aforementioned technical problems, this disclosure proposes a training method for a vehicle motion prediction model and a vehicle control method.

[0004] According to one aspect of this disclosure, a training method for a vehicle motion prediction model is provided. The vehicle motion prediction model includes a current motion prediction network and a subsequent motion prediction network. The method comprises: acquiring training data, the training data including motion data and first state data at a first moment, and second state data at a second moment, the second moment being the moment following the first moment; inputting the motion data and first state data at the first moment into the current motion prediction network, and inputting the second state data into the subsequent motion prediction network; performing performance prediction on the prediction results of the current motion prediction network and the subsequent motion prediction network based on the first network and the second network, respectively; adjusting the parameters of the current motion prediction network, the subsequent motion prediction network, the first network, and the second network according to the performance prediction results; wherein the training data is determined based on vehicle motion data generated by a driver driving the vehicle, vehicle motion data generated by an autonomous driving simulation platform, vehicle motion data generated by autonomous driving trajectory planning, and a high-precision map.

[0005] In some possible implementations, the first network includes at least one first sub-network, the second network includes at least one second sub-network, each of the first sub-networks is connected to the current action prediction network, and each of the second sub-networks is connected to the subsequent action prediction network.

[0006] In some possible implementations, the step of inputting the action data and the first state data at the first moment into the current action prediction network, and inputting the second state data into the subsequent action prediction network, includes: inputting the action data and the first state data at the first moment into the current action prediction network to obtain a first predicted action at the first moment; and inputting the second state data into the subsequent action prediction network to obtain a second predicted action at the second moment.

[0007] In some possible implementations, the first network includes at least one first sub-network, and the second network includes at least one second sub-network. The execution performance prediction based on the prediction results of the current action prediction network and the subsequent action prediction network using the first network and the second network respectively includes: inputting the first estimated action and the first state data into each first sub-network to obtain a corresponding first prediction result, whereby the first prediction result represents the execution result of the vehicle executing the first estimated action under the condition corresponding to the first state data; and inputting the second estimated action and the second state data into each second sub-network to obtain a corresponding second prediction result, whereby the second prediction result represents the execution result of the vehicle executing the second estimated action under the condition corresponding to the second state data.

[0008] In some possible implementations, adjusting the parameters of the current action prediction network, the subsequent action prediction network, the first network, and the second network based on the performance prediction results includes: determining a target value based on each of the second prediction results; determining a model loss based on the target value and each of the first prediction results; and adjusting the parameters of the current action prediction network, the subsequent action prediction network, the first network, and the second network based on the model loss and the first prediction results.

[0009] In some possible implementations, adjusting the parameters of the current action prediction network, the subsequent action prediction network, the first network, and the second network based on the performance prediction results includes: adjusting the parameters of the current action prediction network in the direction of gradient ascent based on each of the first prediction results; adjusting the parameters of the first network based on the model loss; and adjusting the parameters of the subsequent action prediction network and the second network respectively according to the parameters of the current action prediction network and the parameters of the first network.

[0010] According to a second aspect of this disclosure, a vehicle control method is provided, comprising: inputting the state data of a vehicle at a target time into a subsequent action prediction network in a vehicle action prediction model to obtain an estimated action at the target time, wherein the vehicle action prediction model is trained using the training method of the vehicle action prediction model according to any one of the first aspects; solving for parameters in a trajectory polynomial based on the estimated action, wherein the trajectory polynomial is used to describe the driving trajectory of the vehicle; and generating control commands for the vehicle based on the trajectory polynomial.

[0011] According to a third aspect of this disclosure, a training apparatus for a vehicle motion prediction model is provided, comprising: a data acquisition module for acquiring training data, the training data including motion data and first state data at a first moment, and second state data at a second moment, the second moment being the moment following the first moment; a first processing module for inputting the motion data and first state data at the first moment into a current motion prediction network, and inputting the second state data into a subsequent motion prediction network; a second processing module for performing performance prediction based on the prediction results of the first network and the subsequent motion prediction network, respectively; and an optimization module for adjusting the parameters of the current motion prediction network, the subsequent motion prediction network, the first network, and the second network according to the performance prediction results; wherein the training data is determined based on vehicle motion data generated by a driver driving a vehicle, vehicle motion data generated by an autonomous driving simulation platform, vehicle motion data generated by autonomous driving trajectory planning, and a high-precision map.

[0012] According to a fourth aspect of this disclosure, an electronic device is provided, including 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 at least one processor implements a training method for a vehicle motion prediction model as described in any one of the first aspects or a vehicle control method as described in the second aspect by executing the instructions stored in the memory.

[0013] According to a fifth aspect of this disclosure, a computer-readable storage medium is provided, wherein at least one instruction or at least one program is stored therein, the at least one instruction or at least one program being loaded and executed by a processor to implement a training method for a vehicle motion prediction model as described in any of the first aspects or a vehicle control method as described in the second aspect.

[0014] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure.

[0015] Implementing this disclosure has the following beneficial effects: This disclosure provides a training method for a vehicle motion prediction model. Training data is determined using vehicle motion data generated by a driver, an autonomous driving simulation platform, and autonomous driving trajectory planning, along with a high-precision map. This training data includes motion data and first state data at a first moment, and second state data at a second moment. The motion data and first state data at the first moment are input into the current motion prediction network, and the second state data is input into the subsequent motion prediction network. Based on the prediction results from the first and second networks, execution performance prediction is performed on the current and subsequent motion prediction networks, respectively. The parameters of each network are adjusted according to the execution performance prediction results. This method can improve the quality of training data, avoid getting trapped in local optima, and make it applicable to more complex traffic scenarios. Furthermore, it eliminates the need for an additional simulation platform, reducing the workload of technical personnel.

[0016] Other features and aspects of this disclosure will become clear from the following detailed description of exemplary embodiments with reference to the accompanying drawings. Attached Figure Description

[0017] To more clearly illustrate the technical solutions and advantages in the embodiments or prior art of this specification, the drawings used in the description of the embodiments or prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this specification. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A schematic flowchart of a training method for a vehicle motion prediction model according to an embodiment of the present disclosure is shown.

[0019] Figure 2 The input-output diagram of the current action prediction network according to an embodiment of this disclosure is shown;

[0020] Figure 3 The input-output diagram of the subsequent action prediction network according to an embodiment of this disclosure is shown.

[0021] Figure 4 A deployment diagram of each network according to embodiments of this disclosure is shown;

[0022] Figure 5 A flowchart illustrating a method for performing performance prediction based on prediction results according to an embodiment of this disclosure is shown.

[0023] Figure 6 This diagram illustrates a method flow for adjusting various network parameters according to an embodiment of the present disclosure. Figure 1 ;

[0024] Figure 7This diagram illustrates a method flow for adjusting various network parameters according to an embodiment of the present disclosure. Figure 2 ;

[0025] Figure 8 This diagram illustrates a flow chart of a vehicle motion control method according to an embodiment of the present disclosure.

[0026] Figure 9 A schematic diagram of a vehicle motion prediction model training device according to an embodiment of the present disclosure is shown.

[0027] Figure 10 This diagram shows a block diagram of a vehicle motion control device according to an embodiment of the present disclosure;

[0028] Figure 11 A block diagram of an electronic device according to an embodiment of the present disclosure is shown;

[0029] Figure 12 A block diagram of another electronic device according to an embodiment of the present disclosure is shown. Detailed Implementation

[0030] The technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this specification, and not all embodiments. Based on the embodiments in this specification, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0031] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or server that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0032] Various exemplary embodiments, features, and aspects of this disclosure will now be described in detail with reference to the accompanying drawings. The same reference numerals in the drawings denote elements that have the same or similar functions. Although various aspects of the embodiments are shown in the drawings, they are not necessarily drawn to scale unless specifically indicated otherwise.

[0033] The term “exemplary” as used herein means “serving as an example, embodiment, or illustration.” Any embodiment illustrated herein as “exemplary” is not necessarily to be construed as superior to or better than other embodiments.

[0034] In this document, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent three cases: A alone, A and B simultaneously, and B alone. Furthermore, the term "at least one" in this document means any combination of at least two of any one or more elements. For example, including at least one of A, B, and C can mean including any one or more elements selected from the set consisting of A, B, and C.

[0035] Furthermore, to better illustrate this disclosure, numerous specific details are set forth in the following detailed description. Those skilled in the art will understand that this disclosure can be practiced without certain specific details. In some instances, methods, means, components, and circuits well known to those skilled in the art have not been described in detail in order to highlight the main points of this disclosure.

[0036] Machine learning-based vehicle motion prediction model training methods can be subdivided into upper-layer motion methods, lower-layer motion methods, and hybrid motion methods. Upper-layer motion methods refer to agents whose action space consists of commands such as lane changing or lane keeping. In lower-layer motion methods, the agent typically outputs operations with the lowest level of control, such as vehicle acceleration and steering wheel angle. Because lower-layer motion methods do not depend on other lower-layer structures, the agent requires a large amount of high-quality data to learn a complete control strategy. The motion commands of upper-layer motion methods are usually fixed, which limits their application scenarios. Hybrid motion methods, on the other hand, balance upper-layer and lower-layer motions while providing the agent with sufficient flexibility, making them valuable for widespread application.

[0037] Most existing machine learning-based vehicle motion prediction model training methods rely on online deep reinforcement learning algorithms. The agent interacts with the simulation environment, maximizing the reward function through trial and error to find the optimal trajectory strategy. First, besides building the algorithm system, this method requires additional work on setting up the simulation environment, increasing the workload for technical personnel. Second, the agent in this method can only learn based on its own experience and cannot learn from the experience of other optimal or suboptimal strategies, easily getting trapped in local optima. Furthermore, most existing methods are applied to high-speed scenarios, and considering information such as traffic lights and signs may lead to trajectory planning failure. Therefore, this disclosure provides a vehicle motion prediction model training method to address at least one of the above problems.

[0038] The vehicle motion prediction model training method and the vehicle control method based on the aforementioned vehicle motion prediction model provided in this disclosure can be executed by a terminal device, a server, or other types of electronic devices. The terminal device can be a user equipment (UE), mobile device, user terminal, terminal, cellular phone, cordless phone, personal digital assistant (PDA), handheld device, computing device, in-vehicle device, wearable device, etc. In some possible implementations, the vehicle motion prediction model training method and the vehicle control method based on the aforementioned vehicle motion prediction model can be implemented by a processor calling computer-readable instructions stored in memory. The neural network training method and the target re-identification method based on the aforementioned neural network in this disclosure are described below using an electronic device as the execution subject as an example.

[0039] Figure 1 This diagram illustrates a flowchart of a vehicle motion prediction model training method according to an embodiment of the present disclosure. The vehicle motion prediction model includes a current motion prediction network and a subsequent motion prediction network, such as... Figure 1 As shown, the above method includes:

[0040] S101: Obtain training data, which includes action data and first state data at the first moment, and second state data at the second moment, where the second moment is the moment following the first moment.

[0041] In the embodiments of this disclosure, the training data is determined based on vehicle motion data generated by the driver, vehicle motion data generated by the autonomous driving simulation platform, vehicle motion data generated by autonomous driving trajectory planning, and high-precision maps. The vehicle motion data generated by autonomous driving trajectory planning is not limited to the trajectory planning method based on the vehicle motion prediction model training method provided in this disclosure, but can be any trajectory planning strategy in the field of autonomous driving.

[0042] In the embodiments of this disclosure, the state data includes the relative distance parameters between the controlled vehicle and surrounding vehicles, relative speed parameters, relative lane parameters, longitudinal speed parameters of the controlled vehicle, lateral speed parameters of the controlled vehicle, logic values ​​for whether the vehicle can change lanes to the left and right, logic values ​​for the traffic light signals and pedestrian signs of the road segment; the action data includes the longitudinal speed parameters of the controlled vehicle, the time parameters for maintaining longitudinal travel, the time parameters for maintaining lateral travel, and lateral position parameters.

[0043] In other embodiments, the training data may also include labeled data, which indicates whether the training has been terminated. Based on the labeled data, a technician may set the number of training iterations for the vehicle motion prediction model training method of this disclosure. When the specified number of iterations is reached, the labeled data indicates that the training has been terminated.

[0044] Based on the above configuration, the training data can cover complex traffic scenarios, expanding the application scenarios of the vehicle motion prediction model disclosed herein. At the same time, since the acquisition of training data is not limited to vehicle motion data generated by any trajectory planning method, the training method disclosed herein can enrich the types of training data and can use training data as nodes to connect other trajectory planning strategies, treating them as data generation strategies, thereby achieving the continuity of technology updates and iterations. Theoretically, it can learn from any excellent experience data, avoiding getting trapped in local optima due to the singularity of training data, and also saving the workload of building a simulation platform.

[0045] S102: Input the action data and first state data at the first moment into the current action prediction network, and input the second state data into the subsequent action prediction network.

[0046] Figure 2 The input-output diagram of the current action prediction network in this disclosure is shown. Figure 3 The input-output diagram of the subsequent action prediction network in this disclosure is shown, as follows: Figure 2 Figure 3 As shown:

[0047] The motion data and first state data at the first moment are input into the current motion prediction network to obtain the first predicted motion at the first moment. In the figure, S represents the motion data at the first moment, and a represents the first predicted motion. The second state data is input into the subsequent motion prediction network to obtain the second predicted motion at the second moment. In the figure, S' represents the second state data, and a' represents the predicted motion at the second moment. Based on the above configuration, the current motion prediction network and the subsequent motion prediction network can respectively predict the vehicle motion data based on the state data.

[0048] S103: Based on the prediction results of the current action prediction network and the subsequent action prediction network, respectively, the execution performance is predicted using the first network and the second network.

[0049] Figure 4 A deployment diagram of each network in an embodiment of this disclosure is shown. For example... Figure 4As shown, the network includes a current action prediction network, a subsequent action prediction network, a first network connected to the current action prediction network, and a second network connected to the subsequent action prediction network. The first network contains at least one first subnetwork, and the second network contains at least one second subnetwork. Each first subnetwork and each second subnetwork is connected to the current action prediction network and the subsequent action prediction network, respectively. Furthermore, data paths exist between the current action prediction network and the subsequent action prediction network, as well as between the first network and the second network. Based on this configuration, the networks can communicate with each other to optimize them according to an optimization strategy.

[0050] Figure 5 A flowchart illustrating a method for predicting the performance of prediction results in an embodiment of this disclosure is shown. As shown, predicting the performance of the current action prediction network based on the prediction results of the first network includes: the current action prediction network sending first estimated action and first state data to each first sub-network; each first sub-network outputting a first prediction result based on the first estimated action and first state data. Predicting the performance of the subsequent action prediction network based on the prediction results of the second network includes: the subsequent action prediction network sending a second estimated action plus noise data and second state data to each second sub-network; each second sub-network outputting a second prediction result based on the second estimated action and second state data. Based on the above configuration, the vehicle action training model provided by this disclosure can evaluate the prediction results of the current action prediction network and the subsequent action prediction network. Based on the evaluation value, it is possible to effectively adjust the parameters of each network to optimize the relevant networks and train a prediction model with good vehicle action prediction performance. Furthermore, the addition of noise data effectively improves the compatibility of the vehicle action prediction model in this disclosure.

[0051] S104: Adjust the parameters of the current action prediction network, the subsequent action prediction network, the first network, and the second network based on the performance prediction results.

[0052] Figure 6 This diagram illustrates a method for adjusting network parameters according to an embodiment of the present disclosure. As shown in the figure, the method for adjusting the parameters of the current action prediction network, the subsequent action prediction network, the first network, and the second network based on the performance prediction results includes:

[0053] S201: Determine the target value based on each second prediction result.

[0054] In one embodiment, the second network has three second sub-networks, and the target value is y.

[0055]

[0056] In this context, the "terminal" label indicates whether a training session has terminated, and γ is the reward decay coefficient. For each second subnetwork, r represents the second prediction result, where r is the preset reward function.

[0057] S202: Determine the model loss based on the target value and each first prediction result.

[0058] In one embodiment, the second network has three second sub-networks, and the model loss corresponding to each first sub-network is . but in, This represents the first prediction result for each first subnetwork.

[0059] S203: Based on the model loss and the first prediction result, adjust the parameters of the current action prediction network, the subsequent action prediction network, the first network, and the second network.

[0060] Figure 7 This diagram illustrates a method for adjusting network parameters according to an embodiment of the present disclosure. As shown in the figure, the method for adjusting the parameters of the current action prediction network, the subsequent action prediction network, the first network, and the second network based on the performance prediction results includes:

[0061] S301: Adjust the parameters of the current action prediction network in the direction of gradient ascent based on the first prediction results.

[0062] In one embodiment, suppose the first network has three first sub-networks and the second network has three second sub-networks, then θ μ ←θ μ +βθ μ G a It is the update gradient of the current action prediction network, θ μ These are the parameters of the current action prediction network, and β is an adjustable hyperparameter to control the speed of gradient updates.

[0063] S302: Adjust the parameters of the first network based on the model loss.

[0064] In one embodiment, suppose the first network has three first sub-networks and the second network has three second sub-networks, then (i = 1, 2, 3); where G c It updates the gradient. These are the parameters of the first subnetwork, and α is an adjustable hyperparameter to control the speed of gradient updates. This represents the first prediction result for each first subnetwork.

[0065] S303: Based on the parameters of the current action prediction network and the parameters of the first network, adjust the parameters of the subsequent action prediction network and the parameters of the second network accordingly.

[0066] In one embodiment, suppose the first network has three first sub-networks and the second network has three second sub-networks, then θ μ ′←ηθ μ +(1-η)θ μ ′,(i=1,2,3); where, These are the parameters of the second self-network, θ μ ′ represents the parameters of the subsequent action prediction network, and η is the preset learning rate, which ranges from 0 to 1.

[0067] Based on the above configuration, the network parameters of the vehicle motion prediction model disclosed herein can be effectively adjusted. By setting multiple first sub-networks and multiple second sub-networks, the stability of the update of each network parameter can be ensured. At the same time, the adjustable hyperparameters and learning rate parameters can facilitate technicians to make timely adjustments based on the training status of the model.

[0068] Figure 8 A schematic flowchart of a vehicle motion control method according to an embodiment of the present disclosure is shown. As shown in the figure, the method includes:

[0069] S401: Input the vehicle's state data at the target time into the subsequent action prediction network in the vehicle action prediction model to obtain the predicted action at the target time.

[0070] In the embodiments of this disclosure, the vehicle motion prediction model is trained by the training method of the vehicle motion prediction model involved in any of the above embodiments.

[0071] S402: Solve for the parameters in the trajectory polynomial based on the predicted motion.

[0072] In embodiments of this disclosure, the trajectory polynomial generally includes a longitudinal trajectory generating fourth-order polynomial and a transverse trajectory generating fifth-order polynomial: traj long =a0+a1t+a2t 2 +a3t 3 +a4t 4 traj lat =b0+b1t+b2t 2 +b3t 3 +b4t 4 +b5t 5 , t∈{0,dt,2dt,...,a d}, where traj long It is the vehicle's longitudinal trajectory, traj latdt represents the vehicle's lateral trajectory, t is the trajectory planning time, and dt is typically taken as 0.2 seconds. When solving for the longitudinal trajectory, a... d For a lod When solving for the lateral trajectory, a d For a lad Let a0 be the vehicle's current longitudinal position, a1 be the vehicle's current longitudinal velocity, a2 be the vehicle's current longitudinal acceleration, b0 be the vehicle's current lateral position, b1 be the vehicle's current lateral velocity, b2 be the vehicle's current lateral acceleration, and a3, a4, b3, b4, and b5 be the parameters that need to be solved. Based on the prediction results of the vehicle motion prediction model, the parameters that need to be solved in the above-mentioned fourth-order polynomial for generating the longitudinal trajectory and the fifth-order polynomial for generating the lateral trajectory can be obtained.

[0073] S403: Generates control commands for the vehicle based on the trajectory polynomial.

[0074] In the embodiments of this disclosure, the corresponding vehicle control action can be obtained by differentiating the trajectory polynomial: traj long '=a lv traj long =a la traj lat =a lp traj lat '=a lav traj lat =a laa ; where a lv It is the longitudinal speed of the controlled vehicle, a la It controls the target longitudinal acceleration of the vehicle, a lp It refers to the lateral position of the vehicle, a lav It controls the target lateral speed of the vehicle, a laa It controls the target lateral acceleration of the vehicle. Based on the above parameters, control commands for the vehicle can be generated, including the target lateral and longitudinal velocities and the steering wheel angle, in order to achieve vehicle motion control.

[0075] Based on the above configuration, real-time control of the vehicle can be achieved. The vehicle control command is based on the vehicle motion prediction model training method provided in this disclosure. Since the high-precision map scene data is used during model training and the training data is rich enough, the vehicle control command can adapt to relatively complex and changeable traffic scenarios.

[0076] Those skilled in the art will understand that, in the above-described method of the specific implementation, the order in which the steps are written does not imply a strict execution order and does not constitute any limitation on the implementation process. The specific execution order of each step should be determined by its function and possible internal logic.

[0077] It is understood that the various method embodiments mentioned above in this disclosure can be combined with each other to form combined embodiments without violating the principle and logic. Due to space limitations, this disclosure will not elaborate further.

[0078] In some embodiments, the functions or modules of the apparatus provided in this disclosure can be used to execute the vehicle control action model training method described in the above embodiments. The specific implementation can be referred to the description of the relevant embodiments, which will not be repeated here for the sake of brevity.

[0079] Figure 9 A block diagram of a vehicle motion prediction model training apparatus according to an embodiment of the present disclosure is shown, such as Figure 9 As shown, the above-mentioned device includes:

[0080] The data acquisition module 101 is used to acquire training data, which includes action data and first state data at a first time point, and second state data at a second time point, where the second time point is the time point following the first time point.

[0081] The first processing module 102 is used to input the action data and the first state data at the first moment into the current action prediction network, and to input the second state data into the subsequent action prediction network.

[0082] The second processing module 103 performs performance prediction based on the prediction results of the current action prediction network and the subsequent action prediction network, respectively, using the first network and the second network.

[0083] The optimization module 104 is used to adjust the parameters of the current action prediction network, the subsequent action prediction network, the first network, and the second network based on the performance prediction results.

[0084] The training data is determined based on vehicle motion data generated by the driver, vehicle motion data generated by the autonomous driving simulation platform, vehicle motion data generated by autonomous driving trajectory planning, and high-precision maps.

[0085] In some possible implementations, the first network includes at least one first sub-network, the second network includes at least one second sub-network, each of the first sub-networks is connected to the current action prediction network, and each of the second sub-networks is connected to the subsequent action prediction network.

[0086] In some possible implementations, the first processing module is further configured to input the action data and the first state data at the first moment into the current action prediction network to obtain the first predicted action at the first moment; and input the second state data into the subsequent action prediction network to obtain the second predicted action at the second moment.

[0087] In some possible implementations, the second processing module is further configured to input the first estimated action and the first state data into each first sub-network to obtain a corresponding first prediction result, wherein the first prediction result represents the execution result corresponding to the vehicle performing the first estimated action under the condition corresponding to the first state data; and to input the second estimated action and the second state data into each second sub-network to obtain a corresponding second prediction result, wherein the second prediction result represents the execution result corresponding to the vehicle performing the second estimated action under the condition corresponding to the second state data.

[0088] In some possible implementations, the optimization module is further configured to determine a target value based on each second prediction result; determine a model loss based on the target value and each first prediction result; and adjust the parameters of the current action prediction network, the subsequent action prediction network, the first network, and the second network based on the model loss and the first prediction results.

[0089] In some possible implementations, the optimization module is further configured to adjust the parameters of the current action prediction network in the direction of gradient ascent based on each first prediction result; adjust the parameters of the first network based on the model loss; and adjust the parameters of the subsequent action prediction network and the second network respectively according to the parameters of the current action prediction network and the parameters of the first network.

[0090] Figure 10 A block diagram of a vehicle motion control device according to an embodiment of the present disclosure is shown, such as Figure 10 As shown, the above-mentioned device includes:

[0091] Prediction module 201: Used to input the vehicle's state data at the target time into the subsequent action prediction network in the vehicle action prediction model to obtain the predicted action at the target time.

[0092] Parameter determination module 202: used to solve the parameters in the trajectory polynomial based on the estimated action.

[0093] Control command generation module 203: Generates control commands for the vehicle based on the trajectory polynomial.

[0094] The vehicle motion prediction model mentioned above was trained using the same training method described earlier.

[0095] This disclosure also proposes a computer-readable storage medium storing at least one instruction or at least one program segment. When the at least one instruction or at least one program segment is loaded and executed by a processor, it implements the vehicle motion prediction model training method or the vehicle control method. The computer-readable storage medium may be a non-volatile computer-readable storage medium.

[0096] This disclosure also proposes an electronic device, including: a processor; and a memory for storing processor-executable instructions; wherein the processor is configured for the vehicle motion prediction model training method or the vehicle control method.

[0097] Electronic devices can be provided as terminals, servers, or other forms of devices.

[0098] Figure 11 This diagram illustrates a block diagram of an electronic device according to an embodiment of the present disclosure. For example, the electronic device 800 may be a mobile phone, computer, digital broadcasting terminal, messaging device, game console, tablet device, medical device, fitness equipment, personal digital assistant, or other terminal.

[0099] Reference Figure 10 The electronic device 800 may include one or more of the following components: a processing component 802, a memory 804, a power supply component 806, a multimedia component 808, an audio component 810, an input / output (I / O) interface 812, a sensor component 814, and a communication component 816.

[0100] Processing component 802 typically controls the overall operation of electronic device 800, such as operations associated with display, telephone calls, data communication, camera operation, and recording operations. Processing component 802 may include one or more processors 820 to execute instructions to complete all or part of the steps of the methods described above. Furthermore, processing component 802 may include one or more modules to facilitate interaction between processing component 802 and other components. For example, processing component 802 may include a multimedia module to facilitate interaction between multimedia component 808 and processing component 802.

[0101] Memory 804 is configured to store various types of data to support the operation of electronic device 800. Examples of this data include instructions for any application or method operating on electronic device 800, contact data, phonebook data, messages, pictures, videos, etc. Memory 804 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0102] Power supply component 806 provides power to various components of electronic device 800. Power supply component 806 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to electronic device 800.

[0103] Multimedia component 808 includes a screen that provides an output interface between the aforementioned electronic device 800 and the user. In some embodiments, the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touchscreen to receive input signals from the user. The touch panel includes one or more touch sensors to sense touches, swipes, and gestures on the touch panel. The touch sensors may sense not only the boundaries of touch or swipe actions but also the duration and pressure associated with the touch or swipe operation. In some embodiments, multimedia component 808 includes a front-facing camera and / or a rear-facing camera. When the electronic device 800 is in an operating mode, such as a shooting mode or a video mode, the front-facing camera and / or the rear-facing camera may receive external multimedia data. Each front-facing camera and rear-facing camera may be a fixed optical lens system or have focal length and optical zoom capabilities.

[0104] Audio component 810 is configured to output and / or input audio signals. For example, audio component 810 includes a microphone (MIC) configured to receive external audio signals when electronic device 800 is in an operating mode, such as call mode, recording mode, and voice recognition mode. The received audio signals may be further stored in memory 804 or transmitted via communication component 816. In some embodiments, audio component 810 also includes a speaker for outputting audio signals.

[0105] I / O interface 812 provides an interface between processing component 802 and peripheral interface modules, such as keyboards, click wheels, buttons, etc. These buttons may include, but are not limited to, home buttons, volume buttons, power buttons, and lock buttons.

[0106] Sensor assembly 814 includes one or more sensors for providing state assessments of various aspects of electronic device 800. For example, sensor assembly 814 can detect the on / off state of electronic device 800, the relative positioning of components such as the display and keypad of electronic device 800, changes in position of electronic device 800 or a component of electronic device 800, the presence or absence of user contact with electronic device 800, orientation or acceleration / deceleration of electronic device 800, and temperature changes of electronic device 800. Sensor assembly 814 may include a proximity sensor configured to detect the presence of nearby objects without any physical contact. Sensor assembly 814 may also include a light sensor, such as a CMOS or CCD image sensor, for use in imaging applications. In some embodiments, sensor assembly 814 may also include an accelerometer, gyroscope, magnetometer, pressure sensor, or temperature sensor.

[0107] Communication component 816 is configured to facilitate wired or wireless communication between electronic device 800 and other devices. Electronic device 800 can access wireless networks based on communication standards, such as WiFi, 2G, 3G, 4G, 5G, or combinations thereof. In one exemplary embodiment, communication component 816 receives broadcast signals or broadcast-related information from an external broadcast management system via a broadcast channel. In one exemplary embodiment, the aforementioned communication component 816 also includes a near-field communication (NFC) module to facilitate short-range communication. For example, the NFC module may be implemented based on radio frequency identification (RFID) technology, Infrared Data Association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology, and other technologies.

[0108] In an exemplary embodiment, the electronic device 800 may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to execute the above-described vehicle motion prediction model training method or vehicle control method.

[0109] In an exemplary embodiment, a non-volatile computer-readable storage medium is also provided, such as a memory 804 including computer program instructions, which can be executed by the processor 820 of the electronic device 800 to complete the above-described vehicle motion prediction model training method or vehicle control method.

[0110] Figure 12 A block diagram of another electronic device according to an embodiment of the present disclosure is shown. For example, electronic device 1900 may be provided as a server. (Refer to...) Figure 12 The electronic device 1900 includes a processing component 1922, which further includes one or more processors, and memory resources represented by memory 1932 for storing instructions, such as application programs, that can be executed by the processing component 1922. The application programs stored in memory 1932 may include one or more modules, each corresponding to a set of instructions. Furthermore, the processing component 1922 is configured to execute instructions to perform the aforementioned vehicle motion prediction model training method or vehicle control method.

[0111] Electronic device 1900 may also include a power supply component 1926 configured to perform power management of electronic device 1900, a wired or wireless network interface 1950 configured to connect electronic device 1900 to a network, and an input / output (I / O) interface 1958. Electronic device 1900 can operate on an operating system stored in memory 1932, such as Windows Server™, Mac OS X™, Unix™, Linux™, FreeBSD™, or similar.

[0112] In an exemplary embodiment, a non-volatile computer-readable storage medium is also provided, such as a memory 1932 including computer program instructions, which can be executed by the processing component 1922 of the electronic device 1900 to complete the vehicle motion prediction model training method or the vehicle control method described above.

[0113] This disclosure can be a system, method, and / or computer program product. A computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for causing a processor to implement various aspects of this disclosure.

[0114] Computer-readable storage media can be tangible devices capable of holding and storing instructions for use by an instruction execution device. Computer-readable storage media can be, for example—but not limited to—electrical storage devices, magnetic storage devices, optical storage devices, electromagnetic storage devices, semiconductor storage devices, or any suitable combination thereof. More specific examples (a non-exhaustive list) of computer-readable storage media include: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), static random access memory (SRAM), portable compact disc read-only memory (CD-ROM), digital multifunction disc (DVD), memory sticks, floppy disks, mechanical encoding devices, such as punch cards or recessed protrusions storing instructions thereon, and any suitable combination thereof. The computer-readable storage media used herein are not to be construed as transient signals themselves, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through waveguides or other transmission media (e.g., light pulses through fiber optic cables), or electrical signals transmitted through wires.

[0115] The computer-readable program instructions described herein can be downloaded from computer-readable storage media to various computing / processing devices, or downloaded via a network, such as the Internet, local area network, wide area network, and / or wireless network, to an external computer or external storage device. The network may include copper transmission cables, fiber optic transmission, wireless transmission, routers, firewalls, switches, gateway computers, and / or edge servers. A network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards them to the computer-readable storage media in the respective computing / processing device.

[0116] Computer program instructions used to perform the operations of this disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, status setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Smalltalk, C++, etc., and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The computer-readable program instructions may execute entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or may be connected to an external computer (e.g., via the Internet using an Internet service provider). In some embodiments, electronic circuitry, such as programmable logic circuitry, field-programmable gate arrays (FPGAs), or programmable logic arrays (PLAs), is personalized by utilizing the status information of the computer-readable program instructions to implement various aspects of this disclosure.

[0117] Various aspects of this disclosure are described herein with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this disclosure. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer-readable program instructions.

[0118] These computer-readable program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine such that, when executed by the processor of the computer or other programmable data processing apparatus, they create means for implementing the functions / actions specified in one or more blocks of the flowchart and / or block diagram. These computer-readable program instructions can also be stored in a computer-readable storage medium that causes a computer, programmable data processing apparatus, and / or other device to operate in a particular manner; thus, the computer-readable medium storing the instructions comprises an article of manufacture that includes instructions for implementing aspects of the functions / actions specified in one or more blocks of the flowchart and / or block diagram.

[0119] Computer-readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, thereby causing the instructions executed on the computer, other programmable data processing apparatus, or other device to perform the functions / actions specified in one or more boxes of a flowchart and / or block diagram.

[0120] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of an instruction, which includes one or more executable instructions for implementing a specified logical function. In some alternative implementations, the functions specified in the blocks may occur in a different order than those specified in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0121] The various embodiments of this disclosure have been described above. These descriptions are exemplary and not exhaustive, nor are they limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is chosen to best explain the principles, practical application, or technical improvements to the embodiments in the market, or to enable others skilled in the art to understand the embodiments disclosed herein.

Claims

1. A method for training a vehicle motion prediction model, the vehicle motion prediction model being used for vehicle control; the vehicle motion prediction model comprising a current motion prediction network and a subsequent motion prediction network, characterized in that, The method includes: Acquire training data, which includes action data and first state data at a first time point, and second state data at a second time point, where the second time point is the time point following the first time point; The action data and the first state data at the first moment are input into the current action prediction network to obtain the first predicted action at the first moment, and the second state data is input into the subsequent action prediction network to obtain the second predicted action at the second moment. The first predicted action and the first state data are input into each first subnetwork of the first network to obtain the corresponding first prediction result. The first prediction result represents the execution result of the vehicle executing the first predicted action under the condition corresponding to the first state data. The second predicted action and the second state data are input into each second sub-network of the second network to obtain the corresponding second prediction result. The second prediction result represents the execution result of the vehicle executing the second predicted action under the condition corresponding to the second state data. The target value is determined based on each of the second prediction results; The model loss is determined based on the target value and each of the first prediction results; The parameters of the current action prediction network, the subsequent action prediction network, the first network, and the second network are adjusted based on the model loss and the first prediction result. The training data is determined based on vehicle motion data generated by the driver, vehicle motion data generated by the autonomous driving simulation platform, vehicle motion data generated by autonomous driving trajectory planning, and high-precision maps.

2. The method according to claim 1, characterized in that, The first network includes at least one first sub-network, and the second network includes at least one second sub-network. Each first sub-network is connected to the current action prediction network, and each second sub-network is connected to the subsequent action prediction network.

3. The method according to claim 1, characterized in that, The method further includes: The parameters of the current action prediction network are adjusted in the direction of gradient ascent based on each of the first prediction results; Based on the model loss, adjust the parameters of the first network; Based on the parameters of the current action prediction network and the parameters of the first network, the parameters of the subsequent action prediction network and the parameters of the second network are adjusted accordingly.

4. A vehicle control method, characterized in that, The method includes: The vehicle's state data at the target time is input into the subsequent action prediction network of the vehicle action prediction model to obtain the predicted action at the target time. The vehicle action prediction model is trained using the training method of the vehicle action prediction model according to any one of claims 1 to 3. The parameters in the trajectory polynomial are solved based on the predicted action, and the trajectory polynomial is used to describe the driving trajectory of the vehicle. Based on the trajectory polynomial, control commands for the vehicle are generated.

5. A training device for a vehicle motion prediction model, characterized in that, The vehicle motion prediction model is used for vehicle control; the device includes: The data acquisition module is used to acquire training data, which includes action data and first state data at a first moment, and second state data at a second moment, where the second moment is the moment following the first moment. The first processing module is used to input the action data and the first state data at the first moment into the current action prediction network to obtain the first predicted action at the first moment, and to input the second state data into the subsequent action prediction network to obtain the second predicted action at the second moment. The second processing module is configured to input the first estimated action and the first state data into each first sub-network of the first network to obtain a corresponding first prediction result, wherein the first prediction result represents the execution result corresponding to the vehicle executing the first estimated action under the condition corresponding to the first state data; and to input the second estimated action and the second state data into each second sub-network of the second network to obtain a corresponding second prediction result, wherein the second prediction result represents the execution result corresponding to the vehicle executing the second estimated action under the condition corresponding to the second state data. An optimization module is used to determine a target value based on each of the second prediction results; determine a model loss based on the target value and each of the first prediction results; and adjust the parameters of the current action prediction network, the subsequent action prediction network, the first network, and the second network based on the model loss and the first prediction results. The training data is determined based on vehicle motion data generated by the driver, vehicle motion data generated by the autonomous driving simulation platform, vehicle motion data generated by autonomous driving trajectory planning, and high-precision maps.

6. An electronic device, characterized in that, It includes 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 at least one processor implements the training method for the vehicle motion prediction model as described in any one of claims 1 to 3, or the vehicle control method as described in claim 4, by executing the instructions stored in the memory.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores at least one instruction or at least one program, which is loaded and executed by a processor to implement the training method for the vehicle motion prediction model as described in any one of claims 1 to 3, or the vehicle control method as described in claim 4.

Images