Path planning method, device and equipment of autonomous vehicle, vehicle and medium

Abstract

The application provides a path planning method, device and equipment of an automatic driving vehicle, a vehicle and a medium. A plurality of frames of driving data are converted into vectors and input into a path planning vector neural network to obtain a reference driving path of the automatic driving vehicle in a future period of time. The path planning vector neural network is updated iteratively online according to an internal neural network having the same structure. The reference driving path is optimized according to a preset optimization target to obtain an optimal driving path when the optimization target is maximum. The optimal driving path is used as a path planning result of the automatic driving vehicle, the adaptability of the automatic driving vehicle in an unknown working condition of a traffic driving scene is enhanced, and the calculation speed and the update iteration speed of the automatic driving path planning algorithm are improved.

Description

Technical Field

[0001] This invention relates to the field of autonomous driving technology, and in particular to a path planning method, apparatus, device, vehicle, and medium for autonomous vehicles. Background Technology

[0002] In the field of autonomous driving, path planning algorithms are crucial for planning a reference path for the vehicle's future travel in the current driving scenario, serving as the basis for subsequent vehicle motion control. Existing autonomous driving path planning algorithms include heuristic search, random tree search, curve sampling, and rolling time-domain optimization. These algorithms are typically only applicable to known design conditions. In actual operation of autonomous vehicles, unknown conditions such as sudden path changes may occur, making these algorithms less adaptable to traffic driving scenarios with poor computational speed. Furthermore, existing autonomous driving path planning algorithms are fixed after deployment in the vehicle controller. Due to limited resources in the vehicle controller, retraining and automatic updates are not possible. The algorithms themselves do not update automatically; instead, they need to be manually retrained locally and then redeployed to the controller to update and improve the path planning algorithm in the vehicle controller. Therefore, the iteration speed of existing autonomous driving path planning algorithms is slow, hindering continuous optimization. Summary of the Invention

[0003] This invention provides a path planning method and apparatus for autonomous vehicles, which addresses the shortcomings of existing autonomous driving path planning algorithms, such as weak adaptability to unknown traffic driving scenarios, slow calculation speed, and slow algorithm update iteration speed.

[0004] This invention provides a path planning method for autonomous vehicles, comprising:

[0005] After vectorizing and converting multiple frames of driving data, the data is input into a path planning vector neural network to obtain a reference driving path for the autonomous vehicle within a certain period of time. The path planning vector neural network is updated online based on an internal neural network with the same structure.

[0006] The reference driving path is optimized according to the preset optimization target to obtain the optimal driving path when the optimization target is maximized. The optimal driving path is then used as the path planning result for the autonomous vehicle.

[0007] According to a path planning method for an autonomous vehicle provided by the present invention, the step of inputting multi-frame driving data into a path planning vector neural network after vectorization to obtain a reference driving path for the autonomous vehicle within a future period includes:

[0008] The driving data acquired at the current moment is input into the path planning vector neural network to obtain the coordinate position of the next moment in the traffic road coordinate system, with the current position as the origin;

[0009] By inputting the driving data at each future moment into the path planning vector neural network, the coordinate position at each moment in the future within a certain period of time is obtained in the traffic road coordinate system;

[0010] The reference driving path of the autonomous vehicle is obtained based on the coordinates of the vehicle at each moment in the future.

[0011] According to a path planning method for an autonomous vehicle provided by the present invention, the step of optimizing the reference driving path according to a preset optimization objective to obtain the optimal driving path when the optimization objective is maximized includes:

[0012] The driving process of autonomous vehicles is discretized in time, and a Markov decision model is constructed based on the discretized vehicle parameters. The Markov decision model uses the vehicle parameters and surrounding vehicle parameters at each time step as the decision state, and the vehicle position at the next time step output by the path planning vector neural network as the decision action.

[0013] The real-time driving parameters during the driving process are input into the Markov decision model, and the decision action is optimized according to the preset optimization objective to obtain the optimal driving path when the optimization objective is maximized.

[0014] According to the present invention, a path planning method for an autonomous vehicle is provided, wherein the vehicle parameters include at least one of vehicle position, speed and acceleration;

[0015] The surrounding vehicle parameters include at least one of the surrounding vehicle positions and speeds.

[0016] According to a path planning method for an autonomous vehicle provided by the present invention, the optimization objective includes:

[0017] The safe distance between the planned route and surrounding traffic participants, and the compliance of the planned route with road traffic rules.

[0018] According to a path planning method for an autonomous vehicle provided by the present invention, the path planning vector neural network is iteratively updated online based on an internal neural network having the same structure, including:

[0019] Based on the training data, the parameters of the internal neural network with the same structure as the path planning vector neural network are optimized to obtain the updated parameters of the network. The training data includes model simulation data, historical driving data, and real-time driving data.

[0020] According to a preset synchronization period, the parameter change rate of the internal neural network is obtained based on the parameters after the network update, and the parameter change rate of the internal neural network is adjusted according to a preset parameter change compression ratio to obtain the parameter change rate of the path planning vector neural network. The parameter change rate of the path planning vector neural network is less than the parameter change rate of the internal neural network.

[0021] The parameters of the path planning vector neural network are adjusted according to the rate of change of the parameters of the path planning vector neural network in order to perform online iterative updates of the path planning vector neural network.

[0022] According to a path planning method for an autonomous vehicle provided by the present invention, the driving data includes:

[0023] One or more of the following: autonomous driving maps, environmental perception data, behavior prediction data, and vehicle chassis dynamics data.

[0024] According to a path planning method for an autonomous vehicle provided by the present invention, the vehicle chassis dynamics data includes the vehicle speed, and a method for vectorizing the current vehicle speed includes:

[0025] Based on the established vehicle dynamics coordinate system, the vehicle speed is decomposed into the velocity direction in the direction the vehicle is facing and the velocity direction in the direction the vehicle is sideways. The vector data of the vehicle speed is obtained based on the velocity direction and the specific value of the velocity corresponding to each velocity direction.

[0026] The present invention also provides a path planning device for an autonomous vehicle, comprising:

[0027] The acquisition module is used to convert multi-frame driving data into vectors and input them into the path planning vector neural network to obtain the reference driving path of the autonomous vehicle in the future.

[0028] The planning module is used to optimize the reference driving path according to the preset optimization target, obtain the optimal driving path when the optimization target is maximized, and use the optimal driving path as the path planning result of the autonomous vehicle.

[0029] An update module is used for online iterative updates of the path planning vector neural network, which performs online iterative updates based on an internal neural network with the same structure.

[0030] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the path planning method for an autonomous vehicle as described above.

[0031] The present invention also provides an autonomous driving vehicle, including the aforementioned electronic equipment.

[0032] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the path planning method for an autonomous vehicle as described above.

[0033] This invention provides a path planning method, apparatus, device, vehicle, and medium for autonomous vehicles. By vectorizing and inputting multi-frame driving data into a path planning vector neural network, a reference driving path for the autonomous vehicle over a future period is obtained. The path planning vector neural network is iteratively updated online based on an internal neural network with the same structure. The reference driving path is optimized according to a preset optimization objective to obtain the optimal driving path when the optimization objective is maximized. This optimal driving path is used as the path planning result for the autonomous vehicle, enhancing its adaptability in unknown traffic driving scenarios and improving the computational speed and iteration speed of the autonomous driving path planning algorithm. Attached Figure Description

[0034] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0035] Figure 1 This is one of the flowcharts illustrating the path planning method for autonomous vehicles provided by the present invention;

[0036] Figure 2 This is the second flowchart illustrating the path planning method for autonomous vehicles provided by this invention.

[0037] Figure 3 This is the third flowchart of the path planning method for autonomous vehicles provided by the present invention;

[0038] Figure 4 This is the fourth flowchart of the path planning method for autonomous vehicles provided by the present invention;

[0039] Figure 5 This is a schematic diagram of the path planning device for autonomous vehicles provided by the present invention;

[0040] Figure 6 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0041] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0042] Figure 1 A flowchart of the path planning method for autonomous vehicles provided in an embodiment of the present invention is shown below. Figure 1 As shown, the path planning method for autonomous vehicles provided in this embodiment of the invention includes:

[0043] Step 101: After vectorizing the multi-frame driving data, input it into the path planning vector neural network to obtain the reference driving path of the autonomous vehicle in the future. The path planning vector neural network is updated online based on the internal neural network with the same structure.

[0044] In this embodiment of the invention, the path planning vector neural network can be implemented using any suitable neural network model, including but not limited to the Transformer model.

[0045] This invention reduces the computational load of path planning methods and improves their computational speed by vectorizing driving data. At the same time, it optimizes the parameters of the internal neural network based on continuously updated driving data and synchronizes the optimized parameters of the internal neural network to the path planning vector neural network, thereby enhancing the adaptability of autonomous vehicles in traffic driving scenarios with unknown conditions.

[0046] Step 102: Optimize the reference driving path according to the preset optimization target to obtain the optimal driving path when the optimization target is maximized, and use the optimal driving path as the path planning result of the autonomous vehicle.

[0047] Traditional autonomous driving path planning algorithms are typically only applicable to known design conditions. In actual operation, autonomous vehicles encounter unknown conditions such as sudden path changes, making them less adaptable to unforeseen traffic scenarios and resulting in slow computation speeds. Furthermore, existing autonomous driving path planning algorithms are fixed once deployed in the vehicle controller; they do not update automatically but require manual redeployment to update and improve the algorithms already loaded into the controller. Therefore, the iteration speed of existing autonomous driving path planning algorithms is slow, hindering continuous optimization.

[0048] This invention provides a path planning method for autonomous vehicles. By vectorizing multi-frame driving data and inputting it into a path planning vector neural network, a reference driving path for the autonomous vehicle over a future period is obtained. The path planning vector neural network is iteratively updated online based on an internal neural network with the same structure. The reference driving path is optimized according to a preset optimization objective to obtain the optimal driving path when the optimization objective is maximized. This optimal driving path is used as the path planning result for the autonomous vehicle, enhancing its adaptability in unknown traffic driving scenarios and improving the computational speed and iteration speed of the autonomous driving path planning algorithm.

[0049] Based on any of the above embodiments, such as Figure 2 As shown, after vectorizing multi-frame driving data, it is input into a path planning vector neural network to obtain the reference driving path of the autonomous vehicle within a future period, including:

[0050] Step 201: Input the driving data acquired at the current moment into the path planning vector neural network to obtain the coordinate position of the next moment in the traffic road coordinate system, with the current position as the origin;

[0051] In this embodiment of the invention, driving data includes, but is not limited to: autonomous driving maps, environmental perception data, behavior prediction data, vehicle chassis dynamics data, etc.

[0052] Autonomous driving maps include two types of data. The first type is road data, such as lane information including the position, type, width, slope, and curvature of lane lines. The second type is data on fixed objects around the lanes, such as traffic signs.

[0053] Environmental perception data includes the current position of the autonomous vehicle in the traffic road coordinate system and the physical environment information around the autonomous vehicle, including data such as the relative distance, relative speed, angle, and direction of motion between the autonomous vehicle and surrounding objects.

[0054] The behavior prediction data includes two types of data: the first type is the path prediction of surrounding vehicles, which serves as the basis for the subsequent second type is the path prediction of pedestrians.

[0055] Vehicle chassis dynamics data includes vehicle speed, vehicle acceleration, direction of motion, angle, and other data.

[0056] Driving data is obtained from onboard sensors and communication information, providing data support for autonomous vehicle path planning methods and ensuring the safe operation of autonomous vehicles.

[0057] In this embodiment of the invention, the input to the path planning vector neural network is a two-dimensional matrix based on driving data. This is achieved by vectorizing multiple consecutive frames (e.g., three frames) of driving data to obtain corresponding vector data, which are then horizontally concatenated to form the two-dimensional matrix. Each column of the two-dimensional matrix represents a frame of data, and each column is obtained by vertically concatenating the vector data corresponding to the driving data. Vectorization refers to converting the data into a data matrix representation composed of vector data, which is obtained based on the time or spatial direction of the data and the specific data size corresponding to each direction.

[0058] Specifically, in this embodiment of the invention, the vehicle chassis dynamics data includes the vehicle speed, and the method for vectorizing the current vehicle speed includes:

[0059] Based on the established vehicle dynamics coordinate system, the vehicle speed is decomposed into the velocity direction in the direction the vehicle is facing and the velocity direction in the direction the vehicle is sideways. The vector data of the vehicle speed is obtained based on the velocity direction and the specific value of the velocity corresponding to each velocity direction.

[0060] In this embodiment of the invention, a traffic road coordinate system is established with the current location of the autonomous vehicle as the origin, the center line of the driving road as the vertical axis, and the normal of the vertical axis as the horizontal axis. The output of the path planning vector neural network is the coordinate position of the next moment with the current location as the origin in the traffic road coordinate system.

[0061] Step 202: Input the driving data at each future moment into the path planning vector neural network to obtain the coordinate position at each moment in the future under the traffic road coordinate system;

[0062] In some embodiments of the present invention, the distances between the coordinates at each time point are equal.

[0063] Step 203: Obtain the reference driving path of the autonomous vehicle based on the coordinates of each time point in the future.

[0064] By using a path planning vector neural network in traffic driving scenarios with unknown operating conditions, we can avoid using complex algorithms and improve online computation efficiency.

[0065] Based on any of the above embodiments, such as Figure 3 As shown, the reference driving path is optimized according to the preset optimization objective to obtain the optimal driving path when the optimization objective is maximized, including:

[0066] Step 301: Discretize the driving process of the autonomous vehicle in time, and construct a Markov decision model based on the vehicle parameters after time discretization. The Markov decision model uses the vehicle parameters and surrounding vehicle parameters at each time step as the decision state, and uses the vehicle position at the next time step output by the path planning vector neural network as the decision action.

[0067] In this embodiment of the invention, the parameters of the vehicle itself include its position, speed, and acceleration; the parameters of the surrounding vehicles include their positions and speeds.

[0068] Step 302: Input the real-time driving parameters during the driving process into the Markov decision model, optimize the decision action according to the preset optimization objective, and obtain the optimal driving path when the optimization objective is maximized.

[0069] In this embodiment of the invention, optimizing objective J includes:

[0070] Safe distance between planned route and surrounding traffic participants J d And the compliance of the planned route with road traffic rules J c .

[0071] Among them, J d For example, J d = 1 / d1 + 1 / d2 + 1 / d3…, d i J is the distance between the waypoint and the i-th traffic participant. c For example,

[0072]

[0073] It should be noted that this invention does not limit the definition and value of the safe distance between the planned path and surrounding traffic participants, or the compliance of the planned path with road traffic rules. Those skilled in the art can select these values ​​according to actual needs.

[0074] Specifically, autonomous vehicles use a path planning vector neural network to predict the next moment's coordinates based on driving data. If the predicted distance between the next moment's coordinates and the pedestrian is less than a preset distance, failing to meet road traffic safety requirements, the Markov model's decision-making will be adjusted, and the prediction result will be re-output until a predicted position that meets road safety requirements is reached, ultimately generating the optimal driving path. If the predicted next moment's coordinates enter the non-motorized vehicle lane, violating road traffic rules, the Markov model's decision-making will be adjusted again, and the prediction result will be re-output until a predicted position that meets road traffic rules is reached, ultimately generating the optimal driving path.

[0075] By discretizing the driving process of autonomous vehicles in time and continuously optimizing the predicted location points of autonomous vehicles through Markov models, the online updating of the autonomous vehicle's path is ultimately achieved. By considering the constraints of vehicle dynamics, road traffic rules, and dynamic obstacle avoidance constraints of surrounding traffic participants, the optimality of path planning is guaranteed.

[0076] Based on any of the above embodiments, such as Figure 4 As shown, the path planning vector neural network performs online iterative updates based on its internal neural network, which has the same structure, including:

[0077] Step 401: Optimize the parameters of the internal neural network with the same structure as the path planning vector neural network based on the training data to obtain the updated parameters of the network. The training data includes model simulation data, historical driving data and real-time driving data.

[0078] In this embodiment of the invention, the parameters of the internal neural network are updated not only based on real-time driving data, but also based on model simulation data and historical driving data, making the path planning vector neural network more complete.

[0079] Step 402: According to the preset synchronization period, obtain the parameter change rate of the internal neural network based on the parameters after the network update, and adjust the parameter change rate of the internal neural network according to the preset parameter change compression ratio to obtain the parameter change rate of the path planning vector neural network. The parameter change rate of the path planning vector neural network is less than the parameter change rate of the internal neural network.

[0080] Step 403: Adjust the corresponding parameters of the path planning vector neural network according to the rate of change of the path planning vector neural network to perform online iterative updates. In this embodiment of the invention, to ensure the iterative stability of the path planning vector neural network, a preset synchronization period is set to, for example, 10 hours or 1 day, which can be set by those skilled in the art according to actual needs. A small rate of change of the path planning neural network parameters can improve the stability of the path planning vector neural network.

[0081] This invention provides a path planning method for autonomous vehicles. By synchronizing the optimization parameters of the internal neural network to the path planning vector neural network, the method ensures the continuous updating and gradual adaptation of the autonomous vehicle to the driving environment during actual driving.

[0082] The path planning device for autonomous vehicles provided by the present invention will be described below. The path planning device for autonomous vehicles described below can be referred to in correspondence with the path planning method for autonomous vehicles described above.

[0083] Figure 5This is a schematic diagram of a path planning device for an autonomous vehicle provided in an embodiment of the present invention, as shown below. Figure 5 As shown, the path planning device for an autonomous vehicle provided in this embodiment of the invention includes:

[0084] The acquisition module 501 is used to convert multi-frame driving data into vectors and input them into the path planning vector neural network to obtain the reference driving path of the autonomous vehicle in the future period of time.

[0085] The planning module 502 is used to optimize the reference driving path according to the preset optimization target, obtain the optimal driving path when the optimization target is maximized, and use the optimal driving path as the path planning result of the autonomous vehicle.

[0086] The update module 503 is used for online iterative updates of the path planning vector neural network, which performs online iterative updates based on its internal neural network with the same structure.

[0087] This invention provides a path planning device for autonomous vehicles. By vectorizing and inputting multi-frame driving data into a path planning vector neural network, a reference driving path for the autonomous vehicle over a future period is obtained. The path planning vector neural network is iteratively updated online based on an internal neural network with the same structure. The reference driving path is optimized according to a preset optimization objective to obtain the optimal driving path when the optimization objective is maximized. This optimal driving path is used as the path planning result for the autonomous vehicle, enhancing its adaptability in unknown traffic driving scenarios and improving the computational speed and iteration speed of the autonomous driving path planning algorithm.

[0088] Figure 6 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 6 As shown, the electronic device may include a processor 610, a communication interface 620, a memory 630, and a communication bus 640. The processor 610, communication interface 620, and memory 630 communicate with each other via the communication bus 640. The processor 610 can call logical instructions in the memory 630 to execute a path planning method for an autonomous vehicle. This method includes: vectorizing multi-frame driving data and inputting it into a path planning vector neural network to obtain a reference driving path for the autonomous vehicle within a future period; wherein the path planning vector neural network is iteratively updated online based on an internal neural network with the same structure; optimizing the reference driving path according to a preset optimization objective to obtain the optimal driving path when the optimization objective is maximized; and using the optimal driving path as the path planning result for the autonomous vehicle.

[0089] Furthermore, the logical instructions in the aforementioned memory 630 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0090] On the other hand, embodiments of the present invention also provide an autonomous driving vehicle, including the above-mentioned electronic devices.

[0091] In this embodiment of the invention, the autonomous vehicle is a commercial vehicle. Since the path planning algorithm inside the on-board controller of existing commercial vehicles cannot be updated online after use, the path planning cannot be optimized quickly and continuously. Therefore, the path planning method of the autonomous vehicle of the present invention can realize the continuous iterative optimization and update of the path planning algorithm used by commercial vehicles, so as to continuously optimize the path planning and meet the needs of use under multiple working conditions on actual roads.

[0092] In another aspect, embodiments of the present invention also provide a non-transitory computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program implements a path planning method for an autonomous vehicle provided by the methods described above. The method includes: vectorizing multi-frame driving data and inputting it into a path planning vector neural network to obtain a reference driving path for the autonomous vehicle within a future period, wherein the path planning vector neural network is iteratively updated online based on an internal neural network with the same structure; optimizing the reference driving path according to a preset optimization objective to obtain the optimal driving path when the optimization objective is maximized, and using the optimal driving path as the path planning result of the autonomous vehicle.

[0093] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0094] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of various embodiments or some parts of embodiments.

[0095] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A path planning method for an autonomous vehicle, characterized in that, include: After vectorizing and converting multiple frames of driving data, the data is input into a path planning vector neural network to obtain a reference driving path for the autonomous vehicle within a certain period of time. The path planning vector neural network is updated online based on an internal neural network with the same structure. The driving data is obtained from onboard sensors and communication information. The reference driving path is optimized according to the preset optimization target to obtain the optimal driving path when the optimization target is maximized. The optimal driving path is used as the path planning result of the autonomous vehicle. The step of optimizing the reference driving path according to a preset optimization target to obtain the optimal driving path when the optimization target is maximized includes: The driving process of autonomous vehicles is discretized in time, and a Markov decision model is constructed based on the discretized vehicle parameters. The Markov decision model uses the vehicle parameters and surrounding vehicle parameters at each time step as the decision state, and the vehicle position at the next time step output by the path planning vector neural network as the decision action. The real-time driving parameters during the driving process are input into the Markov decision model, and the decision action is optimized according to the preset optimization objective to obtain the optimal driving path when the optimization objective is maximized. The path planning vector neural network is iteratively updated online based on its internal neural network, which has the same structure, including: Based on the training data, the parameters of the internal neural network with the same structure as the path planning vector neural network are optimized to obtain the updated parameters of the network. The training data includes model simulation data, historical driving data, and real-time driving data. According to a preset synchronization period, the parameter change rate of the internal neural network is obtained based on the parameters after the network update, and the parameter change rate of the internal neural network is adjusted according to a preset parameter change compression ratio to obtain the parameter change rate of the path planning vector neural network. The parameter change rate of the path planning vector neural network is less than the parameter change rate of the internal neural network. The parameters of the path planning vector neural network are adjusted according to the rate of change of the parameters of the path planning vector neural network in order to perform online iterative updates of the path planning vector neural network.

2. The path planning method for an autonomous vehicle according to claim 1, characterized in that, The process of vectorizing multi-frame driving data and inputting it into a path planning vector neural network to obtain a reference driving path for the autonomous vehicle over a future period includes: The driving data acquired at the current moment is input into the path planning vector neural network to obtain the coordinate position of the next moment in the traffic road coordinate system, with the current position as the origin; By inputting the driving data at each future moment into the path planning vector neural network, the coordinate position at each moment in the future within a certain period of time is obtained in the traffic road coordinate system; The reference driving path of the autonomous vehicle is obtained based on the coordinates of the vehicle at each moment in the future.

3. The path planning method for an autonomous vehicle according to claim 1, characterized in that, The vehicle parameters include at least one of vehicle position, speed, and acceleration; The surrounding vehicle parameters include at least one of the surrounding vehicle positions and speeds.

4. The path planning method for an autonomous vehicle according to claim 1, characterized in that, The optimization objectives include: The safe distance between the planned route and surrounding traffic participants, and the compliance of the planned route with road traffic rules.

5. The path planning method for an autonomous vehicle according to claim 1, characterized in that, The driving data includes: One or more of the following: autonomous driving maps, environmental perception data, behavior prediction data, and vehicle chassis dynamics data.

6. The path planning method for an autonomous vehicle according to claim 5, characterized in that, The vehicle chassis dynamics data includes vehicle speed, and the method for vectorizing the current vehicle speed includes: Based on the established vehicle dynamics coordinate system, the vehicle speed is decomposed into the velocity direction in the direction the vehicle is facing and the velocity direction in the direction the vehicle is sideways. The vector data of the vehicle speed is obtained based on the velocity direction and the specific value of the velocity corresponding to each velocity direction.

7. A path planning device for an autonomous vehicle, characterized in that, include: The acquisition module is used to convert multi-frame driving data into vectors and input them into the path planning vector neural network to obtain the reference driving path of the autonomous vehicle in the future. The driving data is obtained from onboard sensors and communication information; The planning module is used to optimize the reference driving path according to the preset optimization target, obtain the optimal driving path when the optimization target is maximized, and use the optimal driving path as the path planning result of the autonomous vehicle. An update module is used for online iterative updates of the path planning vector neural network, wherein the path planning vector neural network performs online iterative updates based on an internal neural network with the same structure. The planning module is specifically used for: The driving process of autonomous vehicles is discretized in time, and a Markov decision model is constructed based on the discretized vehicle parameters. The Markov decision model uses the vehicle parameters and surrounding vehicle parameters at each time step as the decision state, and the vehicle position at the next time step output by the path planning vector neural network as the decision action. The real-time driving parameters during the driving process are input into the Markov decision model, and the decision action is optimized according to the preset optimization objective to obtain the optimal driving path when the optimization objective is maximized. The update module is specifically used for: Based on the training data, the parameters of the internal neural network with the same structure as the path planning vector neural network are optimized to obtain the updated parameters of the network. The training data includes model simulation data, historical driving data, and real-time driving data. According to a preset synchronization period, the parameter change rate of the internal neural network is obtained based on the parameters after the network update, and the parameter change rate of the internal neural network is adjusted according to a preset parameter change compression ratio to obtain the parameter change rate of the path planning vector neural network. The parameter change rate of the path planning vector neural network is less than the parameter change rate of the internal neural network. The parameters of the path planning vector neural network are adjusted according to the rate of change of the parameters of the path planning vector neural network in order to perform online iterative updates of the path planning vector neural network.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the path planning method for the autonomous vehicle as described in any one of claims 1 to 6.

9. An autonomous vehicle, characterized in that, Including the electronic device as described in claim 8.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the path planning method for the autonomous vehicle as described in any one of claims 1 to 6.

Images