Vehicle path planning method, system, electronic device and storage medium

Abstract

The present invention provides a vehicle path planning method, system, electronic device and storage medium. First, all possible driving trajectories of surrounding vehicles are predicted based on the status information and traffic environment data of the unmanned vehicle and surrounding vehicles; then, the paths of surrounding vehicles are planned based on the path planning of the unmanned vehicle before the current moment, and the assumed planned trajectories of surrounding vehicles are generated; the assumed planned trajectories are corrected, and the corrected assumed planned trajectories are merged with the predicted trajectories, and the path planning of the unmanned vehicle is performed based on the trajectory merging results, so as to generate a driving path in which the unmanned vehicle and surrounding vehicles will not conflict or collide. The present invention not only takes into account the interaction between the unmanned vehicle and the surrounding vehicles, but also improves the rationality and accuracy of the prediction results of the trajectories of surrounding vehicles, and improves the initiative and anthropomorphism of the path planning of the unmanned vehicle.

Description

Technical Field

[0001] The present application relates to the field of autonomous driving technology, and specifically to a vehicle path planning method, system, electronic device and storage medium. Background Art

[0002] When an unmanned vehicle (or unmanned vehicle) is driving on a motor vehicle lane, it needs to predict multiple possible driving trajectories of surrounding vehicles in the future based on real-time detection of other vehicles around it (referred to as surrounding vehicles), and plan at least one safe path that will not conflict or collide with surrounding vehicles based on the trajectory prediction results of surrounding vehicles, and then use the planned safe path as the basis for the movement of the unmanned vehicle. Therefore, there is interactive coupling between the path planning of the unmanned vehicle and the trajectory prediction of surrounding vehicles, and they affect each other.

[0003] However, in the current methods of trajectory prediction for surrounding vehicles, the fact that the path planning of unmanned vehicles is a known condition is not taken into account, and the uniqueness of unmanned vehicles cannot be reflected. As a result, these methods will affect the trajectory prediction accuracy of surrounding vehicles, especially the trajectory prediction accuracy under the influence of the movement of unmanned vehicles. On the other hand, when the trajectory prediction is used as the input of the path planning of unmanned vehicles, it is easy to make the path planning of unmanned vehicles passive and conservative, and cannot reflect the interactive coupling between trajectory prediction and path planning. In addition, although some literature currently considers the two-way influence between unmanned vehicles and surrounding vehicles, it is not easy to trace the causal relationship and add constraints to correct and optimize the trajectory prediction. Summary of the invention

[0004] In view of the above-mentioned shortcomings of the prior art, the present invention provides a vehicle path planning method, system, electronic device and storage medium to solve the above-mentioned technical problems.

[0005] The present invention provides a vehicle path planning method, the method comprising the following steps:

[0006] According to the state information of the target unmanned vehicle, the state information of the surrounding vehicles, and the traffic environment data of the target unmanned vehicle and the surrounding vehicles, all possible driving trajectories of the surrounding vehicles are predicted and recorded as the predicted trajectories of the surrounding vehicles;

[0007] Performing path planning for the surrounding vehicles according to the path planning of the target unmanned vehicle before the current moment, and generating an assumed planned trajectory of the surrounding vehicles;

[0008] Correcting the assumed planned trajectory, merging the corrected assumed planned trajectory with the predicted trajectory, and performing path planning for the target unmanned vehicle according to the trajectory merging result to generate at least one driving path in which the target unmanned vehicle will not conflict or collide with the surrounding vehicles;

[0009] The target unmanned vehicle includes an unmanned vehicle that is predetermined or determined in real time, and the surrounding vehicles include vehicles that are within a preset area with the target unmanned vehicle.

[0010] In one embodiment of the present invention, based on the state information of the target unmanned vehicle, the state information of the surrounding vehicles, and the traffic environment data of the target unmanned vehicle and the surrounding vehicles, the process of predicting all possible driving trajectories of the surrounding vehicles includes:

[0011] Rendering the historical status information of the target unmanned vehicle and the historical status information of the surrounding vehicles into a grid map that meets a preset accuracy; wherein the grid map that meets the preset accuracy is obtained based on the traffic environment data;

[0012] In a raster map that meets a preset accuracy, different color values ​​are used to represent different semantic information, and gradient colors of the same color are used to represent information at different historical time steps, thereby obtaining a raster map containing dynamic information;

[0013] Use convolutional neural networks to encode raster maps containing dynamic information and obtain spatial features containing vehicle interaction information and traffic environment data;

[0014] Using a gated recurrent unit to encode traffic rules, historical state information of the target driverless vehicle, and historical state information of surrounding vehicles to obtain time features;

[0015] Fusing the spatial feature and the temporal feature to obtain a spatiotemporal fusion feature;

[0016] The spatiotemporal fusion features are input into a multi-layer perceptron for decoding, and all possible driving trajectories of the surrounding vehicles in multiple future time steps are predicted.

[0017] In one embodiment of the present invention, the process of performing path planning for the surrounding vehicles according to the path planning of the target unmanned vehicle before the current moment and generating the assumed planned trajectory of the surrounding vehicles includes:

[0018] Predicting the path initial state sequence of the surrounding vehicles according to the path planning of the target unmanned vehicle before the current moment;

[0019] The initial state sequence of the paths of the surrounding vehicles is processed by using a fast search random tree model, and the paths of the surrounding vehicles are planned from the perspective of the surrounding vehicles to generate an assumed planned trajectory of the surrounding vehicles.

[0020] In one embodiment of the present invention, the process of predicting the path initial state sequence of the surrounding vehicles according to the path planning of the target unmanned vehicle before the current moment includes:

[0021] Establishing a partially observable Markov decision model based on the perspective of the surrounding vehicles;

[0022] The partially observable Markov decision model is used to predict the path initial state sequence of the surrounding vehicles from the path planning of the target unmanned vehicle before the current moment, the traffic environment data and the current state information of the surrounding vehicles.

[0023] In one embodiment of the present invention, the process of correcting the assumed planned trajectory includes:

[0024] Determining the interaction behavior between the target unmanned vehicle and the surrounding vehicles according to the state information of the target unmanned vehicle and the state information of the surrounding vehicles;

[0025] Modifying the assumed planned trajectory based on the interaction behavior between the target unmanned vehicle and the surrounding vehicles;

[0026] The interactive behavior includes one of the following: overtaking behavior, following behavior, avoiding behavior, and meeting behavior.

[0027] In one embodiment of the present invention, the process of correcting the assumed planned trajectory based on the interaction behavior between the target unmanned vehicle and the surrounding vehicles includes:

[0028] When the target unmanned vehicle is in overtaking behavior, shortening and correcting the assumed planned trajectory;

[0029] When the target unmanned vehicle is in a following behavior or an avoidance behavior, performing an increase correction on the assumed planned trajectory;

[0030] When the target unmanned vehicle is in a meeting behavior, the assumed planned trajectory is not corrected.

[0031] In one embodiment of the present invention, when the assumed planned trajectory is corrected, the method further includes:

[0032] Determining the right-of-way priority between the target unmanned vehicle and the surrounding vehicles according to preset traffic prior knowledge, the location information of the target unmanned vehicle, and the location information of the surrounding vehicles;

[0033] Obtain surrounding vehicles with a lower right of way priority than the target unmanned vehicle, and record them as surrounding vehicles to be corrected;

[0034] Performing a shortening correction on the assumed planned trajectory of the surrounding vehicles to be corrected;

[0035] The preset traffic prior knowledge includes traffic rules, traffic scenarios and maps.

[0036] In one embodiment of the present invention, the process of merging the corrected assumed planned trajectory with the predicted trajectory and performing path planning for the target unmanned vehicle according to the trajectory merging result includes:

[0037] Adding the corrected assumed planned trajectory to the predicted trajectory for merging to obtain a new predicted trajectory;

[0038] Using the new predicted trajectory as a path planning condition for the target unmanned vehicle in the current cycle, and using the partially observable Markov decision model to obtain a path initial state sequence of the target unmanned vehicle from the new predicted trajectory, the current state information of the target unmanned vehicle, and the traffic environment data;

[0039] The fast search random tree model is used to process the path initial state sequence of the target unmanned vehicle, and the path of the target unmanned vehicle is planned to generate at least one driving path that prevents the target unmanned vehicle from conflicting or colliding with the surrounding vehicles.

[0040] In one embodiment of the present invention, after adding the corrected assumed planned trajectory to the predicted trajectory for merging to obtain a new predicted trajectory, the method further includes:

[0041] Obtaining the confidence of the predicted trajectory generated when predicting all possible driving trajectories of the surrounding vehicles; wherein the confidence is used to indicate the credibility or credibility probability of the trajectory;

[0042] Assigning a dynamic initial value to the corrected assumed planning trajectory as the confidence of the corrected assumed planning trajectory;

[0043] The confidence of the predicted trajectory before merging and the dynamic initial value are normalized to correct the confidence of the new predicted trajectory.

[0044] The present invention also provides a vehicle path planning system, the system comprising:

[0045] A surrounding vehicle trajectory prediction module, used to predict all possible driving trajectories of the surrounding vehicles according to the state information of the target unmanned vehicle, the state information of the surrounding vehicles, and the traffic environment data of the target unmanned vehicle and the surrounding vehicles, and record them as the predicted trajectories of the surrounding vehicles;

[0046] A surrounding vehicle trajectory planning module, used to perform path planning for the surrounding vehicles according to the path planning of the target unmanned vehicle before the current moment, and generate an assumed planned trajectory of the surrounding vehicles;

[0047] A trajectory correction module, used to correct the assumed planned trajectory and merge the corrected assumed planned trajectory with the predicted trajectory;

[0048] An unmanned vehicle path planning module, used for performing path planning for the target unmanned vehicle according to the trajectory merging result, so as to generate at least one driving path in which the target unmanned vehicle will not conflict or collide with the surrounding vehicles;

[0049] The target unmanned vehicle includes an unmanned vehicle that is predetermined or determined in real time, and the surrounding vehicles include vehicles that are within a preset area with the target unmanned vehicle.

[0050] The present invention also provides an unmanned vehicle, wherein the unmanned vehicle travels according to a preset driving path, and the generation process of the preset driving path includes:

[0051] According to the state information and traffic environment data of the unmanned vehicle, as well as the state information and traffic environment data of surrounding vehicles, all possible driving trajectories of the surrounding vehicles are predicted, and recorded as predicted trajectories of the surrounding vehicles;

[0052] Performing path planning for the surrounding vehicles according to the path planning of the unmanned vehicle before the current moment, and generating an assumed planned trajectory of the surrounding vehicles;

[0053] Correcting the assumed planned trajectory, merging the corrected assumed planned trajectory with the predicted trajectory, and performing path planning for the unmanned vehicle according to the trajectory merging result to generate at least one driving path in which the target unmanned vehicle will not conflict or collide with the surrounding vehicles;

[0054] The surrounding vehicles include vehicles that are within a preset area with the unmanned vehicle.

[0055] The present invention also provides an electronic device, comprising:

[0056] one or more processors;

[0057] A storage device is used to store one or more programs, and when the one or more programs are executed by the one or more processors, the electronic device implements the vehicle path planning method as described in any one of the above.

[0058] The present invention also provides a computer-readable storage medium, characterized in that computer-readable instructions are stored thereon, and when the computer-readable instructions are executed by a processor of a computer, the computer executes the vehicle path planning method as described in any one of the above.

[0059] As described above, the present invention provides a vehicle path planning method, system, electronic device and storage medium, which have the following beneficial effects:

[0060] The present invention first predicts all possible driving trajectories of the surrounding vehicles according to the state information of the target unmanned vehicle, the state information of the surrounding vehicles, and the traffic environment data of the target unmanned vehicle and the surrounding vehicles, and records them as the predicted trajectories of the surrounding vehicles; then, according to the path planning of the target unmanned vehicle before the current moment, the surrounding vehicles are path planned to generate the assumed planning trajectory of the surrounding vehicles; finally, the assumed planning trajectory is corrected, and the corrected assumed planning trajectory is merged with the predicted trajectory, and the target unmanned vehicle is path planned according to the trajectory merging result to generate at least one driving path that does not conflict or collide with the target unmanned vehicle and the surrounding vehicles. Wherein, the target unmanned vehicle includes an unmanned vehicle that is predetermined or determined in real time, and the surrounding vehicles include vehicles that are in a preset area with the target unmanned vehicle. It can be seen from this that the present invention can not only predict all possible driving trajectories of the surrounding vehicles, but also can plan the paths of the surrounding vehicles from the perspective of the surrounding vehicles according to the planned path of the unmanned vehicle before the current cycle, and obtain the assumed planning trajectory of the surrounding vehicles, which can be used to supplement and improve the predicted trajectory of the surrounding vehicles, reduce the uncertainty of the trajectory prediction of the surrounding vehicles, and improve the rationality and accuracy of the trajectory prediction results. At the same time, the present invention can also add the corrected assumed trajectory to the multiple predicted trajectories of the surrounding vehicles, and use them together as the input of the target unmanned vehicle path planning, and perform the path planning of the target unmanned vehicle in the current cycle, and can generate at least one driving path that will prevent the target unmanned vehicle from conflicting or colliding with the surrounding vehicles. This is equivalent to not only considering the interaction between the target unmanned vehicle and the surrounding vehicles, but also improving the rationality and accuracy of the prediction results of the surrounding vehicle trajectories, and improving the initiative and anthropomorphism of the target unmanned vehicle path planning.

[0061] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] The drawings herein are incorporated into the specification and constitute a part of the specification, showing embodiments consistent with the present application, and together with the specification, are used to explain the principles of the present application. Obviously, the drawings described below are only some embodiments of the present application, and for those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative work. In the drawings:

[0063] Figure 1 A schematic diagram of an exemplary system architecture for applying the technical solutions in one or more embodiments of the present application;

[0064] Figure 2 A schematic diagram of a vehicle path planning method according to an embodiment of the present application;

[0065] Figure 3 A schematic diagram of a process for predicting the driving trajectory of surrounding vehicles provided in an embodiment of the present application;

[0066] Figure 4 A schematic diagram of a predicted trajectory before trajectory merging provided in an embodiment of the present application;

[0067] Figure 5 A schematic diagram of a predicted trajectory after trajectory merging is completed in an embodiment of the present application;

[0068] Figure 6 A schematic diagram of a vehicle path planning method provided in another embodiment of the present application;

[0069] Figure 7 A schematic diagram of the hardware structure of a vehicle path planning system provided in an embodiment of the present application;

[0070] Figure 8 The hardware structure diagram of an electronic device suitable for implementing one or more embodiments of the present application. DETAILED DESCRIPTION

[0071] The following will describe the embodiments of the present invention with reference to the accompanying drawings and preferred embodiments. Those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be understood that the preferred embodiments are only for illustrating the present invention, not for limiting the scope of protection of the present invention.

[0072] It should be noted that the illustrations provided in the following embodiments are only schematic illustrations of the basic concept of the present invention, and thus the drawings only show components related to the present invention rather than being drawn according to the number, shape and size of components in actual implementation. In actual implementation, the type, quantity and proportion of each component may be changed arbitrarily, and the component layout may also be more complicated.

[0073] In the following description, numerous details are discussed to provide a more thorough explanation of the embodiments of the present invention. However, it is obvious to those skilled in the art that the embodiments of the present invention can be implemented without these specific details. In other embodiments, well-known structures and devices are shown in the form of block diagrams rather than in detail to avoid making the embodiments of the present invention difficult to understand.

[0074] Figure 1 A schematic diagram of an exemplary system architecture to which the technical solution in one or more embodiments of the present application can be applied is shown. Figure 1 As shown, the system architecture 100 may include a terminal device 110, a network 120, and a server 130. The terminal device 110 may include various electronic devices such as a smart phone, a tablet computer, a laptop computer, and a desktop computer. The server 130 may be an independent physical server, or a server cluster or a distributed system composed of multiple physical servers, or a cloud server providing cloud computing services. The network 120 may be a communication medium of various connection types that can provide a communication link between the terminal device 110 and the server 130, such as a wired communication link or a wireless communication link.

[0075] According to the implementation requirements, the system architecture in the embodiment of the present application can have any number of terminal devices, networks and servers. For example, the server 130 can be a server group composed of multiple server devices. In addition, the technical solution provided in the embodiment of the present application can be applied to the terminal device 110, can also be applied to the server 130, or can be implemented by the terminal device 110 and the server 130 together, and the present application does not make any special restrictions on this.

[0076] In one embodiment of the present application, the terminal device 110 or the server 130 of the present application can predict all possible driving trajectories of the surrounding vehicles based on the state information of the target unmanned vehicle, the state information of the surrounding vehicles, and the traffic environment data of the target unmanned vehicle and the surrounding vehicles, and record them as the predicted trajectories of the surrounding vehicles; then, according to the path planning of the target unmanned vehicle before the current moment, the surrounding vehicles are path planned to generate the assumed planned trajectory of the surrounding vehicles; finally, the assumed planned trajectory is corrected, and the corrected assumed planned trajectory is merged with the predicted trajectory, and the target unmanned vehicle is path planned according to the trajectory merging result to generate at least one driving path that prevents the target unmanned vehicle from conflicting or colliding with the surrounding vehicles. Among them, the target unmanned vehicle includes an unmanned vehicle determined in advance or in real time, and the surrounding vehicles include vehicles in a preset area with the target unmanned vehicle. It can be seen that after executing the vehicle path planning method using the terminal device 110 or the server 130, not only can all possible driving trajectories of the surrounding vehicles be predicted, but also the surrounding vehicles can be path planned from the perspective of the surrounding vehicles according to the planned path of the unmanned vehicle before the current cycle, and the assumed planned trajectory of the surrounding vehicles can be obtained, which can be used to supplement and improve the predicted trajectory of the surrounding vehicles, thereby reducing the uncertainty of the trajectory prediction of the surrounding vehicles and improving the rationality and accuracy of the trajectory prediction results. At the same time, the terminal device 110 or the server 130 can also be used to supplement the corrected assumed trajectory to the multiple predicted trajectories of the surrounding vehicles, and together as the input of the target unmanned vehicle path planning, the path planning of the target unmanned vehicle in the current cycle can be performed, and at least one driving path can be generated so that the target unmanned vehicle will not conflict or collide with the surrounding vehicles. It is equivalent to using the terminal device 110 or the server 130 to execute the vehicle path planning method, which not only takes into account the interactive influence of the target unmanned vehicle and the surrounding vehicles, but also improves the rationality and accuracy of the trajectory prediction results of the surrounding vehicles, and improves the initiative and anthropomorphism of the path planning of the target unmanned vehicle.

[0077] The above section introduces the contents of an exemplary system architecture that applies the technical solution of the present application. Next, we will continue to introduce the method of generating a digital twin instance of the present application.

[0078] Figure 2 FIG. 1 shows a flow chart of a vehicle path planning method provided by an embodiment of the present application. Specifically, in an exemplary embodiment, Figure 2 As shown, this embodiment provides a vehicle path planning method, which includes the following steps:

[0079] S210, based on the state information of the target unmanned vehicle, the state information of the surrounding vehicles, and the traffic environment data of the target unmanned vehicle and the surrounding vehicles, predict all possible driving trajectories of the surrounding vehicles, and record them as the predicted trajectories of the surrounding vehicles. As an example, the target unmanned vehicle includes an unmanned vehicle that is predetermined or determined in real time, and the surrounding vehicles include one or more vehicles that are in a preset area with the target unmanned vehicle, and the state information includes current state information and historical state information. Among them, the preset area can be a preset area, a preset distance area, or other preset areas; this embodiment does not impose any limitation on the form of the preset area, so it will not be repeated here. In the embodiment of the present method, the unmanned vehicle can also be referred to as an unmanned vehicle.

[0080] S220, performing path planning for the surrounding vehicles according to the path planning of the target unmanned vehicle before the current moment, and generating an assumed planned trajectory of the surrounding vehicles. As an example, the path planning of the target unmanned vehicle before the current moment includes but is not limited to: the path planning of the target unmanned vehicle in the previous cycle.

[0081] S230, correcting the assumed planned trajectory, and merging the corrected assumed planned trajectory with the predicted trajectory, and performing path planning for the target unmanned vehicle according to the trajectory merging result to generate at least one driving path that prevents the target unmanned vehicle from conflicting or colliding with the surrounding vehicles.

[0082] It can be seen that this embodiment first predicts all possible driving trajectories of surrounding vehicles based on the state information of the target unmanned vehicle, the state information of surrounding vehicles, and the traffic environment data of the target unmanned vehicle and surrounding vehicles, and then performs path planning for surrounding vehicles from the perspective of surrounding vehicles based on the planned path of the unmanned vehicle before the current cycle, and obtains the assumed planned trajectory of the surrounding vehicles, which is used to supplement and improve the predicted trajectory of the surrounding vehicles, and can reduce the uncertainty of the trajectory prediction of the surrounding vehicles, and improve the rationality and accuracy of the trajectory prediction results. At the same time, this embodiment can also supplement the corrected assumed trajectory to the multiple predicted trajectories of the surrounding vehicles, and use them together as the input of the path planning of the target unmanned vehicle, and perform the path planning of the target unmanned vehicle in the current cycle, and can generate at least one driving path that does not cause conflict or collision between the target unmanned vehicle and the surrounding vehicles. This is equivalent to not only considering the interactive influence of the target unmanned vehicle and the surrounding vehicles, but also improving the rationality and accuracy of the trajectory prediction results of the surrounding vehicles, and improving the initiative and anthropomorphism of the path planning of the target unmanned vehicle.

[0083] like Figure 3As shown, in an exemplary embodiment, based on the state information of the target unmanned vehicle, the state information of the surrounding vehicles, and the traffic environment data of the target unmanned vehicle and the surrounding vehicles, the process of predicting all possible driving trajectories of the surrounding vehicles may be:

[0084] S310, rendering the historical status information of the target unmanned vehicle and the historical status information of the surrounding vehicles into a grid map that meets a preset accuracy; wherein the grid map that meets the preset accuracy is obtained based on the traffic environment data. As an example, for example, the traffic environment data in this embodiment may include a grid map that meets the preset accuracy. wherein the accuracy of the grid map may be preset according to actual conditions, and this embodiment does not impose a specific numerical limit.

[0085] S320, using different color values ​​to represent different semantic information in a raster map that meets a preset accuracy, and using gradient colors of the same color to represent information at different historical time steps, to obtain a raster map containing dynamic information;

[0086] S330, using a convolutional neural network to encode a grid map containing dynamic information to obtain spatial features containing vehicle interaction information and traffic environment data;

[0087] S340, using a gated recurrent unit to encode traffic rules, historical status information of the target unmanned vehicle, and historical status information of surrounding vehicles to obtain a time feature;

[0088] S350, fusing the spatial feature and the temporal feature to obtain a spatiotemporal fusion feature; as an example, this embodiment may use a feature concatenation function to fuse the spatial feature and the temporal feature.

[0089] S360, input the spatiotemporal fusion features into two multi-layer perceptrons for decoding, and predict all possible driving trajectories of the surrounding vehicles in multiple future time steps, as well as the confidence of each driving trajectory. The confidence of each driving trajectory is used to indicate the credibility or credibility probability of the driving trajectory. The greater the confidence, the higher the credibility of the corresponding driving trajectory, or the greater the confidence, the greater the credibility probability of the corresponding driving trajectory.

[0090] Specifically, the historical status information of the unmanned vehicle and surrounding vehicles is rendered into a three-channel raster map that meets the preset accuracy (hereinafter referred to as the raster high-precision map), and different color values ​​are used to represent different semantic information, while the same color gradient is used to represent the information of different historical time steps, so as to obtain a raster high-precision map M containing dynamic information. r .

[0091] Using lightweight convolutional neural network (CNN) EfficientNet to encode the grid high-precision map M r , and obtain the spatial feature F containing vehicle interaction information and traffic environment data S At the same time, the gated recurrent unit GRU (Gated Recurrent Unit, referred to as GRU) is used to encode the historical state data of the driverless vehicle, the historical state information of the surrounding vehicles and traffic rules to obtain the time feature F T ;

[0092] F S =EfficientNet(M r ; W e );

[0093]

[0094] Where EfficientNet(·) is a lightweight convolutional neural network, M r is a high-precision raster map, W e is the weight of the lightweight convolutional neural network, GRU(·) is the gated recurrent unit, The historical status information of the driverless vehicle. is the historical status information of surrounding vehicles, R t is the traffic rules information, W GRU is the gated recurrent unit weight.

[0095] Use the feature concatenation function Concatenate to concatenate the spatial features F S and time feature F T Fusion is performed to obtain the spatiotemporal fusion feature F ST :

[0096] F ST =Concatenate(F S ,F T );

[0097] In the formula, Concatenate(·) is the feature concatenation function, F S is the spatial feature, F T For time characteristics.

[0098] The spatiotemporal fusion feature F ST After two multi-layer perceptron (MLP) decoding, the future T f The multimodal initial prediction trajectory consisting of K initial prediction trajectories at time steps, and the confidence conf of each initial prediction trajectoryk ;have:

[0099]

[0100] In the formula, MLP T (·) is a multi-layer perceptron that generates the predicted trajectory, MLP c (·) is the multi-layer perceptron that generates confidence, F ST is the spatiotemporal fusion feature. For multi-layer perceptron MLP T The weight of (·), For multi-layer perceptron MLP c (·) is the weight, Softmax(·) is the softmax function, Traj k is the kth predicted trajectory, conf k is the confidence of the kth trajectory. The confidence refers to the credibility or probability of the predicted trajectory, k = 1, 2, ..., K. The subscript f refers to the future time, and the time step refers to the time required to pass two consecutive position points on a trajectory. Among them, the multimodal initial prediction trajectory composed of K initial prediction trajectories is as follows Figure 4 shown.

[0101] In this embodiment, the method of obtaining the historical status information of the unmanned vehicle and the historical status information of the surrounding vehicles can be: obtaining the historical status information of the unmanned vehicle through the vehicle-mounted sensing system and positioning system of the unmanned vehicle and historical status information of surrounding vehicles The superscript ego represents the driverless vehicle, the superscript tar represents the surrounding vehicles perceived by the driverless vehicle, and the subscript t = -T p ,-T p +1,…,0 represents the time step, such as t=0 is the current time, -T p is the Tth historical moment in the past, the subscript p refers to the historical moment past, p t represents the position information at time t, v t represents the speed information at time t, y t Indicates the heading angle information at time t, l t Represents the length and width of the vehicle.

[0102] In an exemplary embodiment, the process of performing path planning for the surrounding vehicles based on the path planning of the target unmanned vehicle before the current moment and generating the assumed planned trajectory of the surrounding vehicles includes: predicting the path initial state sequence of the surrounding vehicles based on the path planning of the target unmanned vehicle before the current moment; processing the path initial state sequence of the surrounding vehicles using a fast search random tree model, performing path planning for the surrounding vehicles from the perspective of the surrounding vehicles, and generating the assumed planned trajectory of the surrounding vehicles. As an example, the process of predicting the path initial state sequence of the surrounding vehicles based on the path planning of the target unmanned vehicle before the current moment may be: establishing a partially observable Markov decision model from the perspective of the surrounding vehicles; using the partially observable Markov decision model to predict the path initial state sequence of the surrounding vehicles from the path planning of the target unmanned vehicle before the current moment, the traffic environment data, and the current state information of the surrounding vehicles.

[0103] Specifically, the surrounding traffic environment data info of each surrounding vehicle in the vehicle coordinate system is obtained through the perception system of the unmanned vehicle, the positioning system of the unmanned vehicle, and the coordinate conversion. env , and use the cache mechanism to obtain the planned path of the unmanned vehicle before the current moment or in the previous cycle. Figure 4 As shown, the gray dotted line represents the planned path of the driverless vehicle in the previous cycle In this embodiment, the cache mechanism refers to a medium or method for storing historical information data of the unmanned vehicle.

[0104] A partially observable Markov decision process (POMDP) ​​is established from the perspective of each surrounding vehicle, and then the initial state sequence of each surrounding vehicle path is obtained using the POMDP model. have:

[0105]

[0106] Where POMDP(·) is a partially observable Markov decision model, info env For traffic environment data, The planned path of the driverless vehicle in the previous cycle, It is the current status information of surrounding vehicles.

[0107] According to the initial state sequence of each surrounding vehicle The rapid-exploring random trees (RRT) model is used to plan the future T of surrounding vehicles.f The path trajectory of each time step is obtained to obtain the assumed planning trajectory of each surrounding vehicle

[0108]

[0109] Where RRT(·) is the fast search random tree model, is the initial state sequence of the paths of the surrounding vehicles.

[0110] In an exemplary embodiment, the process of correcting the assumed planned trajectory includes: determining the interaction behavior between the target unmanned vehicle and the surrounding vehicles according to the state information of the target unmanned vehicle and the state information of the surrounding vehicles; and correcting the assumed planned trajectory based on the interaction behavior between the target unmanned vehicle and the surrounding vehicles. Wherein, the interaction behavior includes one of the following: overtaking behavior, following behavior, avoiding behavior, and meeting behavior. Specifically, the process of correcting the assumed planned trajectory based on the interaction behavior between the target unmanned vehicle and the surrounding vehicles may be: when the target unmanned vehicle is in overtaking behavior, shortening correction is made to the assumed planned trajectory. When the target unmanned vehicle is in following behavior or avoiding behavior, lengthening correction is made to the assumed planned trajectory. When the target unmanned vehicle is in meeting behavior, the assumed planned trajectory is not corrected. When correcting the assumed planned trajectory, this embodiment can also: determine the right of way priority between the target unmanned vehicle and the surrounding vehicles based on preset traffic prior knowledge, the location information of the target unmanned vehicle and the location information of the surrounding vehicles; obtain the surrounding vehicles whose right of way priority is lower than that of the target unmanned vehicle, and record them as the surrounding vehicles to be corrected; shorten the assumed planned trajectory of the surrounding vehicles to be corrected; wherein the preset traffic prior knowledge includes traffic rules, traffic scenes and maps.

[0111] Specifically, the process of correcting the assumed planned trajectory and obtaining a new predicted trajectory is as follows:

[0112] According to the current state information and historical state information of the unmanned vehicle and the current state information and historical state information of the surrounding vehicles, the interactive behavior of the unmanned vehicle and the surrounding vehicles is determined. The interactive behavior includes but is not limited to overtaking, following, avoiding and meeting.

[0113] The assumed planned trajectory is corrected according to the interaction stage between the unmanned vehicle and the surrounding vehicles. When the unmanned vehicle is in overtaking behavior, the assumed planned trajectory is shortened; when the unmanned vehicle is in following or avoiding behavior, the assumed planned trajectory is lengthened; when the unmanned vehicle is in meeting behavior, the assumed planned trajectory is not corrected. In order to ensure that the number of trajectory points of the assumed planned trajectory remains unchanged after the correction process, this embodiment can re-sample 10 trajectory points evenly on the corrected trajectory.

[0114] According to the positioning system, traffic rules and high-precision maps, the road priority between the unmanned vehicle and surrounding vehicles is determined. For surrounding vehicles with lower road priority than the unmanned vehicle, their assumed planned trajectory is shortened. Figure 4 As shown, in the traffic scene at the crossroad, A is the unmanned vehicle, B is the currently predicted vehicle among the surrounding vehicles, and C is other vehicles. When vehicle A is going straight and vehicle B in the opposite lane turns left, according to traffic regulations, vehicle A has a higher right of way priority than vehicle B. At this time, vehicle A should continue to go straight and does not have to give way to vehicle B. Therefore, in order to allow the unmanned vehicle to plan a trajectory that continues to go straight as much as possible, this embodiment can shorten the assumed planned trajectory of vehicle B. Figure 5 As shown in the figure, the assumed planned trajectory of the surrounding vehicle B is shortened and corrected as follows Figure 5 The black dashed line indicates the trajectory.

[0115] In an exemplary embodiment, the modified assumed planned trajectory is merged with the predicted trajectory, and the process of performing path planning for the target unmanned vehicle according to the trajectory merging result includes: adding the modified assumed planned trajectory to the predicted trajectory for merging to obtain a new predicted trajectory; using the new predicted trajectory as the path planning condition of the target unmanned vehicle in the current cycle, and using the partially observable Markov decision model to obtain the path initial state sequence of the target unmanned vehicle from the new predicted trajectory, the current state information of the target unmanned vehicle, and the traffic environment data; using the fast search random tree model to process the path initial state sequence of the target unmanned vehicle, and perform path planning for the target unmanned vehicle to generate at least one driving path that prevents the target unmanned vehicle from conflicting or colliding with the surrounding vehicles. In this embodiment, after adding the revised assumed planned trajectory to the predicted trajectory for merging, and obtaining a new predicted trajectory, this embodiment may also include: obtaining the confidence of the predicted trajectory generated when predicting all possible driving trajectories of surrounding vehicles; assigning a dynamic initial value to the revised assumed planned trajectory as the confidence of the revised assumed planned trajectory; the confidence is used to indicate the credibility or credibility probability of the trajectory; the confidence of the predicted trajectory before merging and the dynamic initial value are normalized to correct the confidence of the new predicted trajectory. Among them, the confidence is used to indicate the credibility or credibility probability of the trajectory.

[0116] According to the above records, specifically, this embodiment can modify the assumed planned trajectory Merge it into the K prediction trajectories predicted initially to obtain a new prediction trajectory, that is, to obtain K+1 prediction trajectories Then a dynamic initial value λ is assigned to the assumed planning trajectory before merging as the confidence of the assumed planning trajectory before merging; and the confidence conf of the initial K predicted trajectories is calculated using the normalized function softmax. k And the initial value λ is processed to correct the confidence of the new predicted trajectory have:

[0117]

[0118] In the formula, Softmax(·) is the softmax function, conf k is the confidence of the initial K predicted trajectories, 0<λ<1 is the dynamic initial value, k=1, 2, ..., K. To prevent the unmanned vehicle from being too aggressive in its path behavior, the dynamic initial value of the confidence of all assumed planned trajectories that have undergone the growth process in this embodiment can be a larger value.

[0119] The process of performing path planning for the target unmanned vehicle according to the trajectory merging result to generate at least one driving path that prevents the target unmanned vehicle from conflicting or colliding with the surrounding vehicles may be: using the new K+1 predicted trajectories as the path and planning basis of the current cycle of the unmanned vehicle, and using the POMDP and RRT models to obtain the planned path of the unmanned vehicle

[0120]

[0121] In the formula, info env For traffic environment data, is the current status information of the driverless vehicle, is the predicted K+1 trajectories, The initial state sequence of the path of the unmanned vehicle, k = 1, 2, ..., K.

[0122] like Figure 6 As shown, in an exemplary embodiment, this embodiment also provides a vehicle path planning method, the method comprising the following steps:

[0123] A. Obtain the initial multimodal predicted trajectories of surrounding vehicles. Specifically, the process of obtaining the initial multimodal predicted trajectories of surrounding vehicles is as follows:

[0124] A1. Obtain the historical status information of the unmanned vehicle through the vehicle-mounted perception system and positioning system of the unmanned vehicle and historical status information of surrounding vehicles The superscript ego represents the driverless vehicle, the superscript tar represents the surrounding vehicles sensed by the driverless vehicle, and the subscript t = -T p ,-T p +1,…,0 represents the time step, such as t=0 is the current time, t=-T p is the Tth historical moment in the past, the subscript p refers to the historical moment past, p t represents the position information at time t, v t represents the speed information at time t, y t Indicates the heading angle information at time t, l t Represents the length and width of the vehicle.

[0125] A2. Obtain traffic environment data, including high-precision map data of the driving environment of the unmanned vehicle and surrounding vehicles. Render the historical status information of the unmanned vehicle and surrounding vehicles into a three-channel raster high-precision map, and use different color values ​​in the raster high-precision map to represent different semantic information, and use the same color gradient to represent historical information at different time steps, to obtain a raster high-precision map M containing dynamic information.r .

[0126] A3. Encoding raster high-precision map M using EfficientNet, a lightweight convolutional neural network with 19 convolutional layers r , and obtain the spatial feature F containing the interaction information and traffic environment data S At the same time, a gated recurrent unit GRU with 32 hidden units is used to encode the historical state information of the driverless vehicle, the historical state information of the surrounding vehicles and traffic rules to obtain the temporal feature F T :

[0127] F S =EfficientNet(M r ; W e );

[0128]

[0129] Where EfficientNet(·) is a lightweight convolutional neural network, M r is a high-precision raster map, W e is the weight of the lightweight convolutional neural network, GRU(·) is the gated recurrent unit, The historical status information of the driverless vehicle, is the historical status information of surrounding vehicles, R t is the traffic rules information, W GRU is the gated recurrent unit weight.

[0130] A4. Use the feature concatenation function Concatenate to concatenate the spatial features F S and time feature F T To merge, such as Figure 6 As shown, the spatiotemporal fusion feature F is obtained ST :

[0131] F ST =Concatenate(F S ,F T );

[0132] In the formula, Concatenate(·) is the feature concatenation function, F S is the spatial feature, F T For time characteristics.

[0133] A5. Fusion of spatiotemporal features F ST After two multi-layer perceptron (MLP) decoding, three initial multimodal prediction trajectories for the next 10 time steps of 5 seconds and their related confidence conf are obtained. k ,like Figure 4 As shown in the figure, the black solid lines represent the three predicted trajectories, and the numbers represent the confidence of each trajectory:

[0134]

[0135] In the formula, MLP T (·) is a multi-layer perceptron that generates the predicted trajectory, MLP c (·) is the multi-layer perceptron that generates confidence, F ST is the spatiotemporal fusion feature. For multi-layer perceptron MLP T The weight of (·), For multi-layer perceptron MLP c (·) is the weight, Softmax(·) is the softmax function, Traj k is the kth predicted trajectory, conf k is the confidence of the kth trajectory. Confidence refers to the credibility or probability of the predicted trajectory, k = 1, 2, 3.

[0136] B. Obtaining the assumed planned trajectories of surrounding vehicles. Specifically, the process of obtaining the assumed planned trajectories of surrounding vehicles is as follows:

[0137] B1. Obtain traffic environment data info of surrounding vehicles in the vehicle coordinate system through the perception system of the unmanned vehicle, the positioning system of the unmanned vehicle, and coordinate conversion env , and use the cache mechanism to obtain the planned path of the unmanned vehicle in the previous cycle. Figure 4 As shown, the gray dotted line represents the planned path of the unmanned vehicle in the previous cycle

[0138] B2. Establish a partially observable Markov decision process (POMDP) ​​from the perspective of each surrounding vehicle, and then use the POMDP model to obtain the initial state sequence of each surrounding vehicle path. have:

[0139]

[0140] Where POMDP(·) is a partially observable Markov decision model, info env For traffic environment data, The planned path of the driverless vehicle in the previous cycle, It is the current status information of each surrounding vehicle.

[0141] B3. Based on the initial state sequence of each surrounding vehicle The rapid-exploring random trees (RRT) model is used to plan the future T of surrounding vehicles. f The path trajectory of each time step is obtained to obtain the assumed planning trajectory of each surrounding vehicle

[0142]

[0143] Where RRT(·) is the fast search random tree model, is the initial state sequence of the paths of the surrounding vehicles.

[0144] C. Modify the assumed planned trajectory and obtain a new predicted trajectory. Specifically, the process of obtaining a new predicted trajectory is as follows:

[0145] C1. Determine the interaction between the unmanned vehicle and surrounding vehicles based on the current state information and historical state information of the unmanned vehicle and the current state information and historical state information of surrounding vehicles, such as overtaking, following, avoiding and meeting of the unmanned vehicle and surrounding vehicles.

[0146] C2. According to the interaction between the driverless vehicle and the surrounding vehicles, the assumed planned trajectory is corrected. When the driverless vehicle is in the process of overtaking, the assumed planned trajectory is shortened; when the driverless vehicle is in the process of following or avoiding, the assumed planned trajectory is lengthened; when the driverless vehicle is in the process of meeting, the assumed planned trajectory is not corrected. In order to ensure that the number of trajectory points of the assumed planned trajectory remains unchanged after the correction, 10 trajectory points are uniformly resampled on the corrected trajectory.

[0147] C3. Determine the right of way priority between the unmanned vehicle and surrounding vehicles based on the positioning system, traffic rules and high-precision maps. For surrounding vehicles with lower right of way priority than the unmanned vehicle, shorten their assumed planned trajectory. Figure 4 As shown in the figure, in the crossroad traffic scene, A is the unmanned vehicle, B is the currently predicted vehicle among the surrounding vehicles, and C is other vehicles. When vehicle A is going straight and vehicle B in the opposite lane turns left, according to traffic regulations, vehicle A has a higher right of way than vehicle B. At this time, vehicle A should continue to go straight and does not need to give way to vehicle B. Therefore, in order to allow the unmanned vehicle to plan a trajectory that continues to go straight as much as possible, the assumed planned trajectory of vehicle B is shortened and corrected, as shown in the figure below. Figure 5 The black dashed line indicates the trajectory.

[0148] C4. Corrected assumed planning trajectory Incorporate it into the three predicted trajectories, and assign a dynamic initial value λ to the confidence of the corrected assumed planning trajectory to obtain four new predicted trajectories Then the softmax function is used to process the confidence and initial value λ of the three predicted trajectories, as follows: Figure 5 As shown, the confidence of the corrected predicted trajectory

[0149]

[0150] Where Softmax(·) is the softmax function, conf k is the confidence of the initial three predicted trajectories, 0<λ<1 is the dynamic initial value, and k=1, 2, 3. To prevent the unmanned vehicle path behavior from being too aggressive, the dynamic initial value λ of the confidence of all assumed planned trajectories that have been processed by growth takes a larger value.

[0151] D. Use the new predicted trajectory as the path planning basis for the unmanned vehicle in the current cycle, and perform path planning for the unmanned vehicle in the current cycle. Specifically, the process of path planning for the unmanned vehicle in the current cycle is as follows:

[0152] The new four predicted trajectories are used as the basis for the current cycle decision and path planning of the unmanned vehicle, and POMDP is used to make decisions, and then the path planning of the unmanned vehicle is realized based on the RRT model. Figure 5 As shown, the gray solid line represents the planned path of the current cycle of the unmanned vehicle

[0153]

[0154] In the formula, info env For traffic environment data, The current status information of the driverless vehicle. For the new 4 prediction trajectories, The initial state sequence of the path of the unmanned vehicle, n = 1, 2, 3, 4.

[0155] In summary, the present invention provides a vehicle path planning method, firstly, according to the state information of the target unmanned vehicle, the state information of the surrounding vehicles, and the traffic environment data of the target unmanned vehicle and the surrounding vehicles, all possible driving trajectories of the surrounding vehicles are predicted, and recorded as the predicted trajectories of the surrounding vehicles; then, according to the path planning of the target unmanned vehicle before the current moment, the path planning of the surrounding vehicles is performed to generate the assumed planning trajectory of the surrounding vehicles; finally, the assumed planning trajectory is corrected, and the corrected assumed planning trajectory is merged with the predicted trajectory, and the path planning of the target unmanned vehicle is performed according to the trajectory merging result, so as to generate at least one driving path that does not cause the target unmanned vehicle to conflict or collide with the surrounding vehicles. Wherein, the target unmanned vehicle includes an unmanned vehicle determined in advance or in real time, and the surrounding vehicles include vehicles in a preset area with the target unmanned vehicle. It can be seen from this that the method can not only predict all possible driving trajectories of the surrounding vehicles, but also can perform path planning for the surrounding vehicles from the perspective of the surrounding vehicles according to the planned path of the unmanned vehicle before the current cycle, and obtain the assumed planning trajectory of the surrounding vehicles, which can be used to supplement and improve the predicted trajectory of the surrounding vehicles, reduce the uncertainty of the trajectory prediction of the surrounding vehicles, and improve the rationality and accuracy of the trajectory prediction results. At the same time, this method can also add the corrected assumed trajectory to the multiple predicted trajectories of the surrounding vehicles, and use them together as the input of the target unmanned vehicle path planning, and perform the path planning of the target unmanned vehicle in the current cycle, and can generate at least one driving path that will prevent the target unmanned vehicle from conflicting or colliding with the surrounding vehicles. This is equivalent to not only considering the interaction between the target unmanned vehicle and the surrounding vehicles, but also improving the rationality and accuracy of the prediction results of the surrounding vehicle trajectories, and improving the initiative and anthropomorphism of the target unmanned vehicle path planning.

[0156] like Figure 7 As shown, the present invention also provides a vehicle path planning system, the system comprising:

[0157] The surrounding vehicle trajectory prediction module 710 is used to predict all possible driving trajectories of the surrounding vehicles based on the state information of the target unmanned vehicle, the state information of the surrounding vehicles, and the traffic environment data of the target unmanned vehicle and the surrounding vehicles, and record them as the predicted trajectories of the surrounding vehicles. As an example, the target unmanned vehicle includes an unmanned vehicle that is predetermined or determined in real time, and the surrounding vehicles include one or more vehicles that are in a preset area with the target unmanned vehicle, and the state information includes current state information and historical state information. Among them, the preset area can be a preset area, a preset distance area, or other preset areas; this embodiment does not impose any limitation on the form of the preset area, so it will not be repeated here. In the embodiment of the present system, the unmanned vehicle can also be referred to as an unmanned vehicle.

[0158] The surrounding vehicle trajectory planning module 720 is used to perform path planning for the surrounding vehicles according to the path planning of the target unmanned vehicle before the current moment, and generate the assumed planned trajectory of the surrounding vehicles. As an example, the path planning of the target unmanned vehicle before the current moment includes but is not limited to: the path planning of the target unmanned vehicle in the previous cycle.

[0159] A trajectory correction module 730, configured to correct the assumed planned trajectory and merge the corrected assumed planned trajectory with the predicted trajectory;

[0160] The unmanned vehicle path planning module 740 is used to plan the path of the target unmanned vehicle according to the trajectory merging result, so as to generate at least one driving path that prevents the target unmanned vehicle from conflicting or colliding with the surrounding vehicles.

[0161] It can be seen that this embodiment first predicts all possible driving trajectories of surrounding vehicles based on the state information of the target unmanned vehicle, the state information of surrounding vehicles, and the traffic environment data of the target unmanned vehicle and surrounding vehicles, and then performs path planning for surrounding vehicles from the perspective of surrounding vehicles based on the planned path of the unmanned vehicle before the current cycle, and obtains the assumed planned trajectory of the surrounding vehicles, which is used to supplement and improve the predicted trajectory of the surrounding vehicles, and can reduce the uncertainty of the trajectory prediction of the surrounding vehicles, and improve the rationality and accuracy of the trajectory prediction results. At the same time, this embodiment can also supplement the corrected assumed trajectory to the multiple predicted trajectories of the surrounding vehicles, and use them together as the input of the path planning of the target unmanned vehicle, and perform the path planning of the target unmanned vehicle in the current cycle, and can generate at least one driving path that does not cause conflict or collision between the target unmanned vehicle and the surrounding vehicles. This is equivalent to not only considering the interactive influence of the target unmanned vehicle and the surrounding vehicles, but also improving the rationality and accuracy of the trajectory prediction results of the surrounding vehicles, and improving the initiative and anthropomorphism of the path planning of the target unmanned vehicle.

[0162] In an exemplary embodiment, the surrounding vehicle trajectory prediction module 710 predicts all possible driving trajectories of the surrounding vehicles according to the state information of the target unmanned vehicle, the state information of the surrounding vehicles, and the traffic environment data of the target unmanned vehicle and the surrounding vehicles. The process may be:

[0163] The historical status information of the target unmanned vehicle and the historical status information of the surrounding vehicles are rendered into a grid map that meets a preset accuracy; wherein the grid map that meets the preset accuracy is obtained based on the traffic environment data. As an example, for example, the traffic environment data in this embodiment may include a grid map that meets the preset accuracy. The accuracy of the grid map may be preset according to actual conditions, and this embodiment does not limit it to a specific value.

[0164] In a raster map that meets a preset accuracy, different color values ​​are used to represent different semantic information, and gradient colors of the same color are used to represent information at different historical time steps, thereby obtaining a raster map containing dynamic information;

[0165] Use convolutional neural networks to encode raster maps containing dynamic information and obtain spatial features containing vehicle interaction information and traffic environment data;

[0166] Using a gated recurrent unit to encode traffic rules, historical state information of the target driverless vehicle, and historical state information of surrounding vehicles to obtain time features;

[0167] The spatial feature and the temporal feature are fused to obtain a spatiotemporal fusion feature; as an example, this embodiment may use a feature concatenation function to fuse the spatial feature and the temporal feature.

[0168] The spatiotemporal fusion features are input into a multi-layer perceptron for decoding, and all possible driving trajectories of the surrounding vehicles in multiple future time steps and the confidence of each driving trajectory are predicted. The confidence of each driving trajectory is used to indicate the credibility or credibility probability of the driving trajectory. The greater the confidence, the higher the credibility of the corresponding driving trajectory, or the greater the confidence, the greater the credibility probability of the corresponding driving trajectory.

[0169] Specifically, the historical status information of the unmanned vehicle and surrounding vehicles is rendered into a three-channel raster map that meets the preset accuracy (hereinafter referred to as the raster high-precision map), and different color values ​​are used to represent different semantic information, while the same color gradient is used to represent the information of different historical time steps, so as to obtain a raster high-precision map M containing dynamic information. r .

[0170] Using lightweight convolutional neural network (CNN) EfficientNet to encode the grid high-precision map M r , and obtain the spatial feature F containing vehicle interaction information and traffic environment data S At the same time, the gated recurrent unit GRU (Gated Recurrent Unit, referred to as GRU) is used to encode the historical state data of the driverless vehicle, the historical state information of the surrounding vehicles and traffic rules to obtain the time feature F T ;

[0171] F S =EfficientNet(M r ; W e );

[0172]

[0173] Where EfficientNet(·) is a lightweight convolutional neural network, M r is a high-precision raster map, W e is the weight of the lightweight convolutional neural network, GRU(·) is the gated recurrent unit, The historical status information of the driverless vehicle. is the historical status information of surrounding vehicles, R t is the traffic rules information, W GRU is the gated recurrent unit weight.

[0174] Use the feature concatenation function Concatenate to concatenate the spatial features F S and time feature F T Fusion is performed to obtain the spatiotemporal fusion feature F ST :

[0175] F ST =Concatenate(F S , F T );

[0176] In the formula, Concatenate(·) is the feature concatenation function, F S is the spatial feature, F T For time characteristics.

[0177] The spatiotemporal fusion feature F ST After two multi-layer perceptron (MLP) decoding, the future T f The multimodal initial prediction trajectory consisting of K initial prediction trajectories at time steps, and the confidence conf of each initial prediction trajectory k ;have:

[0178]

[0179] In the formula, MLP T (·) is a multi-layer perceptron that generates the predicted trajectory, MLP c (·) is the multi-layer perceptron that generates confidence, F ST is the spatiotemporal fusion feature. For multi-layer perceptron MLP T The weight of (·), For multi-layer perceptron MLP c (·) is the weight, Softmax(·) is the softmax function, Traj k is the kth predicted trajectory, conf k is the confidence of the kth trajectory. The confidence refers to the credibility or probability of the predicted trajectory, k = 1, 2, ..., K. The subscript f refers to the future time, and the time step refers to the time required to pass two consecutive position points on a trajectory. Among them, the multimodal initial prediction trajectory composed of K initial prediction trajectories is as follows Figure 4 shown.

[0180] In this embodiment, the method of obtaining the historical status information of the unmanned vehicle and the historical status information of the surrounding vehicles can be: obtaining the historical status information of the unmanned vehicle through the vehicle-mounted sensing system and positioning system of the unmanned vehicle and historical status information of surrounding vehicles The superscript ego represents the driverless vehicle, the superscript tar represents the surrounding vehicles perceived by the driverless vehicle, and the subscript t = -T p ,-T p +1,…,0 represents the time step, such as t=0 is the current time, -T p is the Tth historical moment in the past, the subscript p refers to the historical moment past, p t represents the position information at time t, v t represents the speed information at time t, y t Indicates the heading angle information at time t, l t Represents the length and width of the vehicle.

[0181] In an exemplary embodiment, the surrounding vehicle trajectory planning module 720 performs path planning for the surrounding vehicles according to the path planning of the target unmanned vehicle before the current moment, and the process of generating the assumed planned trajectory of the surrounding vehicles includes: predicting the path initial state sequence of the surrounding vehicles according to the path planning of the target unmanned vehicle before the current moment; processing the path initial state sequence of the surrounding vehicles using a fast search random tree model, performing path planning for the surrounding vehicles from the perspective of the surrounding vehicles, and generating the assumed planned trajectory of the surrounding vehicles. As an example, the process of predicting the path initial state sequence of the surrounding vehicles according to the path planning of the target unmanned vehicle before the current moment may be: establishing a partially observable Markov decision model from the perspective of the surrounding vehicles; using the partially observable Markov decision model to predict the path initial state sequence of the surrounding vehicles from the path planning of the target unmanned vehicle before the current moment, the traffic environment data, and the current state information of the surrounding vehicles.

[0182] Specifically, the surrounding traffic environment data info of each surrounding vehicle in the vehicle coordinate system is obtained through the perception system of the unmanned vehicle, the positioning system of the unmanned vehicle, and the coordinate conversion. env , and use the cache mechanism to obtain the planned path of the unmanned vehicle before the current moment or in the previous cycle. Figure 4 As shown, the gray dotted line represents the planned path of the driverless vehicle in the previous cycle In this embodiment, the cache mechanism refers to a medium or method for storing historical information data of the unmanned vehicle.

[0183] A partially observable Markov decision process (POMDP) ​​is established from the perspective of each surrounding vehicle, and then the initial state sequence of each surrounding vehicle path is obtained using the POMDP model. have:

[0184]

[0185] Where POMDP(·) is a partially observable Markov decision model, info env For traffic environment data, The planned path of the driverless vehicle in the previous cycle, It is the current status information of surrounding vehicles.

[0186] According to the initial state sequence of surrounding vehicles The rapid-exploring random trees (RRT) model is used to plan the future T of surrounding vehicles. f The path trajectory of time steps is used to obtain the assumed planning trajectory of the surrounding vehicles

[0187]

[0188] Where RRT(·) is the fast search random tree model, is the initial state sequence of the paths of surrounding vehicles.

[0189] In an exemplary embodiment, the process of the trajectory correction module 730 correcting the assumed planned trajectory includes: determining the interaction behavior between the target unmanned vehicle and the surrounding vehicles according to the state information of the target unmanned vehicle and the state information of the surrounding vehicles; and correcting the assumed planned trajectory based on the interaction behavior between the target unmanned vehicle and the surrounding vehicles. Wherein, the interaction behavior includes one of the following: overtaking behavior, following behavior, avoiding behavior, and meeting behavior. Specifically, the process of correcting the assumed planned trajectory based on the interaction behavior between the target unmanned vehicle and the surrounding vehicles may be: when the target unmanned vehicle is in overtaking behavior, shortening correction is made to the assumed planned trajectory. When the target unmanned vehicle is in following behavior or avoiding behavior, lengthening correction is made to the assumed planned trajectory. When the target unmanned vehicle is in meeting behavior, the assumed planned trajectory is not corrected. When correcting the assumed planned trajectory, this embodiment can also: determine the right of way priority between the target unmanned vehicle and the surrounding vehicles based on preset traffic prior knowledge, the location information of the target unmanned vehicle and the location information of the surrounding vehicles; obtain the surrounding vehicles whose right of way priority is lower than that of the target unmanned vehicle, and record them as the surrounding vehicles to be corrected; shorten the assumed planned trajectory of the surrounding vehicles to be corrected; wherein the preset traffic prior knowledge includes traffic rules, traffic scenes and maps.

[0190] Specifically, the process of the trajectory correction module 730 correcting the assumed planned trajectory and obtaining a new predicted trajectory is as follows:

[0191] According to the current state information and historical state information of the unmanned vehicle and the current state information and historical state information of the surrounding vehicles, the interactive behavior of the unmanned vehicle and the surrounding vehicles is determined. The interactive behavior includes but is not limited to overtaking, following, avoiding and meeting.

[0192] The assumed planned trajectory is corrected according to the interaction stage between the unmanned vehicle and the surrounding vehicles. When the unmanned vehicle is in overtaking behavior, the assumed planned trajectory is shortened; when the unmanned vehicle is in following or avoiding behavior, the assumed planned trajectory is lengthened; when the unmanned vehicle is in meeting behavior, the assumed planned trajectory is not corrected. In order to ensure that the number of trajectory points of the assumed planned trajectory remains unchanged after the correction process, this embodiment can re-sample 10 trajectory points evenly on the corrected trajectory.

[0193] According to the positioning system, traffic rules and high-precision maps, the road priority between the unmanned vehicle and surrounding vehicles is determined. For surrounding vehicles with lower road priority than the unmanned vehicle, their assumed planned trajectory is shortened. Figure 4 As shown, in the traffic scene at the crossroad, A is the unmanned vehicle, B is the currently predicted vehicle among the surrounding vehicles, and C is other vehicles. When vehicle A is going straight and vehicle B in the opposite lane turns left, according to traffic regulations, vehicle A has a higher right of way priority than vehicle B. At this time, vehicle A should continue to go straight and does not have to give way to vehicle B. Therefore, in order to allow the unmanned vehicle to plan a trajectory that continues to go straight as much as possible, this embodiment can shorten the assumed planned trajectory of vehicle B. Figure 5 As shown in the figure, the assumed planned trajectory of the surrounding vehicle B is shortened and corrected as follows Figure 5 The black dashed line indicates the trajectory.

[0194] In an exemplary embodiment, the trajectory correction module 730 merges the corrected assumed planned trajectory with the predicted trajectory, and the process of performing path planning for the target unmanned vehicle based on the trajectory merging result includes: adding the corrected assumed planned trajectory to the predicted trajectory for merging to obtain a new predicted trajectory; using the new predicted trajectory as the path planning condition of the target unmanned vehicle in the current cycle, and using the partially observable Markov decision model to obtain the path initial state sequence of the target unmanned vehicle from the new predicted trajectory, the current state information of the target unmanned vehicle, and the traffic environment data; using the fast search random tree model to process the path initial state sequence of the target unmanned vehicle, and perform path planning for the target unmanned vehicle to generate at least one driving path that prevents the target unmanned vehicle from conflicting or colliding with the surrounding vehicles. In this embodiment, the trajectory correction module 730 adds the corrected assumed planned trajectory to the predicted trajectory for merging, and after obtaining a new predicted trajectory, this embodiment may also include: obtaining the confidence of the predicted trajectory generated when predicting all possible driving trajectories of surrounding vehicles; assigning a dynamic initial value to the corrected assumed planned trajectory as the confidence of the corrected assumed planned trajectory; the confidence is used to indicate the credibility or credibility probability of the trajectory; the confidence of the predicted trajectory before merging and the dynamic initial value are normalized to correct the confidence of the new predicted trajectory. Among them, the confidence is used to indicate the credibility or credibility probability of the trajectory.

[0195] According to the above records, specifically, this embodiment can modify the assumed planned trajectory Merge it into the K prediction trajectories predicted initially to obtain a new prediction trajectory, that is, to obtain K+1 prediction trajectories Then a dynamic initial value λ is assigned to the assumed planning trajectory before merging as the confidence of the assumed planning trajectory before merging; and the confidence conf of the initial K predicted trajectories is calculated using the normalized function softmax. k And the initial value λ is processed to correct the confidence of the new predicted trajectory have:

[0196]

[0197] Where Softmax(·) is the softmax function, conf k is the confidence of the initial K predicted trajectories, 0<λ<1 is the dynamic initial value, k=1, 2, ..., K. To prevent the unmanned vehicle from being too aggressive in its path behavior, the dynamic initial value of the confidence of all assumed planned trajectories that have undergone the growth process in this embodiment can be a larger value.

[0198] The unmanned vehicle path planning module 740 performs path planning for the target unmanned vehicle according to the trajectory merging result to generate at least one driving path that prevents the target unmanned vehicle from conflicting or colliding with the surrounding vehicles. The process may be: using the new K+1 predicted trajectories as the path and planning basis of the current cycle of the unmanned vehicle, and using the POMDP and RRT models to obtain the planned path of the unmanned vehicle.

[0199]

[0200] In the formula, info env For traffic environment data, is the current status information of the driverless vehicle, is the predicted K+1 trajectories, The initial state sequence of the path of the unmanned vehicle, k = 1, 2, ..., K.

[0201] In summary, the present invention provides a vehicle path planning system, firstly, according to the state information of the target unmanned vehicle, the state information of the surrounding vehicles, and the traffic environment data of the target unmanned vehicle and the surrounding vehicles, all possible driving trajectories of the surrounding vehicles are predicted, and recorded as the predicted trajectories of the surrounding vehicles; then, according to the path planning of the target unmanned vehicle before the current moment, the path planning of the surrounding vehicles is performed to generate the assumed planning trajectory of the surrounding vehicles; finally, the assumed planning trajectory is corrected, and the corrected assumed planning trajectory is merged with the predicted trajectory, and the path planning of the target unmanned vehicle is performed according to the trajectory merging result, so as to generate at least one driving path that does not cause the target unmanned vehicle to conflict or collide with the surrounding vehicles. Among them, the target unmanned vehicle includes an unmanned vehicle that is predetermined or determined in real time, and the surrounding vehicles include vehicles that are in a preset area with the target unmanned vehicle. It can be seen from this that the system can not only predict all possible driving trajectories of the surrounding vehicles, but also can perform path planning for the surrounding vehicles from the perspective of the surrounding vehicles according to the planned path of the unmanned vehicle before the current cycle, and obtain the assumed planning trajectory of the surrounding vehicles, which can be used to supplement and improve the predicted trajectory of the surrounding vehicles, reduce the uncertainty of the trajectory prediction of the surrounding vehicles, and improve the rationality and accuracy of the trajectory prediction results. At the same time, the system can also add the corrected assumed trajectory to the multiple predicted trajectories of the surrounding vehicles, and use them together as the input of the target unmanned vehicle path planning, and perform the path planning of the target unmanned vehicle in the current cycle, and can generate at least one driving path that will prevent the target unmanned vehicle from conflicting or colliding with the surrounding vehicles. This means that the system not only considers the interaction between the target unmanned vehicle and the surrounding vehicles, but also improves the rationality and accuracy of the prediction results of the surrounding vehicle trajectories, and improves the initiative and anthropomorphism of the target unmanned vehicle path planning.

[0202] It should be noted that the vehicle path planning system provided in the above embodiment and the vehicle path planning method provided in the above embodiment belong to the same concept, wherein the specific manner in which each module performs the operation has been described in detail in the method embodiment and will not be repeated here. In practical applications, the vehicle path planning system provided in the above embodiment can allocate the above functions to different functional modules as needed, that is, divide the internal structure of the device into different functional modules to complete all or part of the functions described above, and this is not limited here.

[0203] In another exemplary embodiment, this embodiment further provides an unmanned vehicle, wherein the unmanned vehicle travels according to a preset driving path, and the generation process of the preset driving path includes:

[0204] According to the state information and traffic environment data of the unmanned vehicle, as well as the state information and traffic environment data of surrounding vehicles, all possible driving trajectories of the surrounding vehicles are predicted, and recorded as predicted trajectories of the surrounding vehicles;

[0205] Performing path planning for the surrounding vehicles according to the path planning of the unmanned vehicle before the current moment, and generating an assumed planned trajectory of the surrounding vehicles;

[0206] Correcting the assumed planned trajectory, merging the corrected assumed planned trajectory with the predicted trajectory, and performing path planning for the unmanned vehicle according to the trajectory merging result to generate at least one driving path in which the target unmanned vehicle will not conflict or collide with the surrounding vehicles;

[0207] The surrounding vehicles include vehicles that are within a preset area with the unmanned vehicle.

[0208] It should be noted that the method of generating a preset driving path provided in the above embodiment and the method of generating a driving path by the vehicle path planning method provided in the above embodiment belong to the same concept, and the specific method of executing each step has been described in detail in the method embodiment and will not be repeated here.

[0209] An embodiment of the present application also provides an electronic device, comprising: one or more processors; a storage device for storing one or more programs, and when the one or more programs are executed by the one or more processors, the electronic device implements the vehicle path planning method provided in the above-mentioned embodiments.

[0210] Figure 8 The structure diagram of the computer system suitable for implementing the electronic device of the embodiment of the present application is shown. It should be noted that: Figure 8 The computer system 800 of the electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0211] like Figure 8As shown, the computer system 800 includes a central processing unit (CPU) 801, which can perform various appropriate actions and processes according to the program stored in the read-only memory (ROM) 802 or the program loaded from the storage part 808 to the random access memory (RAM) 803, such as executing the method described in the above embodiment. In the RAM 803, various programs and data required for system operation are also stored. The CPU 801, ROM 802 and RAM 803 are connected to each other through a bus 804. An input / output (I / O) interface 805 is also connected to the bus 804.

[0212] The following components are connected to the I / O interface 805: an input section 806 including a keyboard, a mouse, etc.; an output section 807 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker; a storage section 808 including a hard disk, etc.; and a communication section 809 including a network interface card such as a LAN (Local Area Network) card, a modem, etc. The communication section 809 performs communication processing via a network such as the Internet. A drive 810 is also connected to the I / O interface 805 as needed. A removable medium 811, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 810 as needed so that a computer program read therefrom is installed into the storage section 808 as needed.

[0213] In particular, according to an embodiment of the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present application includes a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a computer program for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through a communication section 809, and / or installed from a removable medium 811. When the computer program is executed by a central processing unit (CPU) 801, various functions defined in the system of the present application are executed.

[0214] It should be noted that the computer-readable medium shown in the embodiment of the present application can be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium can be, for example, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media can include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a flash memory, an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present application, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, wherein a computer-readable computer program is carried. This propagated data signal can take a variety of forms, including but not limited to an electromagnetic signal, an optical signal, or any suitable combination of the above. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium, which may send, propagate or transmit a program for use by or in combination with an instruction execution system, apparatus or device. A computer program contained on a computer-readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wired, etc., or any suitable combination of the above.

[0215] The flowchart and block diagram in the accompanying drawings illustrate the possible architecture, functions and operations of the system, method and computer program product according to various embodiments of the present application. Wherein, each box in the flowchart or block diagram can represent a module, a program segment, or a part of the code, and the above-mentioned module, program segment, or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flowchart, and the combination of boxes in the block diagram or flowchart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0216] The units involved in the embodiments described in this application may be implemented by software or hardware, and the units described may also be set in a processor. The names of these units do not, in some cases, constitute limitations on the units themselves.

[0217] Another aspect of the present application also provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the vehicle path planning method as described above is implemented. The computer-readable storage medium may be included in the electronic device described in the above embodiment, or may exist independently without being assembled into the electronic device.

[0218] Another aspect of the present application also provides a computer program product or a computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the vehicle path planning method provided in each of the above embodiments.

[0219] The above embodiments are merely illustrative of the principles and effects of the present invention, and are not intended to limit the present invention. Anyone familiar with the technology may modify or change the above embodiments without violating the spirit and scope of the present invention. Therefore, all equivalent modifications or changes made by a person of ordinary skill in the art without departing from the spirit and technical ideas disclosed by the present invention shall still be covered by the claims of the present invention.

Claims

1. A vehicle path planning method, characterized in that: The method comprises the following steps: According to the state information of the target unmanned vehicle, the state information of the surrounding vehicles, and the traffic environment data of the target unmanned vehicle and the surrounding vehicles, all possible driving trajectories of the surrounding vehicles are predicted and recorded as the predicted trajectories of the surrounding vehicles; Performing path planning for the surrounding vehicles according to the path planning of the target unmanned vehicle before the current moment, and generating an assumed planned trajectory of the surrounding vehicles; Correcting the assumed planned trajectory, merging the corrected assumed planned trajectory with the predicted trajectory, and performing path planning for the target unmanned vehicle according to the trajectory merging result to generate at least one driving path in which the target unmanned vehicle will not conflict or collide with the surrounding vehicles; Wherein, the target unmanned vehicle includes an unmanned vehicle determined in advance or in real time, and the surrounding vehicles include vehicles in a preset area with the target unmanned vehicle; The process of predicting all possible driving trajectories of the surrounding vehicles includes: Rendering the historical status information of the target unmanned vehicle and the historical status information of the surrounding vehicles into a grid map that meets a preset accuracy; wherein the grid map that meets the preset accuracy is obtained based on the traffic environment data; and, in the grid map that meets the preset accuracy, using different color values ​​to represent different semantic information, and using gradient colors of the same color to represent information at different historical time steps, to obtain a grid map containing dynamic information; Convolutional neural networks are used to encode grid maps containing dynamic information to obtain spatial features containing vehicle interaction information and traffic environment data; there are: F S =EfficientNet(M r ; W e ), where F S is the spatial feature containing vehicle interaction information and traffic environment data, EfficientNet(·) is a lightweight convolutional neural network, and M r W is a high-precision grid map containing dynamic information. e is the weight of the lightweight convolutional neural network; The gated recurrent unit is used to encode traffic rules, the historical state information of the target driverless vehicle and the historical state information of the surrounding vehicles to obtain time features; there are: In the formula, F T is the temporal feature, GRU(·) is the gated recurrent unit, The historical status information of the driverless vehicle. is the historical status information of surrounding vehicles, R t is the traffic rules information, W GRU is the gated recurrent unit weight; The spatial feature and the temporal feature are fused to obtain the spatiotemporal fusion feature; there are: ST =Concatenate(F S ,F T ), where F ST is the spatiotemporal fusion feature, Concatenate(·) is the feature concatenation function; The spatiotemporal fusion features are input into a multi-layer perceptron for decoding, and all possible driving trajectories of the surrounding vehicles in multiple future time steps are predicted; there are: In the formula, in the formula, MLP T (·) is a multi-layer perceptron that generates the predicted trajectory, MLP c (·) is the multi-layer perceptron that generates confidence, F ST is the spatiotemporal fusion feature. MLP is a multi-layer perceptron T The weight of (·), MLP is a multi-layer perceptron c (·) is the weight, Softmax(·) is the softmax function, Traj k is the kth predicted trajectory, conf k is the confidence of the kth trajectory. Confidence refers to the credibility or probability of the predicted trajectory, k = 1, 2, ..., K. The subscript f refers to the future time, and the time step refers to the time required to pass two consecutive position points on a trajectory.

2. The vehicle path planning method according to claim 1, characterized in that: The process of performing path planning for the surrounding vehicles according to the path planning of the target unmanned vehicle before the current moment and generating the assumed planned trajectory of the surrounding vehicles includes: Predicting the path initial state sequence of the surrounding vehicles according to the path planning of the target unmanned vehicle before the current moment; The initial state sequence of the paths of the surrounding vehicles is processed by using a fast search random tree model, and the paths of the surrounding vehicles are planned from the perspective of the surrounding vehicles to generate an assumed planned trajectory of the surrounding vehicles.

3. The vehicle path planning method according to claim 2, characterized in that: The process of predicting the path initial state sequence of the surrounding vehicles according to the path planning of the target unmanned vehicle before the current moment includes: Establishing a partially observable Markov decision model based on the perspective of the surrounding vehicles; The partially observable Markov decision model is used to predict the path initial state sequence of the surrounding vehicles from the path planning of the target unmanned vehicle before the current moment, the traffic environment data and the current state information of the surrounding vehicles.

4. The vehicle path planning method according to any one of claims 1 to 3, characterized in that: The process of correcting the assumed planned trajectory includes: Determining the interaction behavior between the target unmanned vehicle and the surrounding vehicles according to the state information of the target unmanned vehicle and the state information of the surrounding vehicles; Modifying the assumed planned trajectory based on the interaction behavior between the target unmanned vehicle and the surrounding vehicles; The interactive behavior includes one of the following: overtaking behavior, following behavior, avoiding behavior, and meeting behavior.

5. The vehicle path planning method according to claim 4, characterized in that: The process of correcting the assumed planned trajectory based on the interactive behavior between the target unmanned vehicle and the surrounding vehicles includes: When the target unmanned vehicle is in overtaking behavior, shortening and correcting the assumed planned trajectory; When the target unmanned vehicle is in a following behavior or an avoidance behavior, performing an increase correction on the assumed planned trajectory; When the target unmanned vehicle is in a meeting behavior, the assumed planned trajectory is not corrected.

6. The vehicle path planning method according to claim 4, characterized in that: When the assumed planned trajectory is corrected, the method further includes: Determining the right-of-way priority between the target unmanned vehicle and the surrounding vehicles according to preset traffic prior knowledge, the location information of the target unmanned vehicle, and the location information of the surrounding vehicles; Obtain surrounding vehicles with a lower right of way priority than the target unmanned vehicle, and record them as surrounding vehicles to be corrected; Performing a shortening correction on the assumed planned trajectory of the surrounding vehicles to be corrected; The preset traffic prior knowledge includes traffic rules, traffic scenarios and maps.

7. The vehicle path planning method according to claim 3, characterized in that: The process of merging the corrected assumed planned trajectory with the predicted trajectory and performing path planning for the target unmanned vehicle according to the trajectory merging result includes: Adding the corrected assumed planned trajectory to the predicted trajectory for merging to obtain a new predicted trajectory; Using the new predicted trajectory as a path planning condition for the target unmanned vehicle in the current cycle, and using the partially observable Markov decision model to obtain a path initial state sequence of the target unmanned vehicle from the new predicted trajectory, the current state information of the target unmanned vehicle, and the traffic environment data; The fast search random tree model is used to process the path initial state sequence of the target unmanned vehicle, and the path of the target unmanned vehicle is planned to generate at least one driving path that prevents the target unmanned vehicle from conflicting or colliding with the surrounding vehicles.

8. The vehicle path planning method according to claim 7, characterized in that: After adding the corrected assumed planned trajectory to the predicted trajectory for merging, and obtaining a new predicted trajectory, the method further includes: Obtaining the confidence of the predicted trajectory generated when predicting all possible driving trajectories of the surrounding vehicles; wherein the confidence is used to indicate the credibility or credibility probability of the trajectory; Assigning a dynamic initial value to the corrected assumed planning trajectory as the confidence of the corrected assumed planning trajectory; The confidence of the predicted trajectory before merging and the dynamic initial value are normalized to correct the confidence of the new predicted trajectory.

9. A vehicle path planning system, characterized in that: The system comprises: A surrounding vehicle trajectory prediction module, used to predict all possible driving trajectories of the surrounding vehicles according to the state information of the target unmanned vehicle, the state information of the surrounding vehicles, and the traffic environment data of the target unmanned vehicle and the surrounding vehicles, and record them as the predicted trajectories of the surrounding vehicles; A surrounding vehicle trajectory planning module, used to perform path planning for the surrounding vehicles according to the path planning of the target unmanned vehicle before the current moment, and generate an assumed planned trajectory of the surrounding vehicles; A trajectory correction module, used to correct the assumed planned trajectory and merge the corrected assumed planned trajectory with the predicted trajectory; An unmanned vehicle path planning module, used for performing path planning for the target unmanned vehicle according to the trajectory merging result, so as to generate at least one driving path in which the target unmanned vehicle will not conflict or collide with the surrounding vehicles; Wherein, the target unmanned vehicle includes an unmanned vehicle determined in advance or in real time, and the surrounding vehicles include vehicles in a preset area with the target unmanned vehicle; The process of the surrounding vehicle trajectory prediction module predicting all possible driving trajectories of the surrounding vehicles includes: Rendering the historical status information of the target unmanned vehicle and the historical status information of the surrounding vehicles into a grid map that meets a preset accuracy; wherein the grid map that meets the preset accuracy is obtained based on the traffic environment data; and, in the grid map that meets the preset accuracy, using different color values ​​to represent different semantic information, and using gradient colors of the same color to represent information at different historical time steps, to obtain a grid map containing dynamic information; Convolutional neural networks are used to encode grid maps containing dynamic information to obtain spatial features containing vehicle interaction information and traffic environment data; there are: F S =EfficientNet(M r ; W e ), where F S is the spatial feature containing vehicle interaction information and traffic environment data, EfficientNet(·) is a lightweight convolutional neural network, and M r W is a high-precision grid map containing dynamic information. e is the weight of the lightweight convolutional neural network; The gated recurrent unit is used to encode traffic rules, the historical state information of the target driverless vehicle and the historical state information of the surrounding vehicles to obtain time features; there are: In the formula, F T is the temporal feature, GRU(·) is the gated recurrent unit, The historical status information of the driverless vehicle. is the historical status information of surrounding vehicles, R t is the traffic rules information, W GRU is the gated recurrent unit weight; The spatial feature and the temporal feature are fused to obtain the spatiotemporal fusion feature; there are: ST =Concatenate(F S ,F T ), where F ST is the spatiotemporal fusion feature, Concatenate(·) is the feature concatenation function; The spatiotemporal fusion features are input into a multi-layer perceptron for decoding, and all possible driving trajectories of the surrounding vehicles in multiple future time steps are predicted; there are: In the formula, in the formula, MLP T (·) is a multi-layer perceptron that generates the predicted trajectory, MLP c (·) is the multi-layer perceptron that generates confidence, F ST is the spatiotemporal fusion feature. MLP is a multi-layer perceptron T The weight of (·), MLP is a multi-layer perceptron c (·) is the weight, Softmax(·) is the softmax function, Traj k is the kth predicted trajectory, conf k is the confidence of the kth trajectory. Confidence refers to the credibility or probability of the predicted trajectory, k = 1, 2, ..., K. The subscript f refers to the future time, and the time step refers to the time required to pass two consecutive position points on a trajectory.

10. An unmanned vehicle, characterized in that: The unmanned vehicle travels according to a preset travel path, and the generation process of the preset travel path includes: According to the state information and traffic environment data of the unmanned vehicle, as well as the state information and traffic environment data of surrounding vehicles, all possible driving trajectories of the surrounding vehicles are predicted, and recorded as predicted trajectories of the surrounding vehicles; Performing path planning for the surrounding vehicles according to the path planning of the unmanned vehicle before the current moment, and generating an assumed planned trajectory of the surrounding vehicles; Correcting the assumed planned trajectory, merging the corrected assumed planned trajectory with the predicted trajectory, and performing path planning for the unmanned vehicle according to the trajectory merging result to generate at least one driving path in which the target unmanned vehicle will not conflict or collide with the surrounding vehicles; The surrounding vehicles include vehicles in a preset area with the unmanned vehicle; The process of predicting all possible driving trajectories of the surrounding vehicles includes: Rendering the historical status information of the target unmanned vehicle and the historical status information of the surrounding vehicles into a grid map that meets a preset accuracy; wherein the grid map that meets the preset accuracy is obtained based on the traffic environment data; and, in the grid map that meets the preset accuracy, using different color values ​​to represent different semantic information, and using gradient colors of the same color to represent information at different historical time steps, to obtain a grid map containing dynamic information; Convolutional neural networks are used to encode grid maps containing dynamic information to obtain spatial features containing vehicle interaction information and traffic environment data; there are: F S =EfficientNet(M r ; W e ), where F S is the spatial feature containing vehicle interaction information and traffic environment data, EfficientNet(·) is a lightweight convolutional neural network, and M r W is a high-precision grid map containing dynamic information. e is the weight of the lightweight convolutional neural network; The gated recurrent unit is used to encode traffic rules, the historical state information of the target driverless vehicle and the historical state information of the surrounding vehicles to obtain time features; there are: In the formula, F T is the temporal feature, GRU(·) is the gated recurrent unit, The historical status information of the driverless vehicle. is the historical status information of surrounding vehicles, R t is the traffic rules information, W GRU is the gated recurrent unit weight; The spatial feature and the temporal feature are fused to obtain the spatiotemporal fusion feature; there are: ST =Concatenate(F S ,F T ), where F ST is the spatiotemporal fusion feature, Concatenate(·) is the feature concatenation function; The spatiotemporal fusion features are input into a multi-layer perceptron for decoding, and all possible driving trajectories of the surrounding vehicles in multiple future time steps are predicted; there are: In the formula, in the formula, MLP T (·) is a multi-layer perceptron that generates the predicted trajectory, MLP c (·) is the multi-layer perceptron that generates confidence, F ST is the spatiotemporal fusion feature. MLP is a multi-layer perceptron T The weight of (·), MLP is a multi-layer perceptron c (·) is the weight, Softmax(·) is the softmax function, Traj k is the kth predicted trajectory, conf k is the confidence of the kth trajectory. Confidence refers to the credibility or probability of the predicted trajectory, k = 1, 2, ..., K. The subscript f refers to the future time, and the time step refers to the time required to pass two consecutive position points on a trajectory.

11. An electronic device, characterized in that: include: one or more processors; A storage device for storing one or more programs, when the one or more programs are executed by the one or more processors, enables the electronic device to implement the vehicle path planning method as described in any one of claims 1 to 8.

12. A computer-readable storage medium, characterized in that: Computer-readable instructions are stored thereon, and when the computer-readable instructions are executed by a processor of a computer, the computer is caused to execute the vehicle path planning method according to any one of claims 1 to 8.

Images