Local path planning method for autonomous vehicle

Abstract

The invention relates to the technical field of automatic driving, and provides a local path planning method for an automatic driving vehicle, which comprises the following steps: constructing an interaction point model, and judging the traffic priority of a target vehicle and an interaction vehicle; under the condition that the target vehicle passes preferentially, a path point sequence containing observation points and interaction points is expanded based on a displacement-time image analysis method, a plurality of planned paths are obtained, and the planned paths comprise a plurality of planned paths of the target vehicle under a plurality of path actions possibly executed by the corresponding interaction vehicle; the output of the interaction point model serves as the input of a multi-layer perceptron model, priority judgment is conducted on multiple planned paths, a target path is obtained, and path sections from the current position point of the target vehicle to the observation point in the multiple planned paths are common paths; and navigating the target vehicle based on the target path. The problems that in the prior art, a path planning method of an automatic driving vehicle is low in safety and accuracy in a complex scene, and parameter debugging is complex are solved.

Description

Technical Field

[0001] The invention relates to the field of autonomous driving technology, and in particular to a local path planning method for an autonomous driving vehicle. Background Art

[0002] As autonomous driving technology develops rapidly, achieving safe and efficient interaction with other traffic participants has become one of the most core requirements of autonomous driving systems. This is particularly prominent at intersections without traffic lights or in driving scenarios with blind spots in the vehicle's field of vision.

[0003] In practical applications, in order to adapt to different road conditions and the behavior patterns of interacting vehicles, researchers need to repeatedly debug and optimize various parameters in the model based on the POMDP (Partially Observable Markov Decision Process) method, which consumes a lot of time and effort. Even after such adjustments, it is difficult to guarantee the safety and accuracy of this method in more complex real-life scenarios.

[0004] The vehicle tracking model method tracks and predicts the motion trajectory of interactive vehicles, mainly for a single driving scenario. When the driving scenario changes, such as switching from urban roads to rural roads, or from a sunny environment to a rainy and snowy weather environment, the parameters in the model need to be re-adapted and adjusted, requiring professionals to spend a lot of time collecting and analyzing data in the new scenario in order to accurately adjust the parameters.

[0005] The path planning methods for autonomous vehicles in related technologies have low safety and accuracy in complex scenarios, and complex parameter debugging. No effective solution has been proposed so far. Summary of the invention

[0006] The present invention provides a local path planning method for an autonomous driving vehicle, which at least solves the problems in the related art of path planning methods for autonomous driving vehicles in complex scenarios, such as low safety and accuracy, and complex parameter debugging.

[0007] The local path planning method for an autonomous driving vehicle provided by an embodiment of the present invention includes: constructing an interaction point model, and judging the passage priority of a target vehicle and an interactive vehicle based on the interaction point model; when the target vehicle has priority passage, constructing a path point sequence including observation points and interaction points, and expanding the path point sequence based on a displacement-time image analysis method to obtain multiple planned paths, wherein the multiple planned paths include multiple planned paths for the target vehicle corresponding to multiple path actions that may be performed by the interactive vehicle; judging the priority of the multiple planned paths based on a multi-layer perceptron model to obtain a target path, wherein the output of the interaction point model is used as the input of the multi-layer perceptron model, and the path segment from the current position point of the target vehicle to the observation point in the multiple planned paths is a common path; and navigating the target vehicle based on the target path.

[0008] The local path planning method for an autonomous driving vehicle provided by an embodiment of the present invention performs priority judgment on multiple planned paths based on a multi-layer perceptron model to obtain a target path, including: comparing the priority probability output by the multi-layer perceptron model with a preset constant threshold layer by layer in order from the lowest layer node to the highest layer node, to determine whether the multiple planned paths are priority paths, wherein the multiple planned paths are determined based on a discrete network obtained by expanding a path point sequence, and the discrete network includes multiple layers of nodes; when the priority probabilities of the multiple planned paths corresponding to the nodes of each layer are all greater than the preset constant threshold, the multiple planned paths are all used as priority paths, wherein the priority of the target vehicle traveling at the position of the nodes of each layer corresponding to the priority path is higher than that of the interactive vehicle; and the farthest priority path is used as the target path.

[0009] The local path planning method for an autonomous driving vehicle provided by an embodiment of the present invention, when the priority probability of a target layer node corresponding to any one of the multiple planned paths is lower than a preset constant threshold, a safety judgment is performed on the planned path to be extended according to the judgment layer node, wherein the target layer node is any layer node among the multi-layer nodes passed by the planned path, the judgment layer node is the previous layer node of the target layer node, and the planned path to be extended is a planned path whose priority probability corresponding to the target layer node is lower than the preset constant threshold; when the planned path to be extended passes the safety judgment, the part of the planned path to be extended from the judgment layer node is discarded, and the planned path to be extended is extended from the judgment layer node to obtain the extended planned path; when the planned path to be extended does not pass the safety judgment, the previous layer node of the judgment layer node is used as a new judgment layer node, and the planned path to be extended is subjected to a safety judgment according to the new judgment layer node until the safety judgment is passed; in the order from the judgment layer node that passes the safety judgment to the highest layer node, the priority probability of the nodes of each layer corresponding to the extended planned path is compared with the preset constant threshold layer by layer until a priority path is obtained.

[0010] The local path planning method for an autonomous driving vehicle provided by an embodiment of the present invention constructs an interaction point model, including: calculating the interaction protection time based on the arrival time interval, wherein the arrival time interval is the difference in time between the target vehicle and the interaction vehicle arriving at the interaction point; constructing the interaction protection time as a function of relative speed and relative angle, and determining the boundary of the interaction protection time according to the function, wherein the relative speed is the difference in speed between the target vehicle and the interaction vehicle arriving at the interaction point, and the relative angle is the difference in yaw angle between the target vehicle and the interaction vehicle arriving at the interaction point; constructing the interaction point model based on the boundary of the interaction protection time.

[0011] The local path planning method for an autonomous driving vehicle provided by an embodiment of the present invention calculates the interaction protection time based on the arrival time interval, including: when the arrival time interval is negative, the target vehicle has priority to pass the interaction point, and the interaction protection time is the maximum value of the arrival time interval; when the arrival time interval is positive, the interaction vehicle has priority to pass the interaction point, and the interaction protection time is the minimum value of the arrival time interval.

[0012] The local path planning method for an autonomous driving vehicle provided by an embodiment of the present invention constructs an interaction point model based on the boundary of the interaction protection time, including: determining a speed limit based on the Euler distance between the target vehicle and the interaction point, and the minimum value of the arrival time interval, wherein the speed limit is the speed limit of the target vehicle traveling to the observation point; and constructing the interaction point model based on the boundary of the interaction protection time and the speed limit.

[0013] The local path planning method for an autonomous driving vehicle provided by an embodiment of the present invention constructs an interaction point model based on the boundary of the interaction protection time and the speed limit, including: calculating the minimum arrival time and the maximum arrival time of the interaction vehicle based on the initial speed of the interaction vehicle, the Euler distance between the interaction vehicle and the interaction point, and the maximum and minimum values ​​of the average acceleration of the interaction vehicle to the interaction point; calculating the minimum arrival time and the maximum arrival time of the target vehicle based on the initial speed of the target vehicle, the Euler distance between the target vehicle and the interaction point, and the maximum and minimum values ​​of the average acceleration of the target vehicle to the interaction point; calculating the overtaking ability of the target vehicle based on the minimum arrival time of the target vehicle, the minimum arrival time of the interaction vehicle, and the maximum value of the arrival time interval; calculating the yielding ability of the target vehicle based on the maximum arrival time of the target vehicle, the maximum arrival time of the interaction vehicle, and the minimum value of the arrival time interval; constructing the interaction point model based on the boundary of the interaction protection time, the speed limit, the overtaking ability and the yielding ability.

[0014] The local path planning method for an autonomous driving vehicle provided by an embodiment of the present invention expands a path point sequence based on a displacement-time image analysis method to obtain multiple planned paths, including: constructing a search space based on the displacement-time image analysis method; using a speed constraint as a spatiotemporal constraint in the search space to obtain a discrete network, wherein the speed constraint includes a curvature of a corresponding position of the path point sequence and a speed upper limit, and a node in the discrete network is a vector related to speed, time and displacement, and each path point sequence includes multiple nodes; expanding the path point sequence in the discrete network to obtain multiple planned paths.

[0015] The local path planning method for an autonomous driving vehicle provided by an embodiment of the present invention expands a path point sequence in a discrete network to obtain multiple planned paths, including: discarding nodes in the discrete network that exceed a preset cost to obtain a target discrete network, wherein the preset cost includes a preset time limit or a preset speed limit; and expanding the path point sequence in the target discrete network to obtain multiple planned paths.

[0016] The local path planning method for an autonomous driving vehicle provided by an embodiment of the present invention navigates a target vehicle based on a target path, including: smoothing the target path based on a quintic segmented Bezier curve to obtain a smooth target path; and navigating the target vehicle based on the smooth target path.

[0017] An electronic device provided by an embodiment of the present invention includes: a processor and a memory storing a program, wherein the program includes instructions, and when the instructions are executed by the processor, the processor executes any of the above methods.

[0018] The invention provides a local path planning method for an autonomous driving vehicle. In the face of complex application scenarios, the traffic priority between the target vehicle and the interactive vehicle is first determined based on the interaction point model, and local path planning is performed when the target vehicle has priority. This can reduce the amount of calculation and the difficulty of parameter debugging. In the path planning, multiple planned paths are generated and priority judgment is performed, which can improve the safety and accuracy of navigation. This solves the problem that the path planning method for autonomous driving vehicles in related technologies has low safety and accuracy in complex scenarios and complex parameter debugging. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some embodiments of the present invention, and for ordinary technicians in this field, other embodiments can be obtained based on these drawings without creative work.

[0020] Figure 1 It is a step flow chart of a local path planning method for an autonomous driving vehicle in an embodiment of the present invention.

[0021] Figure 2 In the embodiment of the invention, the target vehicle and the interactive vehicle pass through the interactive point. Speed ​​diagram.

[0022] Figure 3 It is a schematic diagram of an application scenario for setting a speed upper limit in an embodiment of the invention.

[0023] Figure 4 It is a schematic diagram of expanding nodes in an embodiment of the invention.

[0024] Figure 5 Schematic diagram of comparison of three planning paths in the embodiment of the present invention.

[0025] Figure 6 It is a structural schematic diagram of the electronic device created by the present invention. DETAILED DESCRIPTION

[0026] The embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be construed as being limited to the embodiments described herein, which are instead provided to provide a more thorough and complete understanding of the present invention. It should be understood that the drawings and embodiments of the present invention are only for exemplary purposes and are not intended to limit the scope of protection of the present invention.

[0027] One of the core requirements of an autonomous driving system is to achieve safe and efficient interaction with other vehicles, which is particularly prominent at intersections without traffic lights or in driving scenarios with blind spots. The path planning methods of autonomous driving vehicles in related technologies have low safety and accuracy in complex scenarios, and complex parameter debugging.

[0028] To this end, the present invention provides an embodiment of a local path planning method for an autonomous driving vehicle.

[0029] Please refer to Figure 1 As shown, the above method includes:

[0030] Step S101: construct an interaction point model, and determine the traffic priority between the target vehicle and the interaction vehicle based on the interaction point model.

[0031] Step S102, when the target vehicle has priority, construct a path point sequence including observation points and interaction points, expand the path point sequence based on the displacement-time image analysis method, and obtain multiple planned paths, wherein the multiple planned paths include multiple planned paths for the target vehicle corresponding to multiple path actions that may be performed by the interactive vehicle.

[0032] Step S103, based on the multi-layer perceptron model, priority is determined for multiple planned paths to obtain the target path, wherein the output of the interaction point model is used as the input of the multi-layer perceptron model, and the path segment from the current position point of the target vehicle to the observation point in the multiple planned paths is a common path.

[0033] Step S104: navigating the target vehicle based on the target path.

[0034] It can be understood that the target vehicle is an autonomous driving vehicle to be navigated, and the vehicle is equipped with a navigation system, and the navigation system applies the above method provided by the embodiment of the invention.

[0035] The location information of the target vehicle can be determined through the navigation system carried by the target vehicle, and the location information of the interacting vehicle can be obtained through sensors or vehicle networking technology.

[0036] Sensors include but are not limited to lidar, millimeter wave radar, cameras and ultrasonic sensors.

[0037] Internet of Vehicles technology refers to the communication between the target vehicle and other vehicles or road infrastructure around it.

[0038] There are multiple path actions that the interacting vehicles may perform, including but not limited to turning left, turning right, and slowing down.

[0039] Since the above method provided by the embodiment of the invention focuses on planning the driving path of the target vehicle in complex scenarios, when the target vehicle is driving in a simple environment, such as a highway with non-merged sections or a road with clear traffic signal facilities, other path planning methods can be used for navigation, such as the Rapidly - Exploring Random Tree (RRT) path planning method.

[0040] In complex scenarios, the target vehicle dynamically updates the current position of the target vehicle and the observation points and interaction points corresponding to the current position in real time through the set planning cycle during driving. The selection of observation points and interaction points can be randomly sampled in the state space that determines the current position through sampling planning. The above state space can be a space composed of the position, posture and other states of the target vehicle.

[0041] The above method provided by the embodiment of the invention, facing complex application scenarios, first determines the traffic priority of the target vehicle and the interaction vehicle based on the interaction point model, and then performs local path planning when the target vehicle has priority, which can reduce the amount of calculation and the difficulty of parameter debugging; in the path planning, multiple planned paths will be generated and priority judgment will be performed, which can improve the safety and accuracy of navigation.

[0042] Preferably, in step S101, constructing an interaction point model includes:

[0043] Step S1011, calculating the interaction protection time based on the arrival time interval, wherein the arrival time interval is the difference between the time when the target vehicle and the interaction vehicle arrive at the interaction point.

[0044] Step S1012, modeling the interaction protection time as a function of relative speed and relative angle, and determining the boundary of the interaction protection time, wherein the relative speed is the difference between the speeds of the target vehicle and the interaction vehicle when they arrive at the interaction point, and the relative angle is the difference between the yaw angles of the target vehicle and the interaction vehicle when they arrive at the interaction point.

[0045] Step S1013: construct an interaction point model based on the boundary of the interaction protection time.

[0046] It is understandable that the relative speed , relative angle and the inter-arrival time is the interaction parameter of the interaction point.

[0047] ;

[0048] ;

[0049] ;

[0050] In the formula, is the speed of the target vehicle when it passes the jth interaction point, is the speed of the interactive vehicle when it passes the jth interactive point, such as Figure 2 As shown, is the yaw angle of the target vehicle when it passes the jth interaction point, is the yaw angle of the interactive vehicle when it passes the jth interactive point, is the time it takes for the target vehicle to travel from the observation point to the jth interaction point, is the time it takes for the interaction vehicle to travel from the position observed by the target vehicle to the jth interaction point.

[0051] It is understandable that the arrival time When it is zero, it means that the target vehicle and the interaction vehicle pass the same position at the same time, that is, the target vehicle and the interaction vehicle collide, which is not allowed. Therefore, in the process of path planning, Cannot be zero.

[0052] Preferably, in step S1011, calculating the interaction protection time based on the arrival time interval includes:

[0053] Interaction protection time Divided into and Two situations.

[0054] In the arrival time When it is negative, the target vehicle has priority to pass the interaction point, and the interaction protection time is the maximum value of the arrival time interval.

[0055] In the arrival time When it is positive, the interactive vehicle has priority to pass the interactive point, and the target vehicle needs to give way to the interactive vehicle. The interactive protection time is the minimum value of the arrival time interval.

[0056] ;

[0057] In the formula, is the set of possible values ​​for the time interval in the dataset.

[0058] Furthermore, in step S1012, the interaction protection time is modeled as a function of the relative speed and the relative angle, and the boundary of the interaction protection time is determined, including:

[0059] Interaction protection time Constructed with respect to relative speed and relative angle Function:

[0060] ;

[0061] ;

[0062] ;

[0063] ;

[0064] In the formula, The superscripts 1, 2, 3, 4 and The superscripts 1 and 2 indicate the number of derivatives; the superscript T indicates transposition; is the fitting coefficient, which is selected by technicians based on experience.

[0065] According to the above function, the fitting curve is obtained and , determine the boundary of the interaction protection time based on the fitting curve :

[0066] ;

[0067] ;

[0068] Preferably, in step S1013, based on the boundary of the interactive protection time Build an interaction point model, including:

[0069] Based on the target vehicle and the jth interaction point The Euler distance between , and the minimum value of the arrival time interval , determine the upper speed limit , where the upper speed limit The target vehicle drives to the observation point speed limit.

[0070] ;

[0071] In the formula, A value representing the Euler distance.

[0072] Boundary based on interactive protection time and speed limit Build an interaction point model.

[0073] It is understandable that if Figure 3 As shown in the figure, before observing the interactive vehicle in the blind spot, the target vehicle needs to prepare for the worst-case scenario in the future, that is, slow down and give way to the interactive vehicle. Before, it travels to the interaction point The remaining time Should be greater than .

[0074] You can do this by Set to 0 to quickly estimate the above formula , which means that there is no need to predict the speed of the target vehicle at the corresponding position of the blind spot.

[0075] Furthermore, based on the initial speed of the interacting vehicle , the average acceleration of the interactive vehicle arriving at the interaction point The maximum value of the Euler distance between the interactive vehicle and the interactive point, and the minimum arrival time of the interactive vehicle are calculated. ;

[0076] ;

[0077] In the formula, the superscript j indicates the corresponding j-th interaction point, that is, the j-th interaction point Take this as an example to illustrate; Represents interactive vehicles and interaction points The Euler distance between .

[0078] Based on the initial speed of the interactive vehicle, the minimum average acceleration of the interactive vehicle to the interaction point, and the Euler distance between the interactive vehicle and the interaction point, the maximum arrival time of the interactive vehicle is calculated. ;

[0079] ;

[0080] Based on the initial speed of the target vehicle , the average acceleration of the target vehicle reaching the interaction point The maximum value of the target vehicle and the Euler distance between the interaction point, and the minimum arrival time of the target vehicle are calculated. ;

[0081] ;

[0082] The maximum arrival time of the target vehicle is calculated based on the initial speed of the target vehicle, the minimum average acceleration of the target vehicle to the interaction point, and the Euler distance between the target vehicle and the interaction point. ;

[0083]

[0084] Based on the minimum arrival time of the target vehicle , the minimum arrival time of interactive vehicles , the maximum value of the arrival time interval , calculate the target vehicle’s overtaking capability ;

[0085] ;

[0086] Based on the maximum arrival time of the target vehicle, the maximum arrival time of the interactive vehicle, and the minimum value of the arrival time interval , calculate the yielding ability of the target vehicle ;

[0087] ;

[0088] Boundary based on interactive protection time , Speed ​​limit , Overtaking ability and ability to give way , build an interaction point model.

[0089] It is understandable that the constraints of the above interaction point model include but are not limited to the boundary of the interaction protection time. and speed limit The output of the above interaction point model includes but is not limited to overtaking capability. and ability to give way .

[0090] Understandably, the overtaking ability It is more important when judging the priority of traffic. When it is less than zero, the target vehicle can reach the interaction point before the interaction vehicle ,exist When it is greater than zero, the target vehicle will create safety risks and illegal risks if it wants to overtake. The time difference indicating that the target vehicle and the interacting vehicle decelerate at the same time can be used as a reference.

[0091] It is understandable that the minimum arrival time in the interaction point model adopts a constant acceleration model, which is invalid when the speed limit changes.

[0092] Exemplarily, in step S102, when the target vehicle has priority, a path point sequence including observation points and interaction points is constructed, including:

[0093] The initial path and speed limit of the target vehicle are input to the interactive path point generation algorithm to construct a maximum layer of The waypoint sequence ,in, is the number of the layer, .

[0094] The above interactive path point generation algorithm must follow: the selected path point sequence should reflect the speed limit changes between curves and straights, the path point sequence should involve the locations of observation points and interaction points, and the number of layers of the path point sequence should be of appropriate size, because too many layers will increase the amount of calculation, and too few layers will reduce the search space.

[0095] The above-mentioned interactive path point generation algorithm can be any one of the probabilistic algorithm (PRM), the rapidly expanding random tree algorithm (RRT), the A* algorithm, or other algorithms used by technicians based on experience.

[0096] Preferably, in step S102, the path point sequence is expanded based on the displacement-time image analysis method to obtain multiple planned paths, including:

[0097] Step S1021 , constructing a search space based on a displacement-time image (ST) analysis method.

[0098] Step S1022: Use the speed constraint as a spatiotemporal constraint in the search space to obtain a discrete network, wherein the speed constraint includes the curvature of the corresponding position of the path point sequence and the speed upper limit. , nodes in the discrete network are vectors associated with velocity, time, and displacement, and each pathpoint sequence includes multiple nodes. Velocity constraints from curvature for:

[0099] ;

[0100] In the formula, represents the maximum lateral acceleration of the target vehicle, express The curvature of the .

[0101] Step S1023, expanding the path point sequence in the discrete network to obtain multiple planned paths.

[0102] It is understandable that the search space constructed based on the displacement-time image (ST) analysis method does not generate a collision-free velocity profile, but rather a future trajectory distribution under the variational velocity limit, see Figure 4 Node , where the superscript is the node index, subscript is the layer index, and are speed and timestamp values.

[0103] Preferably, in step S1023, the path point sequence is expanded in the discrete network to obtain multiple planned paths, including:

[0104] Step S10231, discarding nodes in the discrete network that exceed a preset cost to obtain a target discrete network, wherein the preset cost includes a preset time limit or a preset speed limit, including:

[0105] The cost of node i is calculated by the following formula :

[0106] ;

[0107] ;

[0108] ;

[0109] ;

[0110] In the formula, represents the cost of the parent node of node i, represents the control cost, represents a set of discrete control inputs , Represents the speed of the target vehicle at node i With speed limit The deviation is the weight.

[0111] For example, =0.5, =0.5.

[0112] Calculate the cost of all nodes, sort all nodes in order of cost from small to large, filter out the top 50% of nodes, and obtain a set of nodes with lower costs.

[0113] The nodes in the above node set are expanded by the following formula:

[0114] ;

[0115] ;

[0116] In the formula, , and Respectively represent child nodes Corresponding speed and time, child nodes is a sequence of waypoints The node expanded from node i in represents the time corresponding to the target vehicle at node i, , Represents a sequence of waypoints The speed limit of each node in is Indicates the lower speed limit.

[0117] For example, is 0.

[0118] It is understandable that the preset cost includes but is not limited to layer limit, preset time limit, preset speed limit, and the technicians can set it according to actual requirements or experience.

[0119] During the expansion process, any layer limit is exceeded or time limit , or below the speed limit None of the child nodes will be expanded.

[0120] Step S10232, expanding the path point sequence in the target discrete network to obtain multiple planned paths.

[0121] Reference Figure 4 As shown, for a resolution of and Discrete network Among the nodes, only the nodes with lower costs are allowed to be expanded, because the values ​​of the discarded nodes are similar to the values ​​of the nodes with the lowest cost. This can significantly improve the search efficiency while reducing the impact on the search results, making it easier to plan path generation.

[0122] Preferably, in step S103, priority judgment is performed on multiple planned paths based on a multi-layer perceptron model to obtain a target path, including:

[0123] Step S1031, in order from the lowest layer node to the highest layer node, compare the priority probability output by the multilayer perceptron model with a preset constant threshold layer by layer to determine whether multiple planned paths are priority paths, wherein the multiple planned paths are determined based on a discrete network expanded according to a path point sequence, and the discrete network includes multiple layers of nodes.

[0124] It can be understood that the above-mentioned highest-level node refers to the highest-level node that the planned path itself passes through.

[0125] Based on binary classifier , build a 3-layer multi-layer perceptron model (Multi-Layer Perceptron, MLP for short):

[0126] ;

[0127] ;

[0128] In the formula, the eigenvector Corresponding to overtaking ability and ability to give way , is the time threshold, is a constant, among which, Is a positive number.

[0129] Reference Figure 5 As shown, taking the orange planning path as an example, the orange planning path passes through the path point sequence , in the order from the 1st layer node to the fth layer node, the priority probability output by the multilayer perceptron model is compared with the preset constant threshold layer by layer to determine whether the orange planned path is the priority path, where, .

[0130] Step S1032, when the priority probabilities of the nodes at each layer corresponding to the multiple planned paths are all greater than the preset constant threshold, the multiple planned paths are all used as priority paths, wherein the priority of the target vehicle traveling at the position of the nodes at each layer corresponding to the priority path is higher than that of the interactive vehicle.

[0131] Furthermore, when the priority probability of a target layer node corresponding to any one of the multiple planned paths is lower than a preset constant threshold, a safety judgment is made on the planned path to be expanded according to the judgment layer node, wherein the target layer node is any layer node among the multiple layers of nodes passed by the planned path, the judgment layer node is the previous layer node of the target layer node, and the planned path to be expanded is a planned path whose priority probability of the corresponding target layer node is lower than a preset constant threshold.

[0132] ;

[0133] In the formula, Represents the priority probability of the multilayer perceptron model output, represents the i-th node in the l-th layer, is a constant threshold, Indicates the expected interaction point collection of vehicles.

[0134] For example, is 0.5.

[0135] Safety judgment is made through the following formula:

[0136] ;

[0137] In the formula, is an invariant safety set, that is, there is at least one action that keeps the target vehicle safe within a preset time range.

[0138] For example, the above formula is explained by taking the l+1th layer as the target layer and the lth layer as the judgment layer as an example.

[0139] When the planned path to be expanded passes the safety judgment, the part of the planned path to be expanded after the judgment layer node is discarded, and the planned path to be expanded starts from the judgment layer node to obtain the expanded planned path.

[0140] Continuing with the above example, when the l-th layer node corresponding to the planned path to be expanded passes the security judgment, the part of the planned path to be expanded from the l-th layer node is discarded, and the planned path to be expanded is expanded from the l-th layer node to obtain the expanded planned path.

[0141] When the planned path to be expanded fails the safety judgment, the previous layer node of the judgment layer node is used as the new judgment layer node, and the planned path to be expanded is subjected to the safety judgment according to the new judgment layer node until the safety judgment passes.

[0142] Continuing with the above example, if the node at the first layer corresponding to the planned path to be expanded fails to pass the safety judgment, the node at the first layer-1 is used as the new node at the judgment layer for safety judgment. If the node at the first layer-1 corresponding to the planned path to be expanded also fails to pass the safety judgment, the node at the first layer-2 is used as the new node at the judgment layer for safety judgment.

[0143] It is foreseeable that the above safety judgment can be performed until the lowest level, that is, until it is determined that the planned path to be expanded can pass the safety judgment of a layer of nodes, or until the lowest level node corresponding to the planned path to be expanded also fails the safety judgment.

[0144] According to the order from the judgment layer nodes to the highest layer nodes that have passed the safety judgment, the priority probability of each layer node corresponding to the extended planning path is compared with the preset constant threshold layer by layer until the priority path is obtained.

[0145] It is understandable that obtaining the preferred path is an iterative process:

[0146] Continuing with the above example, when the l-th layer node corresponding to the planned path to be expanded passes the security judgment, the part of the planned path to be expanded from the l-th layer node is discarded, and the planned path to be expanded is expanded from the l-th layer node to obtain the extended planned path. For the sake of clarity, the above-mentioned extended planned path obtained by expanding from the l-th layer node is recorded as the first extended planned path.

[0147] When the priority probability of the first extended planning path corresponding to the first target layer node is lower than the preset constant threshold, the first extended planning path is safety judged according to the first judgment layer node, wherein the first target layer node is any layer node from the l+1th layer node to the highest layer node, and the first judgment layer node is the previous layer node of the first target layer node.

[0148] When the first extended planning path passes the safety judgment, the part of the first extended planning path after the first judgment layer node is discarded, and the first extended planning path is extended from the first judgment layer node to obtain the second extended planning path. It can be understood that the first extended planning path here is the planned path to be extended, and the second extended planning path is the corresponding extended planning path.

[0149] When the first extended planned path fails the safety judgment, the previous layer node of the first judgment layer node is used as a new judgment layer node, and the safety judgment of the first extended planned path is performed according to the new judgment layer node until the safety judgment passes.

[0150] According to the order from the judgment layer node to the highest layer node that has passed the safety judgment, the priority probability of each layer node corresponding to the second extended planning path is compared with the preset constant threshold layer by layer until the priority path is obtained.

[0151] Step S1033: taking the farthest priority path as the target path.

[0152] Please refer to Figure 5 As shown, the orange planned path has a low priority at the f-th layer node and can be safely judged and expanded at the f-1th layer node.

[0153] The purple planning path has high priority from the lowest node to the f*th node, and the gray planning path has high priority from the lowest node to the highest node, so both the purple planning path and the gray planning path are priority paths. Since the highest node of the gray planning path is between the f-1th node and the f*th node, the purple planning path is farther. Therefore, compared with the gray planning path, the purple planning path is the target path.

[0154] It is understandable that the above three planned paths correspond to different path point sequences, and the difference in the path point sequences is caused by the expansion. After the orange planned path is expanded, it can continue to be compared with the purple planned path.

[0155] Preferably, in step S104, navigating the target vehicle based on the target path includes:

[0156] The target path is smoothed based on the quintic segmented Bezier curve to obtain a smooth target path.

[0157] Navigate the target vehicle based on a smooth target path.

[0158] It can be understood that, after the above-mentioned smoothing process, the obtained smooth target path can be made more feasible in terms of kinematics and more adaptable to the requirements of practical applications.

[0159] Exemplarily, the present invention also provides test data of the above method to verify its technical effect, which is as follows:

[0160] A comparative test was conducted on the following four path planning methods for autonomous vehicles in a semi-enclosed factory area.

[0161] SSC method: a slowness-slope coherence algorithm generated using a constant velocity (CVel) model.

[0162] MCTS: Monte Carlo Tree Search algorithm.

[0163] MMFN-Expert: An improved expert model algorithm based on multimodal fusion network.

[0164] PD-IPM: The above method provided by the embodiment of the present invention.

[0165] Table 1 Test data

[0166]

[0167] It can be understood that the longer the completed distance, the fewer collisions, and the smaller the average speed fluctuation, the more accurate and safe the corresponding path planning method is.

[0168] In summary, compared with other methods, the PD-IPM method provided by the embodiment of the present invention is more accurate, safer and more stable.

[0169] Among them, MMFN-Expert has the longest completion distance in scenario 2, but MMFN-Expert has the highest number of collisions in both scenarios 1 and 2, which shows that MMFN-Expert is not safe enough. MCTS has the smallest average speed fluctuation in scenario 1, but the completion distance of MCTS in both scenarios 1 and 2 is relatively short, which shows that MCTS is not accurate enough.

[0170] The above-mentioned method PD-IPM provided by the embodiment of the present invention has the longest completion distance and the least number of collisions in scene 1, and has the smallest average speed fluctuation and the least number of collisions in scene 2, and the average speed amplitudes in scene 1 and scene 2 are both the smallest, indicating that PD-IPM is safe and accurate. Although it is relatively conservative, this is exactly what is needed for autonomous driving vehicles to travel in complex environments.

[0171] The present invention also provides a non-transitory machine-readable medium storing a computer program, wherein the computer program, when executed by a processor of a computer, is used to cause the computer to execute the method of the present invention.

[0172] The present invention also provides a computer program product, including a computer program, wherein the computer program, when executed by a processor of a computer, is used to cause the computer to execute the method of the present invention.

[0173] The present invention also provides an electronic device, comprising: at least one processor; and a memory connected to the at least one processor. The memory stores a computer program executable by the at least one processor, and the computer program is used to enable the electronic device to perform the method of the present invention when executed by the at least one processor.

[0174] refer to Figure 6, a block diagram of an electronic device that can be used as a server or client of an embodiment of the invention will now be described, which is an example of hardware devices that can be applied to various aspects of the invention. The electronic device is intended to represent various forms of digital electronic computer devices, such as laptop computers, desktop computers, workbenches, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the invention described and / or required herein.

[0175] like Figure 6 As shown, the electronic device includes a computing unit 601, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 602 or a computer program loaded from a storage unit 608 to a random access memory (RAM) 603. In the RAM 603, various programs and data required for the operation of the electronic device can also be stored. The computing unit 601, the ROM 602, and the RAM 603 are connected to each other via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0176] Multiple components in the electronic device are connected to the I / O interface 605, including: an input unit 606, an output unit 607, a storage unit 608, and a communication unit 609. The input unit 606 can be any type of device that can input information to the electronic device, and the input unit 606 can receive input digital or character information, and generate key signal input related to user settings and / or function control of the electronic device. The output unit 607 can be any type of device that can present information, and can include but is not limited to a display, a speaker, a video / audio output terminal, a vibrator, and / or a printer. The storage unit 608 can include but is not limited to a disk, an optical disk. The communication unit 609 allows the electronic device to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks, and can include but is not limited to a modem, a network card, an infrared communication device, and / or a wireless communication transceiver, such as a Bluetooth device, a WiFi device, a WiMax device, a cellular communication device, and / or the like.

[0177] The computing unit 601 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 601 include, but are not limited to, a CPU, a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing units, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 601 performs the various methods and processes described above. For example, in some embodiments, the method embodiments created by the present invention may be implemented as a computer program, which is tangibly contained in a machine-readable medium, such as a storage unit 608. In some embodiments, part or all of the computer program may be loaded and / or installed on an electronic device via a ROM 602 and / or a communication unit 609. In some embodiments, the computing unit 601 may be configured to perform the above-described method in any other appropriate manner (e.g., by means of firmware).

[0178] The computer program for implementing the method of the present invention can be written in any combination of one or more programming languages. These computer programs can be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the computer program, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The computer program can be executed entirely on the machine, partially on the machine, partially on the machine as a stand-alone software package and partially on a remote machine, or entirely on a remote machine or server.

[0179] In the context of the present invention, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or apparatus. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable signal medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, or infrared systems, devices, or equipment, or any suitable combination of the foregoing. More specific examples of machine-readable storage media may include electrical connections based on one or more lines, portable computer disks, hard disks, random access memories (RAM), read-only memories (ROM), erasable programmable read-only memories (EPROM or flash memory), optical fibers, portable compact disk read-only memories (CD-ROMs), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0180] It should be noted that the term "including" and its variations used in the embodiments of the present invention are open inclusions, that is, "including but not limited to". The term "based on" means "based at least in part on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one other embodiment"; the term "some embodiments" means "at least some embodiments". The modifications of "one" and "multiple" mentioned in the embodiments of the present invention are illustrative and not restrictive. Those skilled in the art should understand that unless otherwise clearly indicated in the context, they should be understood as "one or more". The descriptions of the terms "first", "second", etc. are for descriptive purposes only and cannot be understood as indicating or implying their relative importance or implicitly indicating the number of technical features indicated.

[0181] The user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in the embodiments of the present invention are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of relevant countries and regions, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0182] The various steps described in the method implementation methods provided by the embodiments of the present invention can be performed in different orders and / or in parallel. In addition, the method implementation methods may include additional steps and / or omit the steps shown. The scope of protection of the present invention is not limited in this respect.

[0183] The term "embodiment" in this specification refers to specific features, structures or characteristics described in conjunction with the embodiment that can be included in at least one embodiment of the invention. The appearance of this phrase in various places in the specification does not necessarily mean the same embodiment, nor does it mean that it is mutually exclusive with other embodiments and is independent or optional. The various embodiments in this specification are described in a related manner, and the same and similar parts between the various embodiments refer to each other. In particular, for the device, equipment, and system embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts refer to the partial description of the method embodiment.

[0184] The above-described embodiments only express several implementation methods of the present invention, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of protection. It should be pointed out that, for a person of ordinary skill in the art, several modifications and improvements can be made without departing from the concept of the present invention, and these all belong to the scope of protection of the present invention. Therefore, the scope of protection of the present invention shall be subject to the attached claims.

Claims

1. A local path planning method for an autonomous driving vehicle, characterized in that: include: Constructing an interaction point model, and determining the traffic priority between the target vehicle and the interaction vehicle based on the interaction point model; In the case where the target vehicle has priority, a path point sequence including observation points and interaction points is constructed, and the path point sequence is expanded based on a displacement-time image analysis method to obtain multiple planned paths, wherein the multiple planned paths include multiple planned paths of the target vehicle corresponding to multiple path actions that may be performed by the interaction vehicle; Prioritizing the multiple planned paths based on a multi-layer perceptron model to obtain a target path, wherein the output of the interaction point model is used as the input of the multi-layer perceptron model, and the path segment from the current position point of the target vehicle to the observation point in the multiple planned paths is a common path; The target vehicle is navigated based on the target path.

2. The method according to claim 1, characterized in that: Prioritizing the multiple planned paths based on a multi-layer perceptron model to obtain a target path includes: Comparing the priority probability output by the multilayer perceptron model with a preset constant threshold layer by layer in the order from the lowest layer node to the highest layer node, to determine whether the multiple planned paths are priority paths, wherein the multiple planned paths are determined according to a discrete network expanded from the path point sequence, and the discrete network includes multiple layers of nodes; In the case where the priority probabilities of the nodes at each layer corresponding to the multiple planned paths are all greater than the preset constant threshold, the multiple planned paths are all used as priority paths, wherein the priority of the target vehicle traveling at the position of the nodes at each layer corresponding to the priority path is higher than that of the interactive vehicle; The farthest priority path is used as the target path.

3. The method according to claim 2, characterized in that In the case that the priority probability of the target layer node corresponding to any one of the multiple planned paths is lower than the preset constant threshold, a safety judgment is made on the planned path to be extended according to the judgment layer node, wherein the target layer node is any layer node among the multiple layers of nodes passed by the planned path, the judgment layer node is the previous layer node of the target layer node, and the planned path to be extended is a planned path whose priority probability of the corresponding target layer node is lower than the preset constant threshold; If the planned path to be expanded passes the safety judgment, discard the part of the planned path to be expanded after the judgment layer node, and expand the planned path to be expanded from the judgment layer node to obtain an expanded planned path; If the planned path to be expanded fails to pass the safety judgment, the previous layer node of the judgment layer node is used as a new judgment layer node, and the planned path to be expanded is subjected to safety judgment according to the new judgment layer node until the safety judgment passes; According to the order from the judgment layer node to the highest layer node that has passed the safety judgment, the priority probability of each layer node corresponding to the extended planning path is compared with the preset constant threshold layer by layer until the priority path is obtained.

4. The method according to claim 1, characterized in that: Build an interaction point model, including: Calculating the interaction protection time based on the arrival time interval, wherein the arrival time interval is the difference between the time when the target vehicle and the interaction vehicle arrive at the interaction point; The interaction protection time is constructed as a function of relative speed and relative angle, and a boundary of the interaction protection time is determined according to the function, wherein the relative speed is the difference between the speeds of the target vehicle and the interaction vehicle when they arrive at the interaction point, and the relative angle is the difference between the yaw angles of the target vehicle and the interaction vehicle when they arrive at the interaction point; The interaction point model is constructed based on the boundary of the interaction protection time.

5. The method according to claim 4, characterized in that The interactive protection time is calculated based on the inter-arrival time interval, including: When the arrival time interval is negative, the target vehicle has priority to pass the interaction point, and the interaction protection time is the maximum value of the arrival time interval; When the arrival time interval is positive, the interactive vehicle has priority in passing the interactive point, and the interactive protection time is the minimum value of the arrival time interval.

6. The method according to claim 5, characterized in that Constructing the interaction point model based on the boundary of the interaction protection time includes: Determine a speed upper limit based on the Euler distance between the target vehicle and the interaction point and the minimum value of the arrival time interval, wherein the speed upper limit is the speed upper limit of the target vehicle traveling to the observation point; The interaction point model is constructed based on the boundary of the interaction protection time and the speed upper limit.

7. The method according to claim 6, characterized in that The interaction point model is constructed based on the boundary of the interaction protection time and the speed upper limit, including: Calculate the minimum arrival time and the maximum arrival time of the interactive vehicle based on the initial speed of the interactive vehicle, the Euler distance between the interactive vehicle and the interactive point, and the maximum and minimum values ​​of the average acceleration of the interactive vehicle when arriving at the interactive point; Calculate the minimum arrival time and the maximum arrival time of the target vehicle based on the initial speed of the target vehicle, the Euler distance between the target vehicle and the interaction point, and the maximum and minimum values ​​of the average acceleration of the target vehicle when reaching the interaction point; Calculating the overtaking capability of the target vehicle based on the minimum arrival time of the target vehicle, the minimum arrival time of the interacting vehicles, and the maximum value of the arrival time intervals; Calculating the yielding capability of the target vehicle based on the maximum arrival time of the target vehicle, the maximum arrival time of the interactive vehicle, and the minimum value of the arrival time interval; The interaction point model is constructed based on the boundary of the interaction protection time, the speed upper limit, the overtaking capability and the yielding capability.

8. The method according to claim 1, characterized in that The path point sequence is expanded based on the displacement-time image analysis method to obtain multiple planning paths, including: Constructing the search space based on the displacement-time image analysis method; Using the speed constraint as a spatiotemporal constraint in the search space to obtain a discrete network, wherein the speed constraint includes the curvature of the corresponding position of the path point sequence and the speed upper limit, the nodes in the discrete network are vectors related to speed, time and displacement, and each path point sequence includes a plurality of the nodes; The path point sequence is expanded in the discrete network to obtain multiple planned paths.

9. The method according to claim 8, characterized in that Expanding the path point sequence in the discrete network to obtain multiple planned paths includes: Discarding nodes in the discrete network that exceed a preset cost to obtain a target discrete network, wherein the preset cost includes a preset time limit or a preset speed limit; The path point sequence is expanded in the target discrete network to obtain multiple planned paths.

10. The method according to claim 1, characterized in that Navigating the target vehicle based on the target path includes: Smoothing the target path based on a quintic segmented Bezier curve to obtain a smooth target path; The target vehicle is navigated based on the smoothed target path.

11. An electronic device, comprising: A processor and a memory storing a program, wherein the program comprises instructions, which, when executed by the processor, cause the processor to perform the method according to any one of claims 1 to 10.

Images