Variable weight-based path planning method and device, vehicle, and storage medium

Abstract

The application relates to a variable weight-based path planning method and device, a vehicle and a storage medium, wherein the method comprises the following steps: determining a cruise termination state, a lane change termination state and a following termination state of a target vehicle; acquiring environment information in which the target vehicle is located, and generating optimal cruise weight parameters, optimal lane change weight parameters and optimal following weight parameters in combination with a weight learning model to calculate a cost function value; acquiring current state information of the target vehicle, and generating a plurality of to-be-planned paths in combination with the cruise termination state, the lane change termination state and the following termination state; and obtaining a final planning path corresponding to the target vehicle based on the plurality of to-be-planned paths, the cost function value and a preset motion constraint checking and collision detection strategy. The application introduces variable weight parameters in path planning, so that different weight coefficients can be selected in different scenarios to execute different path planning strategies, and the generalization performance of path planning is effectively improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of automatic driving, and in particular to a path planning method and device based on variable weights, a vehicle and a storage medium. BACKGROUND

[0002] An autonomous vehicle needs to perceive the surrounding environment, track and predict traffic participants, and complete the local path planning task of the ego vehicle under the condition of ensuring safety. A local path planning algorithm based on sampling usually calculates the cost value of each path according to a defined cost function within a set of drivable planning paths; collision detection and kinematic constraint detection are performed on each path from low to high according to the cost value; finally, the optimal path is selected, and the control module downstream follows the planning output of the optimal path to ensure that the ego vehicle travels along the planned path.

[0003] At present, the existing technology can select the optimal path through an evaluation function after sampling the planning path, wherein the evaluation function selects three items of the distance deviating from the reference path, the path smoothness, and the distance from the obstacle, and sets a weight coefficient for each item. However, the weight coefficient of the existing technology is a fixed value, which is set at the beginning, so that the generalization performance is poor when dealing with some long-tail scenarios. In addition, the existing technology can generate a plurality of predicted driving trajectories by using a method based on time sampling, speed sampling and angle sampling; then the vehicle maximum acceleration, vehicle maximum cornering speed, vehicle speed, distance of the vehicle from the road center line, vehicle driving distance and vehicle heading angle within the sampling time range are used as evaluation factors, different cost weight is set for each evaluation factor, and the optimal planning path is selected according to the calculated cost weight. However, the existing technology does not detail the setting process of the weight coefficient, and the optimal path selection made according to the fixed weight coefficient cannot guarantee that it is applicable in all scenarios.

[0004] In summary, the existing technology cannot dynamically adjust the weight coefficient according to the environmental state, which greatly affects the generalization and scenario applicability of path planning, and needs to be solved urgently. SUMMARY

[0005] The present application provides a path planning method and device based on variable weights, a vehicle and a storage medium to solve the problem that the existing technology cannot dynamically adjust the weight coefficient according to the environmental state, which greatly affects the generalization and scenario applicability of path planning.

[0006] The first aspect of the present application provides a path planning method based on variable weights, comprising the following steps: determining a cruise termination state, a lane change termination state and a following termination state of a target vehicle based on a preset path planning duration; obtaining environmental information of the target vehicle, and generating optimal cruise weight parameters, optimal lane change weight parameters and optimal following weight parameters corresponding to the cruise termination state, the lane change termination state and the following termination state according to the environmental information and a pre-trained weight learning model, to calculate a cost function value based on the optimal cruise weight parameters, the optimal lane change weight parameters and the optimal following weight parameters; obtaining current state information of the target vehicle, and generating a plurality of paths to be planned through the current state information, the cruise termination state, the lane change termination state and the following termination state, to obtain a final planning path corresponding to the target vehicle based on the plurality of paths to be planned, the cost function value and a preset motion constraint checking and collision detection strategy.

[0007] Optionally, in an embodiment of the present application, the cruise termination state, the lane change termination state and the following termination state of the target vehicle are determined based on the preset path planning duration, comprising: determining a path planning duration of the target vehicle, and obtaining a vehicle state of the target vehicle after the path planning duration; generating the cruise termination state, the lane change termination state and the following termination state based on the vehicle state, a preset cruise strategy, a preset lane change strategy and a preset following strategy.

[0008] Optionally, in an embodiment of the present application, before the optimal cruise weight parameters, the optimal lane change weight parameters and the optimal following weight parameters corresponding to the cruise termination state, the lane change termination state and the following termination state are generated according to the environmental information and the pre-trained weight learning model, it further comprises: constructing a weight learning model based on a preset BEV feature extraction network and a Transformer architecture; collecting historical environmental information of the target vehicle and a historical optimal path corresponding to the historical environmental information, inputting the historical environmental information into the weight learning model to obtain weight parameters corresponding to the historical environmental information; calculating a cost function value of the historical optimal path according to the weight parameters, and determining a loss function of the weight learning model based on the cost function value of the historical optimal path, to train the weight learning model through the loss function and a preset imitation learning strategy.

[0009] Optionally, in an embodiment of the present application, after the plurality of to-be-planned paths are generated based on the current state information, the cruise termination state, the lane change termination state and the following termination state, the method further comprises: sorting the plurality of to-be-planned paths in ascending order based on the cost function values of the plurality of to-be-planned paths, to obtain a to-be-planned path sequence corresponding to the plurality of to-be-planned paths.

[0010] Optionally, in an embodiment of the present application, the obtaining the final planned path corresponding to the target vehicle based on the plurality of to-be-planned paths, the cost function values and the preset motion constraint checking and collision detection strategy comprises: obtaining a to-be-planned path with the minimum cost function value in the to-be-planned path sequence corresponding to the plurality of to-be-planned paths; determining a motion constraint checking threshold of the target vehicle, and filtering the to-be-planned path based on the motion constraint checking threshold to obtain a candidate path corresponding to the to-be-planned path, wherein the motion constraint checking threshold comprises a minimum vehicle speed threshold, a maximum vehicle speed threshold, an acceleration absolute value threshold and a turning curvature threshold; predicting driving trajectories of other traffic participants in an environment in which the target vehicle is located to obtain predicted driving trajectories of the other traffic participants; determining whether the target vehicle has a collision risk according to the predicted driving trajectories and the candidate path corresponding to the target vehicle; if the target vehicle has the collision risk, extracting a new to-be-planned path with the minimum cost function value from the to-be-planned path sequence except the to-be-planned path, and re-executing the motion constraint checking and collision detection operation based on the preset motion constraint checking and collision detection strategy; if the target vehicle does not have the collision risk, taking the candidate path as the final planned path to control the target vehicle to perform a corresponding driving operation according to the final planned path.

[0011] The second aspect embodiment of the application provides a path planning device based on variable weights, comprising: a determination module configured to determine a cruise termination state, a lane change termination state and a following termination state of a target vehicle based on a preset path planning duration; a first calculation module configured to obtain environmental information of the target vehicle, and generate optimal cruise weight parameters, optimal lane change weight parameters and optimal following weight parameters corresponding to the cruise termination state, the lane change termination state and the following termination state according to the environmental information and a pre-trained weight learning model, so as to calculate a cost function value based on the optimal cruise weight parameters, the optimal lane change weight parameters and the optimal following weight parameters; a checking module configured to obtain current state information of the target vehicle, and generate a plurality of to-be-planned paths through the current state information, the cruise termination state, the lane change termination state and the following termination state, and check a preset motion constraint and a collision detection strategy based on the plurality of to-be-planned paths, the cost function value and the preset motion constraint and collision detection strategy, so as to obtain a final planned path corresponding to the target vehicle.

[0012] Optionally, in an embodiment of the application, the determination module comprises: a first acquisition unit configured to determine a path planning duration of the target vehicle, and acquire a vehicle state of the target vehicle after the path planning duration; and a generation unit configured to generate the cruise termination state, the lane change termination state and the following termination state based on the vehicle state, a preset cruise strategy, a preset lane change strategy and a preset following strategy.

[0013] Optionally, in an embodiment of the application, the device further comprises: a construction module configured to construct a weight learning model based on a preset BEV feature extraction network and a Transformer architecture before generating the optimal cruise weight parameters, the optimal lane change weight parameters and the optimal following weight parameters corresponding to the cruise termination state, the lane change termination state and the following termination state according to the environmental information and the pre-trained weight learning model; an acquisition module configured to acquire historical environmental information of the target vehicle and a historical optimal path corresponding to the historical environmental information, and input the historical environmental information into the weight learning model to obtain weight parameters corresponding to the historical environmental information; and a training module configured to calculate a cost function value of the historical optimal path according to the weight parameters, and determine a loss function of the weight learning model based on the cost function value of the historical optimal path, so as to train the weight learning model through the loss function and a preset imitation learning strategy.

[0014] Optionally, in an embodiment of the present application, the sorting module is further configured to sort the plurality of to-be-planned paths in ascending order based on a cost function value of each to-be-planned path in the plurality of to-be-planned paths to obtain a to-be-planned path sequence corresponding to the plurality of to-be-planned paths after the plurality of to-be-planned paths are generated based on the current state information, the cruise termination state, the lane change termination state, and the following termination state.

[0015] Optionally, in an embodiment of the present application, the checking module comprises: a second acquisition unit configured to acquire a to-be-planned path with a minimum cost function value in a to-be-planned path sequence corresponding to the plurality of to-be-planned paths; a filtering unit configured to determine a motion checking threshold of the target vehicle, and filter the to-be-planned path based on the motion checking threshold to obtain a candidate path corresponding to the to-be-planned path, wherein the motion checking threshold comprises a minimum vehicle speed threshold, a maximum vehicle speed threshold, an acceleration absolute value threshold, and a turning curvature threshold; a prediction unit configured to predict a driving trajectory of other traffic participants in an environment in which the target vehicle is located to obtain a predicted driving trajectory of the other traffic participants; a judgment unit configured to determine whether the target vehicle has a collision risk according to the predicted driving trajectory and the candidate path corresponding to the target vehicle; a loop unit configured to extract a new to-be-planned path with a minimum cost function value in the to-be-planned path sequence except the to-be-planned path if the target vehicle has the collision risk, and re-perform motion constraint checking and collision detection operations on the new to-be-planned path based on the preset motion constraint checking and collision detection strategy; and a control unit configured to take the candidate path as the final planned path to control the target vehicle to perform a corresponding driving operation according to the final planned path if the target vehicle does not have the collision risk.

[0016] An embodiment of the third aspect of the present application provides a vehicle, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the path planning method based on variable weights as described in the above embodiments.

[0017] An embodiment of the fourth aspect of the present application provides a computer readable storage medium, which stores a computer program executable by a processor to implement the path planning method based on variable weights as described above.

[0018] Therefore, the embodiments of the present application have the following beneficial effects:

[0019] Embodiments of the present application can determine the cruise termination state, lane change termination state and following termination state of the target vehicle based on a preset path planning duration; obtain environmental information in which the target vehicle is located, and generate optimal cruise weight parameters, optimal lane change weight parameters and optimal following weight parameters corresponding to the cruise termination state, lane change termination state and following termination state according to the environmental information and a pre-trained weight learning model, to calculate a cost function value based on the optimal cruise weight parameters, optimal lane change weight parameters and optimal following weight parameters; obtain current state information of the target vehicle, and generate a plurality of to-be-planned paths through the current state information, cruise termination state, lane change termination state and following termination state, based on the plurality of to-be-planned paths, cost function value and preset motion constraint checking and collision detection strategy, to obtain a final planning path corresponding to the target vehicle. The present application introduces variable weight parameters in path planning, so that different weight coefficients can be selected in different scenarios to execute different path planning strategies, effectively improving the generalization performance of path planning. Thus, the problem that the prior art cannot dynamically adjust the weight coefficients according to the environmental state, greatly affecting the generalization and scenario applicability of path planning, is solved.

[0020] Additional aspects and advantages of the present application will be made apparent by the following description and the accompanying drawings. BRIEF DESCRIPTION OF DRAWINGS

[0021] The above and / or additional aspects and advantages of the present application will become apparent and be readily understood from the following description, taken in conjunction with the accompanying drawings, in which:

[0022] Figure 1 A flowchart of a path planning method based on variable weights according to an embodiment of the present application is provided;

[0023] Figure 2 A structure diagram of a weight learning model according to an embodiment of the present application is provided;

[0024] Figure 3 An example diagram of a path planning device based on variable weights according to an embodiment of the present application is provided;

[0025] Figure 4 A structure diagram of a vehicle according to an embodiment of the present application is provided.

[0026] Among them, 10 - path planning device based on variable weights; 100 - determination module, 200 - first calculation module, 300 - checking module; 401 - memory, 402 - processor, 403 - communication interface. DETAILED DESCRIPTION

[0027] Embodiments of the present application are described below in detail, examples of which are shown in the accompanying drawings, wherein the same or similar notations represent the same or similar elements or elements having the same or similar functions throughout. The embodiments described below by reference to the drawings are exemplary and are intended to explain the present application, and cannot be understood as a limitation of the present application.

[0028] The variable weight-based path planning method, device, vehicle and storage medium of the embodiments of the present application are described below with reference to the accompanying drawings. In view of the problems mentioned in the above background art, the present application provides a variable weight-based path planning method, in which the cruise termination state, the lane change termination state and the following termination state of the target vehicle are determined based on a preset path planning duration; the environmental information in which the target vehicle is located is obtained, and the optimal cruise weight parameter, the optimal lane change weight parameter and the optimal following weight parameter corresponding to the cruise termination state, the lane change termination state and the following termination state are generated according to the environmental information and a pre-trained weight learning model, so as to calculate the cost function value based on the optimal cruise weight parameter, the optimal lane change weight parameter and the optimal following weight parameter; the current state information of the target vehicle is obtained, and a plurality of paths to be planned are generated through the current state information, the cruise termination state, the lane change termination state and the following termination state, and the final planning path corresponding to the target vehicle is obtained based on the plurality of paths to be planned, the cost function value and a preset motion constraint checking and collision detection strategy. The present application introduces variable weight parameters in path planning, so that different weight coefficients can be selected in different scenarios to execute different path planning strategies, thereby effectively improving the generalization performance of path planning. Thus, the problem that the prior art cannot dynamically adjust the weight coefficients according to the environmental state, which greatly affects the generalization and scenario applicability of path planning, is solved.

[0029] Specifically, Figure 1 A flowchart of a variable weight-based path planning method provided by an embodiment of the present application is shown in FIG. 1.

[0030] As Figure 1 shown, the variable weight-based path planning method includes the following steps:

[0031] In step S101, the cruise termination state, the lane change termination state and the following termination state of the target vehicle are determined based on a preset path planning duration.

[0032] The embodiments of the present application first generate a plurality of sets of termination state generation paths according to the cruise strategy, the following strategy and the lane change strategy according to the path planning duration T, and generate the self-vehicle state after the path planning duration T, thereby providing reliable data guidance and basis for the implementation of subsequent path planning.

[0033] Optionally, in an embodiment of the present application, based on the preset path planning duration, the cruise termination state, the lane change termination state and the following termination state of the target vehicle are determined, comprising: determining the path planning duration of the target vehicle, and obtaining the vehicle state of the target vehicle after the path planning duration; based on the vehicle state, the preset cruise strategy, the preset lane change strategy and the preset following strategy, the cruise termination state, the lane change termination state and the following termination state are generated.

[0034] In actual execution process, the embodiment of the present application can first determine the path planning duration T of the vehicle, and obtain the vehicle state (i.e. the ego state) of the target vehicle after the path planning duration T, so as to generate 20 groups of cruise termination state, lane change termination state and following termination state through the vehicle state, the cruise strategy, the lane change strategy and the following strategy; the termination state is a group of vectors, which represents the position and state of the ego vehicle in the Frenet coordinate system after T time, and the position and state in the Frenet coordinate system is a group of 6-dimensional vectors Wherein, S and L are the tangent direction and normal direction of the reference line in the Frenet coordinate system.

[0035] Specifically, the termination state of the cruise strategy can be expressed as Wherein, Nan is a non-zero default value, which is not used in subsequent calculation process; is the speed of the ego vehicle in the reference line direction, and L is the lateral deviation of the ego vehicle in the reference line direction; T is equally divided into n time periods, and n C state are generated, which respectively represent C state after T / n, 2T / n, 3T / n, …, T time.

[0036] The termination state of the lane change strategy can be expressed as S represents the displacement of the ego vehicle on the reference line, is the speed of the ego vehicle on the reference line, and L is the lateral deviation of the ego vehicle on the reference line, is the lateral speed of the ego vehicle on the reference line; T is equally divided into n time periods, and n LC state are generated, which respectively represent LC state after T / n, 2T / n, 3T / n, …, T time, wherein L can be set as the lateral deviation of the left lane center and the ego lane center, or the lateral deviation of the right lane center and the ego lane center, which respectively represent the ego vehicle state after left lane change and right lane change, so there are 2n LC state .

[0037] The termination state of the following strategy can be expressed as Wherein, S lead represents the displacement of the following target on the reference line relative to the ego vehicle, S safe is the safety distance of following, represents the speed of the following target on the reference line, L lead represents the lateral deviation of the following target on the reference line, represents the lateral speed of the following target on the reference line; T is divided into n time periods, and n F state , respectively, represent F after T / n, 2T / n, 3T / n, …, T state .

[0038] Therefore, the embodiments of the present application determine the cruise termination state, the lane change termination state and the following termination state of the target vehicle according to the path planning duration, thereby providing reliable data support for the calculation of the corresponding weight parameters.

[0039] In step S102, the environmental information in which the target vehicle is located is obtained, and the optimal cruise weight parameter, the optimal lane change weight parameter and the optimal following weight parameter corresponding to the cruise termination state, the lane change termination state and the following termination state are generated according to the environmental information and the pre-trained weight learning model, so as to calculate the cost function value based on the optimal cruise weight parameter, the optimal lane change weight parameter and the optimal following weight parameter.

[0040] After determining the path planning duration, the cruise termination state, the lane change termination state and the following termination state of the target vehicle, further, the embodiments of the present application can also select three evaluation factors, which are ΔL, Δt and Δv, wherein ΔL represents the absolute value of the difference between the L of the termination state and the L of the current state of the ego vehicle; Δt represents the time iT / n (i=1, 2, 3, …n) at which the termination state is located; and Δv is the absolute value of the difference between the speed of the termination state and the speed of the current state of the ego vehicle.

[0041] In the embodiments of the present application, the cost function value is:

[0042] cost = ω1*ΔL + ω2*Δt + ω3*Δv

[0043] Wherein, ω1, ω2, ω3 are variable weight parameters, which are learned by a deep learning neural network, and the specific learning steps will be described in detail later, which will not be repeated here; the three terms of the cost function represent different path planning strategies, the weight ω1 of the first term is a positive weight, and the smaller the first term, the more conservative the path planning strategy, which will not change lanes rashly; the second term ω2 is a negative weight, and the smaller the second term, the more the path planning strategy pays attention to the paths with larger Δt values; the weight ω3 of the first term is a positive value, and the smaller the third term, the path planning strategy will not rashly accelerate or decelerate.

[0044] Therefore, the embodiments of the present application calculate the cost function value of the target vehicle, thereby effectively guaranteeing the efficiency and reliability of the path planning.

[0045] Optionally, in one embodiment of this application, before generating the optimal cruise weight parameters, optimal lane change weight parameters, and optimal following weight parameters corresponding to the cruise termination state, lane change termination state, and following termination state based on environmental information and a pre-trained weight learning model, the method further includes: constructing a weight learning model based on a preset BEV feature extraction network and Transformer architecture; collecting historical environmental information of the target vehicle and the historical optimal path corresponding to the historical environmental information, inputting the historical environmental information into the weight learning model to obtain the weight parameters corresponding to the historical environmental information; calculating the cost function value of the historical optimal path based on the weight parameters, and determining the loss function of the weight learning model based on the cost function value of the historical optimal path, so as to train the weight learning model through the loss function and a preset imitation learning strategy.

[0046] It should be noted that the embodiments of this application select different weight parameters ω1, ω2, and ω3 to learn according to the different environments in which the vehicle is located, which will improve the generalization performance of the system.

[0047] Specifically, embodiments of this application first require constructing a weight learning model using a BEV feature extraction network and a Transformer architecture, such as... Figure 2 As shown, the backbone of this weight learning model is a BEV feature extraction network, which extracts BEV features from multi-view image data of the vehicle. The extracted BEV features, along with BEV features from historical frames, are input into a Transformer-based neural network. This weight learning model can learn the optimal weight parameters from the BEV features based on an attention mechanism. The result processed by the Transformer is then passed through an MLP neural network, outputting the weight parameters ω1, ω2, and ω3.

[0048] The embodiments of this application can use the imitation learning paradigm to train the neural network, that is, to collect data on human drivers driving vehicles and train the neural network.

[0049] Specifically, given the environmental perception conditions, the human driver selects path A as the optimal path, and calculates the cost value of path A as cost based on the pre-trained weight learning model outputting weight parameters ω1, ω2, and ω3. A Then the loss function for network training is set as Loss = (cost) A -in cost ) 2 And update the parameters of the neural network using backpropagation, where min cost This represents the lower limit of the cost range; the purpose of the loss function is to enable the neural network to adjust its parameters during learning, mimicking the experience of human drivers, and using the planned path chosen by the human driver as the planned path with the lowest cost value.

[0050] After the training weight learning model is completed, the embodiment of the application can calculate the corresponding results based on the BEV feature extraction according to the Q, K and V matrices trained under the Transformer framework under the given environment perception condition, input them to the MLP network to obtain the optimal ω1, ω2 and ω3 weight parameters, and calculate the cost function according to the weight parameters, so as to select the optimal planning path.

[0051] Therefore, in the embodiment of the application, the weight coefficients of each item can be dynamically adjusted according to the environmental state when setting the cost function, so that different weight coefficients can be used for different environmental scenarios to perform different path planning strategies, thereby effectively improving the generalization performance of path planning.

[0052] In step S103, the current state information of the target vehicle is obtained, and a plurality of to-be-planned paths are generated through the current state information, the cruise termination state, the lane change termination state and the following termination state. Based on the plurality of to-be-planned paths, the cost function value and the preset motion constraint checking and collision detection strategy, the final planning path corresponding to the target vehicle is obtained.

[0053] Further, the embodiment of the application also needs to obtain the current state information of the vehicle, and generate a plurality of to-be-planned paths through the cruise termination state, the lane change termination state and the following termination state. According to the cost function value, the motion constraint checking and collision detection operation is performed on the plurality of to-be-planned paths, so as to obtain the final planning path corresponding to the target vehicle.

[0054] Optionally, in an embodiment of the application, after generating a plurality of to-be-planned paths through the current state information, the cruise termination state, the lane change termination state and the following termination state, it further includes: based on the cost function value of each to-be-planned path in the plurality of to-be-planned paths, the plurality of to-be-planned paths are sorted in ascending order to obtain a to-be-planned path sequence corresponding to the plurality of to-be-planned paths.

[0055] It should be noted that in the embodiment of the application, the current state of the ego vehicle and the 4n termination states can be smoothly connected in space to generate 4n planned paths, and sorted in order from small to large according to the cost function value, thereby obtaining a to-be-planned path sequence corresponding to the plurality of to-be-planned paths. In the embodiment of the application, the path with a smaller cost function value will be preferentially selected, thereby improving the efficiency of path planning.

[0056] Optionally, in an embodiment of the present application, based on the plurality of to-be-planned paths, the cost function value, and the preset kinematic constraint checking and collision detection strategy, a final planned path corresponding to the target vehicle is obtained, including: obtaining a to-be-planned path with the minimum cost function value in a to-be-planned path sequence corresponding to the plurality of to-be-planned paths; determining a kinematic constraint checking threshold of the target vehicle, and filtering the to-be-planned path through the kinematic constraint checking threshold to obtain a candidate path corresponding to the to-be-planned path, wherein the kinematic constraint checking threshold includes a minimum vehicle speed threshold, a maximum vehicle speed threshold, an acceleration absolute value threshold, and a turning curvature threshold; predicting a driving trajectory of other traffic participants in an environment in which the target vehicle is located to obtain a predicted driving trajectory of the other traffic participants; judging whether the target vehicle has a collision risk according to the predicted driving trajectory and the candidate path corresponding to the target vehicle; if the target vehicle has a collision risk, extracting a new to-be-planned path with the minimum cost function value in the to-be-planned path sequence except the to-be-planned path, and re-executing the kinematic constraint checking and collision detection operation on the new to-be-planned path based on the preset kinematic constraint checking and collision detection strategy; if the target vehicle does not have a collision risk, taking the candidate path as the final planned path to control the target vehicle to perform a corresponding driving operation according to the final planned path.

[0057] In actual execution, the embodiments of the present application also need to extract a to-be-planned path with the minimum cost function value in a to-be-planned path sequence corresponding to a plurality of to-be-planned paths, and determine a kinematic constraint checking threshold of the target vehicle, while checking the selected to-be-planned path with the minimum cost function value according to the kinematic constraint, including the minimum vehicle speed, the maximum vehicle speed, the acceleration absolute value, and the turning curvature, which are filtered according to the set kinematic constraint threshold to maintain the comfort of the whole vehicle, to screen out a candidate path meeting the requirements.

[0058] Then, the embodiments of the present application need to predict the driving trajectory of other traffic participants to obtain a predicted driving trajectory of the other traffic participants, and judge whether there is a collision risk with other traffic participants in time and space according to the selected candidate path and the predicted driving trajectory, if there is no collision risk, taking the candidate path as the final planned path to control the target vehicle to perform a corresponding driving operation according to the final planned path. Otherwise, extract a new to-be-planned path with the minimum cost function value in the to-be-planned path sequence except the to-be-planned path, and re-execute the kinematic constraint checking and collision detection operation on the new to-be-planned path, thereby ensuring the reliability and accuracy of the planned path.

[0059] According to the variable weight-based path planning method provided in the embodiments of the present application, the cruise termination state, the lane change termination state and the following termination state of the target vehicle are determined based on a preset path planning duration; the environment information in which the target vehicle is located is acquired, and the optimal cruise weight parameter, the optimal lane change weight parameter and the optimal following weight parameter corresponding to the cruise termination state, the lane change termination state and the following termination state are generated according to the environment information and a pre-trained weight learning model, so as to calculate the cost function value based on the optimal cruise weight parameter, the optimal lane change weight parameter and the optimal following weight parameter; the current state information of the target vehicle is acquired, and a plurality of to-be-planned paths are generated through the current state information, the cruise termination state, the lane change termination state and the following termination state, so as to obtain the final planning path corresponding to the target vehicle based on the plurality of to-be-planned paths, the cost function value and a preset motion constraint checking and collision detection strategy. The variable weight parameter is introduced in the path planning, so that different weight coefficients can be selected in different scenarios to perform different path planning strategies, and the generalization performance of the path planning is effectively improved.

[0060] Secondly, the variable weight-based path planning device provided in the embodiments of the present application is described with reference to the accompanying drawings.

[0061] Figure 3 is a block schematic diagram of the variable weight-based path planning device in the embodiments of the present application.

[0062] As shown in Figure 3 , the variable weight-based path planning device 10 comprises a determination module 100, a first calculation module 200 and a checking module 300.

[0063] The determination module 100 is configured to determine the cruise termination state, the lane change termination state and the following termination state of the target vehicle based on a preset path planning duration.

[0064] The first calculation module 200 is configured to acquire the environment information in which the target vehicle is located, and generate the optimal cruise weight parameter, the optimal lane change weight parameter and the optimal following weight parameter corresponding to the cruise termination state, the lane change termination state and the following termination state according to the environment information and a pre-trained weight learning model, so as to calculate the cost function value based on the optimal cruise weight parameter, the optimal lane change weight parameter and the optimal following weight parameter.

[0065] The checking module 300 is configured to acquire the current state information of the target vehicle, and generate a plurality of to-be-planned paths through the current state information, the cruise termination state, the lane change termination state and the following termination state, so as to obtain the final planning path corresponding to the target vehicle based on the plurality of to-be-planned paths, the cost function value and a preset motion constraint checking and collision detection strategy.

[0066] Optionally, in an embodiment of the present application, the determining module 100 comprises a first acquisition unit and a generating unit.

[0067] The first acquisition unit is configured to determine a path planning duration of the target vehicle and acquire a vehicle state of the target vehicle after the path planning duration.

[0068] The generating unit is configured to generate a cruise termination state, a lane change termination state and a following termination state based on the vehicle state, a preset cruise strategy, a preset lane change strategy and a preset following strategy.

[0069] Optionally, in an embodiment of the present application, the variable weight-based path planning device 10 of the present application further comprises a constructing module, an acquiring module and a training module.

[0070] The constructing module is configured to, before generating the optimal cruise weight parameter, the optimal lane change weight parameter and the optimal following weight parameter corresponding to the cruise termination state, the lane change termination state and the following termination state according to the environment information and the pre-trained weight learning model, construct the weight learning model based on a preset BEV feature extraction network and a Transformer architecture.

[0071] The acquiring module is configured to acquire historical environment information of the target vehicle and a historical optimal path corresponding to the historical environment information, input the historical environment information into the weight learning model, and obtain a weight parameter corresponding to the historical environment information.

[0072] The training module is configured to calculate a cost function value of the historical optimal path according to the weight parameter, determine a loss function of the weight learning model based on the cost function value of the historical optimal path, and train the weight learning model through the loss function and a preset imitation learning strategy.

[0073] Optionally, in an embodiment of the present application, the variable weight-based path planning device 10 of the present application further comprises a sorting module.

[0074] The sorting module is configured to, after generating a plurality of to-be-planned paths through the current state information, the cruise termination state, the lane change termination state and the following termination state, sort the plurality of to-be-planned paths in ascending order based on a cost function value of each to-be-planned path in the plurality of to-be-planned paths, and obtain a to-be-planned path sequence corresponding to the plurality of to-be-planned paths.

[0075] Optionally, in an embodiment of the present application, the checking module 300 comprises a second acquisition unit, a filtering unit, a prediction unit, a judgment unit, a loop unit and a control unit.

[0076] The second acquisition unit is configured to acquire a to-be-planned path with a minimum cost function value in the to-be-planned path sequence corresponding to the plurality of to-be-planned paths.

[0077] a filtering unit configured to determine a motion school threshold of the target vehicle, and filter the to-be-planned path by the motion school threshold to obtain a candidate path corresponding to the to-be-planned path, wherein the motion school threshold comprises a minimum vehicle speed threshold, a maximum vehicle speed threshold, an acceleration absolute value threshold, and a turning curvature threshold.

[0078] a prediction unit configured to predict a driving trajectory of another traffic participant in an environment in which the target vehicle is located to obtain a predicted driving trajectory of the another traffic participant.

[0079] a judgment unit configured to determine whether the target vehicle has a collision risk according to the predicted driving trajectory and the candidate path corresponding to the target vehicle.

[0080] a circulation unit configured to extract a new to-be-planned path with a minimum cost function value from the to-be-planned path sequence if the target vehicle has the collision risk, and re-perform the motion constraint checking and the collision detection operation on the new to-be-planned path based on the preset motion constraint checking and collision detection strategy.

[0081] a control unit configured to take the candidate path as a final planned path to control the target vehicle to perform a corresponding driving operation according to the final planned path if the target vehicle does not have the collision risk.

[0082] It should be noted that the foregoing explanation and description of the embodiment of the path planning method based on variable weights also applies to the embodiment of the path planning device based on variable weights, which will not be described here again.

[0083] The path planning device based on variable weights provided by the embodiment of the present application comprises a determination module configured to determine a cruise termination state, a lane change termination state, and a following termination state of a target vehicle based on a preset path planning duration; a first calculation module configured to obtain environmental information in which the target vehicle is located, and generate optimal cruise weight parameters, optimal lane change weight parameters, and optimal following weight parameters corresponding to the cruise termination state, the lane change termination state, and the following termination state according to the environmental information and a pre-trained weight learning model, to calculate a cost function value based on the optimal cruise weight parameters, the optimal lane change weight parameters, and the optimal following weight parameters; and a checking module configured to obtain current state information of the target vehicle, and generate a plurality of to-be-planned paths through the current state information, the cruise termination state, the lane change termination state, and the following termination state, to obtain a final planned path corresponding to the target vehicle based on the plurality of to-be-planned paths, the cost function value, and a preset motion constraint checking and collision detection strategy. The present application introduces variable weight parameters in path planning, so that different weight coefficients can be selected in different scenarios to perform different path planning strategies, effectively improving the generalization performance of path planning.

[0084] Figure 4 A structural schematic diagram of a vehicle is provided for the embodiments of the present application. The vehicle can include:

[0085] The memory 401, the processor 402 and the computer program stored in the memory 401 and executable on the processor 402.

[0086] The processor 402 implements the variable weight based path planning method provided in the above embodiments when executing the program.

[0087] Further, the vehicle further includes:

[0088] The communication interface 403 is used for communication between the memory 401 and the processor 402.

[0089] The memory 401 is used to store the computer program executable on the processor 402.

[0090] The memory 401 can include a high-speed RAM memory, and can also include a non-volatile memory, for example, at least one disk memory.

[0091] If the memory 401, the processor 402 and the communication interface 403 are independently implemented, the communication interface 403, the memory 401 and the processor 402 can be connected to each other through a bus and complete communication between each other. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For the convenience of representation, Figure 4 In the figure, only one thick line is used to represent, but it does not mean that there is only one bus or one type of bus.

[0092] Optionally, in specific implementation, if the memory 401, the processor 402 and the communication interface 403 are integrated on a chip, the memory 401, the processor 402 and the communication interface 403 can complete communication between each other through an internal interface.

[0093] The processor 402 can be a central processing unit (CPU), or an application specific integrated circuit (ASIC), or one or more integrated circuits configured to perform the operations of the embodiments of the application.

[0094] The embodiments of the application further provide a computer readable storage medium, which stores a computer program. The computer program is executed by a processor to implement the variable weight based path planning method.

[0095] In the description of the specification, the description of the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the application. In the specification, the illustrative description of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any appropriate manner in any one or N embodiments or examples. In addition, the person skilled in the art can combine and combine the different embodiments or examples described in the specification and the features of the different embodiments or examples without contradiction.

[0096] In addition, the terms "first", "second" are only for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined with "first", "second" can explicitly or implicitly include at least one of the features. In the description of the application, the meaning of "N" is at least two, for example, two, three, etc., unless otherwise specifically limited.

[0097] Any process or method descriptions in flow charts or otherwise described herein can be understood as representing code modules, segments, or portions of code that include one or more executable instructions for performing a step of a process described in the specification, and the various embodiments of the preferred implementation of the application can include additional or fewer steps performing the described functions in the illustrated or discussed order, including as an embodiment of the application, the functions can be performed in an order different from that described, and / or can be performed substantially concurrently with each other, as will be appreciated by those skilled in the art.

[0098] The logic and / or steps represented in the flowcharts and / or described herein, for example, can be considered as a sequence of instructions to implement logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, apparatus, or device, such as a computer-based system, processor- based system, or other system that can fetch the instructions from the instruction execution system, apparatus, or device and execute the instructions. For purposes of this specification, a "computer-readable medium" can be any apparatus that can contain, store, communicate, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus, or device. The computer-readable medium can be a computer- readable storage medium or a computer-readable signal medium. The computer- readable storage medium can be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples (a non-exhaustive list) of the computer-readable storage medium include the following: an electrical connection having one or more wires (electrical connections), a portable computer diskette (a magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, and a portable compact disc read-only memory (CDROM). In addition, the computer-readable medium can even be paper or another suitable medium upon which the program is printed, as the program can be electronically captured, via, for example, optical scanning of the paper or other medium, then compiled, interpreted, or otherwise processed in a suitable manner, if necessary, and then stored in a computer memory.

[0099] It should be understood that aspects of the application can be implemented in hardware, software, firmware or combinations thereof. In the above embodiments, the N steps or methods can be implemented in software or firmware stored in a memory and executed by a suitable instruction execution system. If implemented in hardware and in another embodiment, any of the following technologies, known in the art, or their combinations can be used: discrete logic circuitry having logic gates for implementing logic functions on data signals, application specific integrated circuits having appropriate combinational logic gates, programmable gate arrays (PGA), field programmable gate arrays (FPGA), etc.

[0100] Those skilled in the art can understand that all or part of the steps carried out by the above-mentioned embodiment methods can be completed by programs instructing related hardware, and the programs can be stored in a computer-readable storage medium. When the programs are executed, one or a combination of the steps of the method embodiments is included.

[0101] In addition, each of the functional units in the various embodiments of the present application can be integrated in one processing module, or each of the units can be physically present separately, or two or more units can be integrated in one module. The integrated module can be implemented in the form of hardware or in the form of a software functional module. When the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it can also be stored in a computer readable storage medium.

[0102] The storage medium mentioned above can be a read-only memory, a magnetic disk or an optical disk, etc. Although the embodiments of the present application have been shown and described above, it should be understood that the above embodiments are exemplary and should not be construed as limiting the present application, and those skilled in the art can make changes, modifications, replacements and variations to the above embodiments within the scope of the present application.

Claims

1. A variable weight based path planning method, characterized in that, The method comprises the following steps: determining a cruise termination state, a lane change termination state and a following termination state of a target vehicle based on a preset path planning duration; obtaining environmental information of the target vehicle, and generating optimal cruise weight parameters, optimal lane change weight parameters and optimal following weight parameters corresponding to the cruise termination state, the lane change termination state and the following termination state according to the environmental information and a pre-trained weight learning model to calculate a cost function value based on the optimal cruise weight parameters, the optimal lane change weight parameters and the optimal following weight parameters; obtaining current state information of the target vehicle, and generating a plurality of to-be-planned paths through the current state information, the cruise termination state, the lane change termination state and the following termination state, and obtaining a final planning path corresponding to the target vehicle based on the plurality of to-be-planned paths, the cost function value and a preset motion constraint checking and collision detection strategy; wherein, before generating the optimal cruise weight parameters, the optimal lane change weight parameters and the optimal following weight parameters corresponding to the cruise termination state, the lane change termination state and the following termination state according to the environmental information and the pre-trained weight learning model, the method further comprises: constructing a weight learning model based on a preset BEV feature extraction network and a Transformer architecture; collecting historical environmental information of the target vehicle and a historical optimal path corresponding to the historical environmental information, inputting the historical environmental information into the weight learning model to obtain weight parameters corresponding to the historical environmental information; calculating a cost function value of the historical optimal path according to the weight parameters, and determining a loss function of the weight learning model based on the cost function value of the historical optimal path to train the weight learning model through the loss function and a preset imitation learning strategy.

2. The method of claim 1, wherein, The method of determining a cruise termination state, a lane change termination state and a following termination state of a target vehicle based on a preset path planning duration comprises: determining a path planning duration of the target vehicle, and obtaining a vehicle state of the target vehicle after the path planning duration; generating the cruise termination state, the lane change termination state and the following termination state based on the vehicle state, a preset cruise strategy, a preset lane change strategy and a preset following strategy.

3. The method of claim 1, wherein, After generating the plurality of to-be-planned paths through the current state information, the cruise termination state, the lane change termination state and the following termination state, the method further comprises: sorting the plurality of to-be-planned paths in ascending order based on a cost function value of each to-be-planned path in the plurality of to-be-planned paths to obtain a to-be-planned path sequence corresponding to the plurality of to-be-planned paths.

4. The method of claim 3, wherein, The method of obtaining a final planning path corresponding to the target vehicle based on the plurality of to-be-planned paths, the cost function value and a preset motion constraint checking and collision detection strategy comprises: obtaining a to-be-planned path with a minimum cost function value in a to-be-planned path sequence corresponding to the plurality of to-be-planned paths; determine a motion school threshold of the target vehicle, and filter the to-be-planned path through the motion school threshold to obtain a candidate path corresponding to the to-be-planned path, wherein the motion school threshold comprises a minimum vehicle speed threshold, a maximum vehicle speed threshold, an acceleration absolute value threshold, and a turning curvature threshold; predict driving trajectories of other traffic participants in an environment in which the target vehicle is located to obtain predicted driving trajectories of the other traffic participants; determine whether the target vehicle has a collision risk according to the predicted driving trajectories and the candidate path corresponding to the target vehicle; if the target vehicle has the collision risk, extract a new to-be-planned path with a minimum cost function value from the to-be-planned path sequence other than the to-be-planned path, and re-perform motion constraint checking and collision detection operations on the new to-be-planned path based on the preset motion constraint checking and collision detection strategy; if the target vehicle does not have the collision risk, take the candidate path as the final planned path to control the target vehicle to perform corresponding driving operations according to the final planned path.

5. A variable weight-based path planning device, characterized by, comprise: a determination module configured to determine a cruise termination state, a lane change termination state, and a following termination state of a target vehicle based on a preset path planning duration; a first calculation module configured to obtain environmental information in which the target vehicle is located, and generate optimal cruise weight parameters, optimal lane change weight parameters, and optimal following weight parameters corresponding to the cruise termination state, the lane change termination state, and the following termination state according to the environmental information and a pre-trained weight learning model, to calculate a cost function value based on the optimal cruise weight parameters, the optimal lane change weight parameters, and the optimal following weight parameters; a checking module configured to obtain current state information of the target vehicle, and generate a plurality of to-be-planned paths through the current state information, the cruise termination state, the lane change termination state, and the following termination state, to obtain a final planned path corresponding to the target vehicle based on the plurality of to-be-planned paths, the cost function value, and a preset motion constraint checking and collision detection strategy; wherein the path planning device based on variable weights further comprises: a construction module configured to, before generating the optimal cruise weight parameters, the optimal lane change weight parameters, and the optimal following weight parameters corresponding to the cruise termination state, the lane change termination state, and the following termination state according to the environmental information and the pre-trained weight learning model, construct a weight learning model based on a preset BEV feature extraction network and a Transformer architecture; an acquisition module configured to acquire historical environmental information of the target vehicle and a historical optimal path corresponding to the historical environmental information, and input the historical environmental information into the weight learning model to obtain weight parameters corresponding to the historical environmental information; The training module is configured to calculate a cost function value of the historical optimal path according to the weight parameter, and determine a loss function of the weight learning model based on the cost function value of the historical optimal path, so as to train the weight learning model through the loss function and a preset imitation learning strategy.

6. The apparatus of claim 5, wherein, Further comprising: The sorting module is configured to sort the plurality of to-be-planned paths in ascending order based on the cost function value of each to-be-planned path in the plurality of to-be-planned paths, to obtain a to-be-planned path sequence corresponding to the plurality of to-be-planned paths.

7. The apparatus of claim 5, wherein, The checking module comprises: The second acquisition unit is configured to acquire a to-be-planned path with the minimum cost function value in the to-be-planned path sequence corresponding to the plurality of to-be-planned paths; The filtering unit is configured to determine a motion checking threshold of the target vehicle, and filter the to-be-planned path through the motion checking threshold to obtain a candidate path corresponding to the to-be-planned path, wherein the motion checking threshold comprises a minimum vehicle speed threshold, a maximum vehicle speed threshold, an acceleration absolute value threshold and a turning curvature threshold; The prediction unit is configured to predict a driving trajectory of other traffic participants in an environment in which the target vehicle is located, to obtain a predicted driving trajectory of the other traffic participants; The judgment unit is configured to determine whether the target vehicle has a collision risk according to the predicted driving trajectory and the candidate path corresponding to the target vehicle; The cycle unit is configured to extract a new to-be-planned path with the minimum cost function value in the to-be-planned path sequence except the to-be-planned path, and re-perform motion checking and collision detection operations on the new to-be-planned path based on the preset motion checking and collision detection strategy, if the target vehicle has the collision risk. The control unit is configured to take the candidate path as the final planning path to control the target vehicle to perform a corresponding driving operation according to the final planning path, if the target vehicle does not have the collision risk.

8. A vehicle characterized by comprising: Comprise: A memory, a processor and a computer program stored on the memory and executable on the processor, the processor executes the program to implement the variable weight-based path planning method according to any one of claims 1-4.

9. A computer readable storage medium having stored thereon a computer program, characterized in that, The program is executed by the processor to implement the variable weight-based path planning method according to any one of claims 1-4.

Images