Automatic driving decision planning method, electronic equipment and readable storage medium

Abstract

The invention provides an automatic driving decision planning method, electronic equipment and a readable storage medium, and the method comprises the steps: carrying out the path planning of an automatic driving vehicle, and obtaining a target path; projecting the multi-modal prediction trajectory of the dynamic obstacle to the target path in the process that the autonomous vehicle travels along the target path to obtain a multi-scene projection drawing in time and space dimensions; a tree-shaped velocity curve is searched on the multi-scene projection drawing, a tree-shaped velocity curve set is obtained, a target tree-shaped velocity curve is determined in the tree-shaped velocity curve set, and the target tree-shaped velocity curve is composed of multiple tracks behind a bifurcation point and a common track before the bifurcation point; the plurality of trajectories are used for respectively describing speed plans of different scenes which are predicted to occur, and the public trajectory is used for describing speed plans determined according to occurrence probabilities of different scenes; and integrating the target path and the target tree-shaped speed curve to obtain a target tree-shaped trajectory, thereby solving the problem of unreasonable automatic driving decision and planning in related technologies.

Description

Technical Field

[0001] The present application relates to the field of autonomous driving, and in particular to an autonomous driving decision-making planning method, an electronic device, and a readable storage medium. Background Art

[0002] The autonomous driving system needs to understand the intentions of other traffic participants in the environment and make decisions and plans for the vehicle's behavior, ultimately generating a safe, comfortable, and efficient trajectory. In real driving scenarios, there are many uncertain events, such as vehicles in the next lane cutting into the vehicle's lane, ghosts poking their heads out of the blocked area, etc.

[0003] However, the relevant technology cannot provide accurate predictions of future risk events, resulting in irrational decision-making and planning. Regarding this technical problem, the relevant technology has not yet proposed an effective solution. Summary of the invention

[0004] The embodiments of the present application provide an autonomous driving decision-making planning method, an electronic device, and a readable storage medium to solve one or more of the above-mentioned technical problems.

[0005] In a first aspect, an embodiment of the present application provides an autonomous driving decision-making planning method, comprising: performing path planning for an autonomous driving vehicle to obtain a target path; projecting the multimodal predicted trajectory of dynamic obstacles during the autonomous driving vehicle's driving along the target path onto the target path to obtain a multi-scene projection map in time and space dimensions, wherein the multi-scene projection map at least includes an area where dynamic obstacles and autonomous driving vehicles are expected to conflict in multiple scenes; searching for a tree-like speed curve on the multi-scene projection map to obtain a tree-like speed curve set, and determining a target tree-like speed curve from the tree-like speed curve set, wherein the target tree-like speed curve is composed of multiple trajectories after a bifurcation point and a common trajectory before the bifurcation point, the multiple trajectories are used to respectively describe the speed planning of different scenarios expected to occur, and the common trajectory is used to describe the speed planning determined according to the probability of occurrence of the different scenarios; integrating the target path and the target tree-like speed curve to obtain a target tree-like trajectory, and using the target tree-like trajectory as the result of the autonomous driving vehicle decision planning.

[0006] In a second aspect, an embodiment of the present application provides an electronic device, including a memory, a processor, and a computer program stored in the memory, wherein the processor implements any of the methods described above when executing the computer program.

[0007] In a third aspect, an embodiment of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, any of the methods described above is implemented.

[0008] In a fourth aspect, an embodiment of the present application provides a computer program product, including computer instructions, which implement any of the methods described above when executed by a processor.

[0009] Compared with the related art, this application has the following advantages:

[0010] In an embodiment of the present application, a path is planned for an autonomous driving vehicle to obtain a target path; a multimodal predicted trajectory of a dynamic obstacle during the autonomous driving vehicle's driving along the target path is projected onto the target path to obtain a multi-scene projection map in time and space dimensions, wherein the multi-scene projection map at least includes an area where a conflict between dynamic obstacles and the autonomous driving vehicle is expected to occur in multiple scenes; a tree-like speed curve is searched on the multi-scene projection map to obtain a tree-like speed curve set, and a target tree-like speed curve is determined from the tree-like speed curve set, wherein the target tree-like speed curve is composed of multiple trajectories after a bifurcation point and a common trajectory before the bifurcation point, the multiple trajectories are used to respectively describe the speed planning of different scenarios that are expected to occur, and the common trajectory is used to describe the speed planning determined according to the probability of occurrence of the different scenarios; the target path and the target tree-like speed curve are integrated to obtain a target tree-like trajectory, and the target tree-like trajectory is used as the result of the decision planning of the autonomous driving vehicle. That is to say, when faced with a scenario where the movement intentions of dynamic obstacles (for example, other traffic participants) in the environment are uncertain, the embodiment of the present application proposes a plan-based planning method. Compared with the solution in the related technology that only outputs a single planned trajectory, the embodiment of the present application generates a tree-like trajectory. The multiple trajectory branches after the bifurcation point of the tree-like trajectory respectively handle different scenarios that may occur in the future, and the common segment trajectory before the bifurcation point comprehensively considers the probabilities of occurrence of different scenarios in the future to generate the most reasonable behavior, thereby achieving the technical effect of delaying decision-making as much as possible while ensuring safety, which is more in line with the idea of ​​human driving, and thus solves the problem that the related technology cannot provide accurate predictions of future risk events, resulting in unreasonable decision-making and planning, and achieves the technical effect of improving the rationality of decision-making and planning.

[0011] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] In the accompanying drawings, unless otherwise specified, the same reference numerals throughout the multiple drawings represent the same or similar parts or elements. These drawings are not necessarily drawn to scale. It should be understood that these drawings only depict some embodiments according to the present application and should not be regarded as limiting the scope of the present application.

[0013] Figure 1 A flowchart of an autonomous driving decision-making planning method provided in an embodiment of the present application is shown;

[0014] Figure 2 A multi-scene space-time projection diagram provided in an embodiment of the present application is shown;

[0015] Figure 3 A schematic diagram of a pre-planned planning scenario provided in an embodiment of the present application is shown;

[0016] Figure 4 A schematic diagram of a lane changing scenario with uncertain intention provided in an embodiment of the present application is shown;

[0017] Figure 5 A schematic diagram of forward traversal provided in an embodiment of the present application is shown;

[0018] Figure 6 A schematic diagram of reverse traversal of the main scene provided in an embodiment of the present application is shown;

[0019] Figure 7 A schematic diagram of reverse traversal of uncertain scenarios provided in an embodiment of the present application is shown;

[0020] Figure 8 A schematic diagram of tree-like track splicing provided in an embodiment of the present application is shown;

[0021] Fig. 9 A schematic diagram of selection and backtracking provided in an embodiment of the present application is shown;

[0022] Fig.10 The structure block diagram of the automatic driving decision-making and planning device provided in the embodiment of the present application is shown;

[0023] Fig.11 A block diagram of an electronic device used to implement an embodiment of the present application is shown. DETAILED DESCRIPTION

[0024] In the following, only some exemplary embodiments are briefly described. As those skilled in the art will appreciate, the described embodiments may be modified in various ways without departing from the concept or scope of the present application. Therefore, the drawings and descriptions are considered to be exemplary in nature and not restrictive.

[0025] To facilitate understanding of the technical solutions of the embodiments of the present application, the following describes the related technologies of the embodiments of the present application. The following related technologies can be combined with the technical solutions of the embodiments of the present application as optional solutions, and they all belong to the protection scope of the embodiments of the present application.

[0026] Explanation of terms

[0027] Path planning, or lateral planning, refers to planning the future movements of the steering wheel, which determines the geometry of the trajectory.

[0028] Speed ​​planning, also known as longitudinal planning, refers to planning the future actions of the accelerator / brake to determine when, at what speed, and at what acceleration to reach different positions on the path.

[0029] Multimodal prediction: The prediction module outputs multiple possible future motion trajectories of traffic participants in the environment.

[0030] The frozen robot problem. The main components of the autonomous driving system include: perception system, predictor, decision system, control system, and driving actuator. Once the environment perceived by the perception system exceeds a certain level of complexity, the decision system will determine that all forward paths are unsafe, and the robot (driving actuator) will freeze (or perform unnecessary actions) to avoid a collision of the autonomous driving vehicle.

[0031] The application scenarios in the embodiments of the present application include but are not limited to: long-distance freight logistics fleets, urban public transportation fleets, high-speed train formations, etc. In this application scenario, the main idea of ​​the current autonomous driving decision-making and planning scheme is to first predict the motion trajectory of other surrounding traffic participants in the next few seconds through a neural network-based predictor, and then plan the path and speed of the vehicle based on the predicted results, and finally combine the two to obtain a space-time trajectory. However, the predictor cannot provide accurate predictions for the future, and the following methods are generally used: Method 1, only use the predicted trajectory with the maximum probability. This type of method only makes decision-making plans for the vehicle based on the maximum probability predicted trajectory, which can work normally in most scenarios, but when other traffic participants in the environment make more radical behaviors (low-probability behaviors in the data distribution), only considering their most likely behaviors will cause the vehicle to respond too late, resulting in dangerous behaviors or even collision risks. Method 2, consider multimodal predicted trajectories. This type of method will represent several possible scenarios in the future based on the multimodal predicted trajectories of other traffic participants in the environment, and process them by simultaneously considering multimodal predicted trajectories or converting multimodal predicted trajectories into single-modal outputs in a certain way to ensure the safety of the vehicle's behavior. However, this approach considers all possible realizations of uncertainty, and its result is too conservative for the behavior of the ego vehicle, which may lead to inaccessible results in crowded environments. This phenomenon is also called the freezing robot problem in the field of robotics.

[0032] In view of the above technical problems, the technical solution of the present application and how the technical solution of the present application solves the above technical problems are described in detail with specific embodiments below. The several specific embodiments listed can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of the present application will be described in detail below in conjunction with the accompanying drawings.

[0033] An embodiment of the present application provides an autonomous driving decision-making planning method. In the embodiment of the present application, in order to simplify the complexity of the problem, the complete three-dimensional space-time trajectory planning problem is decoupled into two two-dimensional spatial planning, namely, path planning (lateral planning) and speed planning (longitudinal planning) are performed separately.

[0034] like Figure 1 The flowchart of the autonomous driving decision planning method according to an embodiment of the present application is shown, including:

[0035] Step S102, performing path planning for the autonomous driving vehicle to obtain a target path.

[0036] Optionally, in an embodiment of the present application, the specific implementation of step S102 may be: path planning is implemented based on information such as environmental perception, map data, static / dynamic obstacles, etc., and a geometric curve is output to describe the future path of the autonomous driving vehicle, that is, the above-mentioned target path, thereby completing the path planning. It should be noted that, with respect to environmental perception, the autonomous driving vehicle is equipped with a variety of sensors, such as lidar, camera, radar and ultrasonic sensor, which collect information about the surrounding environment in real time, including road shape, lane lines, vehicles and pedestrians, etc. With respect to map data, the autonomous driving vehicle uses high-resolution map data, which contains detailed road information, such as vehicle driving path, lane lines, traffic signs, traffic light locations, etc. With respect to static obstacles, these are fixed objects, such as roadside buildings, railings, traffic islands, etc. Based on high-precision maps and real-time perception data, the system can accurately locate these obstacles. With respect to dynamic obstacles, these are moving objects, such as other vehicles, pedestrians and bicycles, to clarify their speed and trajectory to predict their future positions. The above-mentioned path planning algorithm may include global path planning and local path planning, wherein global path planning is the overall path planning from the starting point to the target, which is usually guided by high-precision maps and road networks. Local path planning is to consider the changes in the surrounding environment in real time under the guidance of the global path, and adjust the specific driving path, including avoiding dynamic obstacles such as slow pedestrians and vehicles that suddenly change lanes in front. Typical algorithms include dynamic window approach (DWA), Bezier curves, dynamic path trees, etc. Regarding trajectory generation, the final generated trajectory is a geometric curve, which is usually a smooth and feasible vehicle driving path. These curves may take mathematical forms such as polynomial curves, spline curves, and Bezier curves to ensure that autonomous driving vehicles travel on smooth and safe routes.

[0037] The above step S102 is equivalent to completing the lateral planning. Since in the real driving scene, the driving principle of "giving way but not giving way" is adhered to, most effective interaction events occur in the longitudinal direction. Therefore, the embodiment of the present application proposes the following steps S104 to S106 for longitudinal planning. Optionally, in step S104, the multimodal predicted trajectory of the dynamic obstacle during the driving of the autonomous driving vehicle along the target path is projected onto the target path to obtain a multi-scene projection map in the time and space dimensions, wherein the multi-scene projection map at least includes the area where the dynamic obstacle and the autonomous driving vehicle are expected to conflict in multiple scenes. The conflicting area in this step represents the place where the autonomous driving vehicle and the obstacle may collide at a certain time point and position in the future. For example, assuming that there is a car in front as a dynamic obstacle, three possible trajectories are predicted based on its behavior, continuing to go straight, turning right at a certain intersection, and stopping. The embodiment of the present application projects these trajectories onto the target path of the autonomous driving vehicle and represents them in a two-dimensional time and space map to form a conflict area. The multi-scene projection map in the time and space dimensions further ensures that the autonomous driving system can identify potential collision risks in advance and take appropriate avoidance measures, thereby improving the safety and reliability of driving.

[0038] Optionally, in an embodiment of the present application, the above-mentioned multimodal predicted trajectory generally refers to the autonomous driving system not only considering one possible trajectory but also considering multiple possible trajectories at the same time when predicting the future movement paths of other traffic participants (such as pedestrians, bicycles, cars, etc.), and its determination method may include: data collection and perception, including sensor data, data fusion (through multi-sensor data fusion, each dynamic obstacle can be accurately located and tracked in three-dimensional space); state estimation, including motion state (for each dynamic obstacle, its motion state needs to be estimated, including position, speed, acceleration and direction), motion model (based on historical motion data, build a motion model of the obstacle, such as the commonly used uniform speed, uniform acceleration or more complex trajectory model. For example, Kalman Filter and Extended Kalman Filter Filter (EKF) is commonly used for linear and nonlinear motion state estimation); behavior modeling, including behavior understanding (understanding the possible behavior patterns of obstacles, such as pedestrians may stop and wait at a certain intersection, vehicles may suddenly change lanes, etc.), statistical learning (using a large amount of historical traffic data to train machine learning models that can recognize and predict different behaviors. Commonly used methods include decision trees, random forests, neural networks, etc.); trajectory generation, including multimodal prediction (generating multiple possible future trajectories for each dynamic obstacle, which reflect the possible movement of the obstacle in different situations. For example, a pedestrian may continue to go straight, suddenly slow down or turn), probability distribution (assign a probability to each predicted trajectory, indicating the probability that different trajectories may occur. This is usually achieved through Bayesian methods or deep learning models, such as Generative Adversarial Networks (GANs). Networks, GAN) or Variational Autoencoder (VAE); model integration and updating, including model fusion (combining the results of different prediction models to obtain more reliable prediction results through integration methods. For example, it can be achieved by weighted averaging or selecting the trajectory with the maximum probability), real-time updating (as new sensor data is continuously input, the system updates the prediction results in real time to make them more accurate and reliable).

[0039] The multimodal predicted trajectory of the above-mentioned dynamic obstacle projected onto the target path may include spatial projection and temporal projection, wherein the spatial projection may be to convert the two-dimensional or three-dimensional trajectory into position coordinates on the actual road, and the temporal projection may add a corresponding timestamp to each trajectory point to determine the position of the obstacle at different time points. Assuming that the above-mentioned multiple scenes include two scenes, for example, the main scene (referring to those most common and main driving environments and situations in autonomous driving applications. These scenes are usually predictable and relatively stable, and the autonomous driving system can make decisions more reliably in these scenes) and uncertain scenes (referring to those driving environments and situations with more uncertainties and complex changes. In these scenes, the autonomous driving system faces more challenges and needs to have higher levels of perception, prediction and decision-making capabilities). Then the multi-scene projection diagram in the time and space dimensions and the time-space conflict area can be as follows Figure 2 In addition, there will be a pre-decision module to make a decision on whether to rush in or give way in each conflict area.

[0040] Step S106, searching for a tree-like speed curve on the multi-scene projection graph to obtain a tree-like speed curve set, and determining a target tree-like speed curve from the tree-like speed curve set, wherein the target tree-like speed curve is composed of a plurality of trajectories after a bifurcation point and a common trajectory before the bifurcation point, the plurality of trajectories are used to respectively describe speed plans of different scenes that are expected to occur, and the common trajectory is used to describe a speed plan determined according to the probability of occurrence of the different scenes.

[0041] It should be noted that the algorithms used in the above search process include but are not limited to: dynamic programming algorithm, genetic algorithm, Monte Carlo tree search algorithm, deep reinforcement learning algorithm, etc. This is not limited here. In addition, it should be noted that the tree-like speed curve set may include one or more tree-like speed curves, which is not limited here. Determining the target tree-like speed curve in the tree-like speed curve set includes but is not limited to: determining the target tree-like speed curve based on minimizing the expected cost of all possible scenarios.

[0042] Step S108, integrating the target path and the target tree-like speed curve to obtain a target tree-like trajectory, and using the target tree-like trajectory as a result of the decision planning of the autonomous driving vehicle.

[0043] Through the above steps S102 to S108, the path of the autonomous driving vehicle is planned to obtain a target path; the multi-modal predicted trajectory of the dynamic obstacles of the autonomous driving vehicle during the driving process along the target path is projected onto the target path to obtain a multi-scene projection map in time and space dimensions, wherein the multi-scene projection map at least includes the area where the dynamic obstacles and the autonomous driving vehicle are expected to conflict in multiple scenes; the tree-like speed curve is searched on the multi-scene projection map to obtain a tree-like speed curve set, and the target tree-like speed curve is determined in the tree-like speed curve set, wherein the target tree-like speed curve is composed of multiple trajectories after the bifurcation point and the common trajectory before the bifurcation point, the multiple trajectories are used to respectively describe the speed planning of different scenes expected to occur, and the common trajectory is used to describe the speed planning determined according to the probability of occurrence of the different scenes; the target path and the target tree-like speed curve are integrated to obtain a target tree-like trajectory, and the target tree-like trajectory is used as the result of the decision-making planning of the autonomous driving vehicle. That is to say, when facing a scene where the movement intention of a dynamic obstacle (for example, other traffic participants) in the environment is uncertain, the embodiment of the present application proposes a plan-based planning method (such as Figure 3 As shown in the figure, the vehicle on the left is the self-vehicle and the other vehicle on the right. The probability of 0.7 is the main scenario and the probability of 0.3 is the uncertain scenario. There are two trajectories after the bifurcation point, which are the speed plans for different scenarios that are expected to occur. The speed plans before the bifurcation point are determined according to the probability of occurrence of different scenarios. Compared with the solution of only outputting a single planned trajectory in the related art (such as Figure 4 As shown, the vehicle has only one planned trajectory), the embodiment of the present application generates a tree-like trajectory. The multiple trajectory branches after the bifurcation point of the tree-like trajectory respectively handle different scenarios that may occur in the future. The common segment trajectory before the bifurcation point comprehensively considers the probabilities of different scenarios in the future to generate the most reasonable behavior, thereby achieving the technical effect of delaying decision-making as much as possible under the premise of ensuring safety, which is more in line with the idea of ​​human driving, and further solves the problem that the relevant technology cannot provide accurate predictions of future risk events, resulting in unreasonable decisions and planning, and achieves the technical effect of improving the rationality of decision-making and planning.

[0044] In one possible implementation, a tree-like speed curve is searched on a multi-scene projection map to obtain a tree-like speed curve set, including: S11, setting a bifurcation time, wherein the bifurcation time is used to describe the time point at which the autonomous driving vehicle has confirmed the risk event. That is, the bifurcation time description assumes that at this time point, the vehicle can obtain more information to eliminate or confirm the risk event. S12, discretizing the multi-scene projection map to obtain a discretized multi-scene prediction matrix; S13, using a dynamic programming algorithm to search in the discretized multi-scene prediction matrix to obtain the tree-like speed curve set. Through the above S11 to S13, by discretizing the multi-scene projection map, continuous environmental data can be converted into discrete grid units, simplifying the computational complexity, and then using a dynamic programming algorithm to search in the discretized multi-scene prediction matrix, thereby improving processing speed and decision-making efficiency.

[0045] The above-mentioned multiple scenes may include at least a first scene and a second scene, and the probability of occurrence of the first scene is greater than the probability of occurrence of the second scene. That is, the first scene is the main scene, and the second scene is an uncertain scene, wherein, in the discretized multi-scene prediction matrix, a dynamic programming algorithm is used to search to obtain the tree-like speed curve set, which may include: S21, forward search for the conflicting areas in all scenes in the multiple scenes to obtain a forward traversal result; S22, reverse search for the conflicting areas in the first scene and the second scene respectively to obtain a reverse traversal result; S23, splicing the forward traversal result and the reverse traversal result to obtain the tree-like speed curve set. That is to say, when searching the tree-like data curve, the space-time conflict in all scenes is considered, and the forward traversal is performed, and the space-time conflict of only the main scene is considered, and the reverse traversal is performed, and the space-time conflict of only the uncertain scene is considered, and the reverse traversal is performed, and the forward traversal result and the reverse traversal result are spliced ​​to obtain the tree-like speed curve set, thereby realizing efficient path planning while maintaining high adaptability to environmental changes. This method provides strong technical support for the safe driving of autonomous driving systems in complex environments.

[0046] Optionally, forward search is performed on the conflicting areas in all the scenarios in the multiple scenarios to obtain the forward traversal results, which may include: S31, forward search in the traversable set from the starting point until the bifurcation time, obtain a bifurcation point candidate set, and determine the top-ranked cumulative cost from the starting point to each bifurcation point in the bifurcation point candidate set to obtain a first cumulative cost set. It should be noted that the top-ranked cumulative cost may be the optimal cumulative cost from the starting point to the bifurcation point.

[0047] That is, considering the space-time conflicts in all scenarios, we search forward from the starting point in the passable set until the bifurcation time, and obtain all feasible bifurcation points, that is, the bifurcation point candidate set, and the optimal cumulative cost from the starting point to the bifurcation point N. C(N) represents the cost evaluation result of a bifurcation point. Generally, the cumulative cost includes the cost corresponding to at least one of the following factors during the driving process of the autonomous driving vehicle: smoothness, efficiency, and safety. Figure 5 As shown, there are three optimal curves from the starting point to the candidate set of bifurcation points.

[0048] S32, respectively, performs a reverse search on the areas where conflicts occur in the first scenario and the second scenario to obtain a reverse traversal result, including: starting from the end point, reversely searching in the passable set until the bifurcation time, and determining a cumulative cost set with a top ranking from the end point to each bifurcation point in the candidate set of the bifurcation point, to obtain a second cumulative cost set and a third cumulative cost set.

[0049] That is, considering only the space-time conflict of the main scenario, starting from the potential end point and searching backward in the passable set until the bifurcation time, the optimal cumulative cost from the bifurcation point N to the end point in the main scenario can be calculated. like Figure 6 As shown in Figure 1, there are three optimal curves from the end point to the candidate set of bifurcation points. Considering only the space-time conflict in the uncertain scenario, starting from the potential end point and searching backward in the passable set until the bifurcation time, the optimal cumulative cost from the bifurcation point N to the end point in the uncertain scenario can be calculated. like Figure 7 As shown, there are two optimal curves from the potential end point to the fork candidate node N.

[0050] Determining the target tree-like speed curve in the tree-like speed curve set includes: S41, combining the occurrence probabilities of the first scene and the second scene, calculating the cumulative total costs belonging to the same bifurcation point in the first cumulative cost set, the second cumulative cost set, and the third cumulative cost set to obtain a cumulative total cost set; S42, determining the target cumulative total cost that meets the preset conditions from the cumulative total cost set; S43, setting the bifurcation point corresponding to the target cumulative total cost as the target bifurcation point; S44, determining the target tree-like speed curve according to the target bifurcation point. That is, the results of the forward traversal and the reverse traversal are spliced ​​together, such as Figure 8 As shown. Considering the probability distribution of different scenarios, the total cost of each bifurcation point N in the bifurcation point candidate set is calculated, where p represents the probability of each scenario occurring, as shown in Formula 1:

[0051] Cumulative total cost

[0052] The target tree-like speed curve is determined according to the target bifurcation point, including: S51, based on the target bifurcation point, backtracking in the direction of the starting point and the direction of the end point respectively, to obtain a first backtracking result and a second backtracking result; S52, determining the target tree-like speed curve through the first backtracking result and the second backtracking result. That is, in the bifurcation point candidate set, the bifurcation point with the smallest cumulative total cost is found, and recorded as the optimal bifurcation point. Based on the optimal bifurcation point, backtracking in the direction of the starting point and the end point respectively, a complete tree-like speed curve is obtained. Fig. 9 As shown in the figure, the bifurcation point in the middle of the bifurcation point candidate set is determined as the target bifurcation point, and the complete tree-like speed curve is as follows: Fig. 9 shown.

[0053] In summary, when faced with a scenario where the movement intentions of other traffic participants in the environment are uncertain, the embodiment of the present application proposes a plan-based planning method. Compared with the traditional idea of ​​outputting a single trajectory, this method ultimately generates a tree-like trajectory. The multiple trajectory branches after the bifurcation point handle different scenarios that may occur in the future respectively. The common segment trajectory before the bifurcation point will comprehensively consider the probability of different scenarios in the future to generate the most reasonable behavior. This method achieves the effect of delaying decision-making as much as possible while ensuring safety, which is more in line with the idea of ​​human driving. In addition, the embodiment of the present application innovatively proposes a tree-like speed curve search method based on dynamic programming. This method can search for the speed curve with the minimum expected cost for all scenarios based on the probability distribution of different scenarios, and has the optimal completeness under resolution.

[0054] Corresponding to the application scenario and method of the method provided in the embodiment of the present application, the embodiment of the present application also provides an automatic driving decision-making planning device. Fig.10 The figure shows a structural block diagram of an automatic driving decision-making planning device according to an embodiment of the present application, and the device may include:

[0055] A first determination module 102 is used to perform path planning for the autonomous driving vehicle to obtain a target path;

[0056] The projection module 104 is used to project the multimodal predicted trajectory of the dynamic obstacles of the autonomous driving vehicle during the driving process along the target path onto the target path to obtain a multi-scene projection map in time and space dimensions, wherein the multi-scene projection map at least includes an area where the dynamic obstacles and the autonomous driving vehicle are expected to conflict in multiple scenes;

[0057] The second determination module 106 searches for a tree-like speed curve on the multi-scenario projection graph to obtain a tree-like speed curve set, and determines a target tree-like speed curve from the tree-like speed curve set, wherein the target tree-like speed curve is composed of a plurality of trajectories after a bifurcation point and a common trajectory before the bifurcation point, the plurality of trajectories are used to respectively describe speed plans of different scenarios that are expected to occur, and the common trajectory is used to describe speed plans determined according to the probability of occurrence of the different scenarios;

[0058] The third determination module 108 is used to integrate the target path and the target tree-like speed curve to obtain a target tree-like trajectory, and use the target tree-like trajectory as a result of the decision planning of the autonomous driving vehicle.

[0059] pass Fig.10 The device shown in the figure performs path planning for the autonomous driving vehicle to obtain a target path; the multi-modal predicted trajectory of the dynamic obstacles of the autonomous driving vehicle during its driving along the target path is projected onto the target path to obtain a multi-scene projection map in time and space dimensions, wherein the multi-scene projection map at least includes the area where the dynamic obstacles and the autonomous driving vehicle are expected to conflict in multiple scenes; the tree-like speed curve is searched on the multi-scene projection map to obtain a tree-like speed curve set, and a target tree-like speed curve is determined in the tree-like speed curve set, wherein the target tree-like speed curve is composed of multiple trajectories after the bifurcation point and a common trajectory before the bifurcation point, the multiple trajectories are used to respectively describe the speed planning of different scenes expected to occur, and the common trajectory is used to describe the speed planning determined according to the probability of occurrence of the different scenes; the target path and the target tree-like speed curve are integrated to obtain a target tree-like trajectory, and the target tree-like trajectory is used as the result of the decision-making planning of the autonomous driving vehicle. That is to say, when facing a scene where the movement intention of a dynamic obstacle (for example, other traffic participants) in the environment is uncertain, the embodiment of the present application proposes a plan-based planning method (such as Figure 3 As shown in the figure, the vehicle on the left is the self-vehicle and the other vehicle on the right. The probability of 0.7 is the main scenario and the probability of 0.3 is the uncertain scenario. There are two trajectories after the bifurcation point, which are the speed plans for different scenarios that are expected to occur. The speed plans before the bifurcation point are determined according to the probability of occurrence of different scenarios. Compared with the solution of only outputting a single planned trajectory in the related art (such as Figure 4As shown, there is only one planned trajectory itself), the embodiment of the present application generates a tree-like trajectory. The multiple trajectory branches after the bifurcation point of the tree-like trajectory respectively handle different scenarios that may occur in the future. The common segment trajectory before the bifurcation point comprehensively considers the probabilities of different scenarios in the future to generate the most reasonable behavior, thereby achieving the technical effect of delaying decision-making as much as possible under the premise of ensuring safety, which is more in line with the idea of ​​human driving, and further solves the problem that the relevant technology cannot provide accurate predictions of future risk events, resulting in unreasonable decision-making and planning, and achieves the technical effect of improving the rationality of decision-making and planning.

[0060] In one possible implementation, the second determination module 106 includes: a setting unit, used to set the bifurcation time, wherein the bifurcation time is used to describe the time point when the autonomous driving vehicle has confirmed the risk event; a processing unit, used to discretize the multi-scene projection map to obtain a discretized multi-scene prediction matrix; a first determination unit, used to search in the discretized multi-scene prediction matrix using a dynamic programming algorithm to obtain the tree-like speed curve set.

[0061] The above-mentioned multiple scenes include at least a first scene and a second scene, and the probability of occurrence of the first scene is greater than the probability of occurrence of the second scene, wherein the first determination unit includes: a first search subunit, used to perform a forward search on the areas where conflicts occur in all scenes in the multiple scenes to obtain a forward traversal result; a second search subunit, used to perform a reverse search on the areas where conflicts occur in the first scene and the second scene respectively to obtain a reverse traversal result; a splicing subunit, used to splice the forward traversal result and the reverse traversal result to obtain the tree-like speed curve set.

[0062] Optionally, the first search subunit is also used to search forward in the passable set from the starting point until the bifurcation time, obtain a candidate set of bifurcation points, and determine the cumulative cost from the starting point to each bifurcation point in the candidate set of bifurcation points, to obtain a first cumulative cost set; the second search subunit is also used to search backward in the passable set from the end point until the bifurcation time, and determine the cumulative cost set from the end point to each bifurcation point in the candidate set of bifurcation points, to obtain a second cumulative cost set and a third cumulative cost set.

[0063] Optionally, the second determination module 106 includes: a calculation unit, used to calculate the cumulative total cost belonging to the same bifurcation point in the first cumulative cost set, the second cumulative cost set and the third cumulative cost set in combination with the occurrence probabilities of the first scenario and the second scenario, to obtain a cumulative total cost set; a second determination unit, used to determine a target cumulative total cost that meets preset conditions from the cumulative total cost set; a third determination unit, used to set the bifurcation point corresponding to the target cumulative total cost as the target bifurcation point; and a fourth determination unit, used to determine the target tree-like speed curve based on the target bifurcation point.

[0064] Optionally, the fourth determination unit includes: a backtracking subunit, used to backtrack in the starting direction and the end direction respectively based on the target bifurcation point to obtain a first backtracking result and a second backtracking result; a determination subunit, used to determine the target tree-like speed curve through the first backtracking result and the second backtracking result.

[0065] The above-mentioned cumulative cost includes the cost corresponding to at least one of the following factors during the driving process of the autonomous driving vehicle: smoothness, efficiency, and safety.

[0066] The functions of each module in each device in the embodiments of the present application can be found in the corresponding description in the above method, and have corresponding beneficial effects, which will not be repeated here.

[0067] Fig.11 FIG. 1 is a block diagram of an electronic device used to implement an embodiment of the present application. Fig.11 As shown, the electronic device includes: a memory 1101 and a processor 1102. The memory 1101 stores a computer program that can be run on the processor 1102. When the processor 1102 executes the computer program, the method in the above embodiment is implemented. The number of the memory 1101 and the processor 1102 can be one or more.

[0068] The electronic device also includes:

[0069] The communication interface 1103 is used to communicate with external devices and perform data exchange transmission.

[0070] If the memory 1101, the processor 1102 and the communication interface 1103 are implemented independently, the memory 1101, the processor 1102 and the communication interface 1103 can be connected to each other through a bus and communicate with each other. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Fig.11 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0071] Optionally, in a specific implementation, if the memory 1101, the processor 1102 and the communication interface 1103 are integrated on a chip, the memory 1101, the processor 1102 and the communication interface 1103 can communicate with each other through an internal interface.

[0072] An embodiment of the present application provides a computer-readable storage medium storing a computer program, which implements the method provided in the embodiment of the present application when the program is executed by a processor.

[0073] An embodiment of the present application also provides a chip, which includes a processor for calling and executing instructions stored in the memory from the memory, so that a communication device equipped with the chip executes the method provided by the embodiment of the present application.

[0074] An embodiment of the present application also provides a chip, including: an input interface, an output interface, a processor and a memory, wherein the input interface, the output interface, the processor and the memory are connected via an internal connection path, and the processor is used to execute the code in the memory. When the code is executed, the processor is used to execute the method provided in the embodiment of the application.

[0075] It should be understood that the processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc. It is worth noting that the processor may be a processor supporting the Advanced RISC Machines (ARM) architecture.

[0076] Further, optionally, the above-mentioned memory may include a read-only memory and a random access memory. The memory may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memory. Among them, the non-volatile memory may include a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory may include a random access memory (RAM), which is used as an external cache. By way of exemplary but not limiting description, many forms of RAM are available. For example, static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM) and direct memory bus random access memory (DR RAM).

[0077] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function according to the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium.

[0078] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine different embodiments or examples described in this specification and the features of different embodiments or examples, unless they are contradictory.

[0079] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include at least one of the features. In the description of this application, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0080] Any process or method described in the flow chart or otherwise described herein can be understood as a module, fragment or portion of a code representing one or more executable instructions for implementing the steps of a specific logical function or process. And the scope of the preferred embodiment of the present application includes other implementations, in which the functions may not be performed in the order shown or discussed, including in a substantially simultaneous manner or in a reverse order according to the functions involved.

[0081] The logic and / or steps described in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, which can be specifically implemented in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or used in combination with these instruction execution systems, devices or apparatuses.

[0082] It should be understood that the various parts of the present application can be implemented with hardware, software, firmware or a combination thereof. In the above embodiments, multiple steps or methods can be implemented with software or firmware stored in a memory and executed by a suitable instruction execution system. All or part of the steps of the above embodiment method can be completed by instructing the relevant hardware through a program, which can be stored in a computer-readable storage medium, and when the program is executed, it includes one of the steps of the method embodiment or a combination thereof.

[0083] In addition, each functional unit in each embodiment of the present application can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into one module. The above-mentioned integrated module can be implemented in the form of hardware or in the form of a software functional module. If the above-mentioned 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. The storage medium can be a read-only memory, a disk or an optical disk, etc.

[0084] The above is only an exemplary embodiment of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of various changes or substitutions within the technical scope recorded in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application shall be based on the protection scope of the claims.

Claims

1. An autonomous driving decision planning method, comprising: Perform path planning for the autonomous driving vehicle to obtain the target path; Projecting the multimodal predicted trajectory of the dynamic obstacles of the autonomous driving vehicle during its driving along the target path onto the target path to obtain a multi-scene projection map in time and space dimensions, wherein the multi-scene projection map at least includes an area where the dynamic obstacles and the autonomous driving vehicle are expected to conflict in multiple scenarios; Searching for a tree-like speed curve on the multi-scenario projection graph to obtain a tree-like speed curve set, and determining a target tree-like speed curve from the tree-like speed curve set, wherein the target tree-like speed curve is composed of a plurality of trajectories after a bifurcation point and a common trajectory before the bifurcation point, the plurality of trajectories are used to respectively describe speed plans of different scenarios that are expected to occur, and the common trajectory is used to describe a speed plan determined according to the probability of occurrence of the different scenarios; The target path and the target tree-like speed curve are integrated to obtain a target tree-like trajectory, and the target tree-like trajectory is used as a result of the decision planning of the autonomous driving vehicle.

2. The method according to claim 1, wherein: Searching for tree-like speed curves on the multi-scene projection graph to obtain a tree-like speed curve set includes: Setting a bifurcation time, wherein the bifurcation time is used to describe the time point at which the autonomous driving vehicle has confirmed a risk event; Discretize the multi-scene projection graph to obtain a discretized multi-scene prediction matrix; In the discretized multi-scenario prediction matrix, a dynamic programming algorithm is used to search and obtain the tree-like speed curve set.

3. The method according to claim 2, wherein the multiple scenarios include at least a first scenario and a second scenario, the probability of occurrence of the first scenario is greater than the probability of occurrence of the second scenario, wherein: In the discretized multi-scenario prediction matrix, a dynamic programming algorithm is used to search to obtain the tree-like speed curve set, including: Perform a forward search on the conflicting areas in all the scenarios in the multiple scenarios to obtain a forward traversal result; Performing a reverse search on the conflicting areas in the first scenario and the second scenario respectively to obtain a reverse traversal result; The forward traversal result and the reverse traversal result are concatenated to obtain the tree-like speed curve set.

4. According to the method of claim 3, forward search is performed on the conflicting areas in all the scenarios in the multiple scenarios to obtain forward traversal results, including: Starting from the starting point, forward searching in the passable set until the bifurcation time is reached, obtaining a bifurcation point candidate set, and determining a cumulative cost from the starting point to each bifurcation point in the bifurcation point candidate set that is ranked first, obtaining a first cumulative cost set; Perform a reverse search on the areas where conflicts occur in the first scenario and the second scenario respectively to obtain a reverse traversal result, including: starting from the end point, searching backward in the passable set until the bifurcation time, and determining a cumulative cost set with a top ranking from the end point to each bifurcation point in the bifurcation point candidate set, to obtain a second cumulative cost set and a third cumulative cost set.

5. The method according to claim 4, wherein: Determining a target tree-like speed curve from the tree-like speed curve set includes: In combination with the occurrence probabilities of the first scenario and the second scenario, the cumulative total cost belonging to the same bifurcation point in the first cumulative cost set, the second cumulative cost set, and the third cumulative cost set is calculated to obtain a cumulative total cost set; Determine a target cumulative total cost that meets a preset condition from the cumulative total cost set; Setting the bifurcation point corresponding to the target cumulative total cost as the target bifurcation point; The target tree-like speed curve is determined according to the target bifurcation point.

6. The method according to claim 5, wherein: Determining the target tree-like speed curve according to the target bifurcation point includes: Based on the target bifurcation point, backtracking is respectively performed in the direction of the starting point and the direction of the end point to obtain a first backtracking result and a second backtracking result; The target tree-like speed curve is determined by the first back-tracing result and the second back-tracing result.

7. The method according to claim 6, wherein: The cumulative cost includes the cost corresponding to at least one of the following factors during the driving process of the autonomous driving vehicle: smoothness, efficiency, and safety.

8. An electronic device comprising a memory, a processor and a computer program stored in the memory, wherein the processor implements the method according to claims 1 to 7 when executing the computer program.

9. A computer-readable storage medium, wherein a computer program is stored in the computer-readable storage medium, and when the computer program is executed by a processor, the method according to claims 1 to 7 is implemented.

10. A computer program product, comprising computer instructions, wherein when the computer instructions are executed by a processor, the method according to claims 1-7 is implemented.

Images