Speed ​​planning methods and devices, control equipment, vehicles, and storage media

Abstract

This invention discloses a speed planning method and apparatus. The method includes acquiring a preset obstacle ST risk field and speed decision results, wherein the obstacle ST risk field is pre-constructed based on multi-frame prediction information of obstacles; determining the constraint boundary of a selected speed planning model based on the obstacle ST risk field and speed decision results; determining the objective optimization function of the selected speed planning model based on a cost function; and solving the speed planning model according to the determined constraint boundary and objective optimization function to obtain a speed planning S-T curve. The solution of this invention can ensure the consistency of changes between consecutive frames, avoid the adverse effects of sudden changes in speed planning results between consecutive frames due to low accuracy of real-time predicted trajectory, and mitigate the impact of input jitter on speed planning results. Furthermore, it effectively reduces the impact of prediction result fluctuations on vehicle speed, improving the system's fault tolerance and the comfort of autonomous driving.

Description

Technical Field

[0001] This invention relates to the field of autonomous driving technology, and more particularly to a speed planning method, a speed planning device, a control device, a vehicle including the control device, and a storage medium. Background Technology

[0002] Artificial intelligence (AI) is the theory, methods, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to achieve optimal results. Autonomous driving is a mainstream application of AI. Autonomous driving technology relies on the collaborative efforts of computer vision, radar, monitoring devices, and global positioning systems to enable motor vehicles to drive autonomously without human intervention.

[0003] In the field of autonomous driving technology, the main key technologies currently include autonomous localization, environmental perception, behavioral decision-making, trajectory planning, and motion control. Speed ​​planning, as a crucial component of trajectory planning, refers to planning a speed curve for the autonomous vehicle over a future time period. The quality of this speed curve determines the safety of the vehicle's motion control and the comfort of driving actions, such as the comfort of longitudinal driving actions and even lateral driving actions. Currently, commonly used speed planning methods generally involve first performing collision detection between the predicted trajectory of obstacles and the vehicle's path. When a collision occurs, a corresponding obstacle ST diagram (distance-time graph between the obstacle and the vehicle, e.g., ...) is generated based on the projection of the predicted trajectory. Figure 1 The diagram shows the speed planning (ST) graph projected onto the left vehicle based on its predicted trajectory. According to the predicted trajectory, the obstacle vehicle on the left cuts in front of the vehicle at a distance s1 meters after time t1. Then, based on the obstacle's ST graph, dynamic programming or numerical optimization methods are used to obtain an ST curve that does not intersect with the obstacle's ST graph. Finally, the corresponding VT / AT curves are obtained by taking the first / second derivative of this curve. However, these commonly used speed planning methods only consider the ST graph of a single frame of the obstacle (the distance-time graph between the obstacle and the vehicle), which requires high quality in the predicted trajectory of the obstacle. The quality of the predicted trajectory also depends on the accuracy and stability of perception and recognition. Therefore, when the quality of the predicted trajectory of the obstacle is low or the perception and recognition module produces errors, using the obstacle's ST graph projected from the predicted trajectory for real-time speed planning can easily lead to abrupt changes in the speed planning results between frames, causing adverse phenomena such as sudden braking and affecting the comfort of autonomous driving. Summary of the Invention

[0004] This invention provides a speed planning scheme to solve the problems in the prior art where speed planning based on the accurate real-time predicted trajectory projection of obstacles is highly dependent on the accuracy and stability of perception and recognition, and is prone to sudden changes in speed planning results between frames, causing sudden braking.

[0005] In a first aspect, embodiments of the present invention provide a speed planning method, the method comprising:

[0006] Obtain a preset obstacle ST risk field and velocity decision results, wherein the obstacle ST risk field is pre-constructed based on multi-frame prediction information of the obstacle;

[0007] The constraint boundaries of the selected speed planning model are determined based on the obstacle ST risk field and the speed decision results;

[0008] The objective optimization function of the velocity planning model is determined based on the cost function.

[0009] The velocity planning model is solved based on the constraint boundary and objective optimization function to obtain the velocity planning ST curve.

[0010] Secondly, embodiments of the present invention provide a speed planning device, the device comprising:

[0011] The data acquisition module is used to acquire the preset obstacle ST risk field and velocity decision results, wherein the obstacle ST risk field is pre-constructed based on multi-frame prediction information of the obstacle;

[0012] The constraint determination module is used to determine the constraint boundary of the selected speed planning model based on the obstacle ST risk field and the speed decision result;

[0013] The optimization objective determination module is used to determine the objective optimization function of the velocity planning model based on the cost function; and

[0014] The curve generation module is used to solve the velocity planning model based on the constraint boundary and the objective optimization function to obtain the velocity planning ST curve output.

[0015] Thirdly, embodiments of the present invention provide another speed planning device, which includes:

[0016] Memory, used to store executable instructions; and

[0017] A processor for executing executable instructions stored in memory, which, when executed by the processor, implement the steps of the method described above.

[0018] Fourthly, embodiments of the present invention provide a control device, which includes:

[0019] A planner for speed planning according to the method described above; and

[0020] The controller is used to control the speed of the vehicle according to the speed planning ST curve determined by the planner.

[0021] Fifthly, the present invention provides a storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described method.

[0022] In a sixth aspect, the present invention provides a vehicle that includes the control device described in the fourth aspect above.

[0023] In a seventh aspect, the present invention provides a computer program product comprising a computer program stored on a non-volatile computer-readable storage medium, the computer program comprising program instructions that, when executed by a computer, cause the computer to perform the speed planning method described above.

[0024] The beneficial effects of the embodiments of the present invention are as follows: The method provided by the embodiments of the present invention utilizes the obstacle ST risk field pre-constructed based on multi-frame obstacle prediction information for speed planning. In particular, when determining the constraint boundary, the constraint boundary is determined based on the obstacle ST risk field and the speed decision result output by the front module. This ensures that the determined constraint conditions simultaneously consider the multi-frame obstacle prediction information, thus fuzzing the uncertainty of the obstacle position. Therefore, the speed ST curve planned based on the constraint conditions can ensure the consistency of changes between consecutive frames, avoiding the adverse effects of sudden changes in the speed planning results between consecutive frames due to the low accuracy of real-time prediction trajectory and the impact of input jitter on the speed planning results. Furthermore, it effectively reduces the impact of prediction result fluctuations on vehicle speed, improving the system's fault tolerance and the comfort of autonomous driving. Attached Figure Description

[0025] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0026] Figure 1 This is a schematic diagram illustrating the processing method of speed planning in existing technologies;

[0027] Figure 2 This is a flowchart of a speed planning method according to an embodiment of the present invention;

[0028] Figure 3This is a schematic diagram of an obstacle risk field environment modeling method provided in an embodiment of the present invention;

[0029] Figures 4A to 4C This is a schematic diagram of SL coordinate system update provided in an embodiment of the present invention;

[0030] Figures 5A to 5C This is a schematic diagram of ST coordinate system update provided in an embodiment of the present invention;

[0031] Figure 6 This is a schematic diagram of a method for determining the currently occupied area of ​​a static obstacle in the obstacle grid map of the current frame, as described in an embodiment of the present invention.

[0032] Figure 7 This is a schematic diagram of a method for determining the currently occupied area of ​​a dynamic obstacle in the obstacle grid map of the current frame, as described in an embodiment of the present invention.

[0033] Figure 8 This is a schematic diagram of the current frame ST risk field determination method for dynamic obstacles in an embodiment of the present invention;

[0034] Figure 9 yes Figure 8 A schematic diagram of the implementation method of step S132;

[0035] Figure 10 This is a flowchart of an embodiment of the velocity decision method based on probabilistic grid diagrams of the present invention;

[0036] Figure 11 This is a flowchart of another embodiment of the velocity decision method based on probabilistic grid diagrams of the present invention;

[0037] Figure 12 This is a flowchart of another embodiment of the velocity decision method based on probabilistic grid diagrams of the present invention;

[0038] Figure 13 This is a flowchart of another embodiment of the velocity decision method based on probabilistic grid diagrams of the present invention;

[0039] Figure 14 This is a flowchart of another embodiment of the velocity decision method based on probabilistic grid diagrams of the present invention;

[0040] Figure 15 A schematic diagram of a feasible decision result provided in an embodiment of the present invention in an ST probability grid diagram;

[0041] Figure 16 A flowchart of a feasible solution search based on a probabilistic grid graph is provided for one embodiment of the present invention;

[0042] Figure 17 A schematic flowchart illustrating the specific process of determining constraint boundaries is shown.

[0043] Figure 18 An embodiment of the present invention is illustrated schematically. Figure 17 Detailed process flow charts for steps S31 and S32;

[0044] Figure 19 This is a schematic block diagram of a speed planning device according to an embodiment of the present invention;

[0045] Figure 20 This is a schematic block diagram of a speed planning device according to an embodiment of the present invention;

[0046] Figure 21 This is a schematic diagram of an autonomous vehicle according to another embodiment of the present invention;

[0047] Figure 22 This is a schematic diagram of an embodiment of the speed planning device of the present invention. Detailed Implementation

[0048] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0049] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other.

[0050] This invention can be described in the general context of computer-executable instructions, such as program modules, that are executed by a computer. Generally, program modules include routines, programs, objects, elements, data structures, etc., that perform a specific task or implement a specific abstract data type. This invention can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0051] In this invention, terms such as "module," "device," and "system" refer to relevant entities applied to a computer, such as hardware, combinations of hardware and software, software, or software in execution. More specifically, for example, an element can be, but is not limited to, a process running on a processor, a processor, an object, an executable element, an execution thread, a program, and / or a computer. Furthermore, an application program or script running on a server, and the server itself, can also be an element. One or more elements may be in an execution process and / or thread, and elements may be localized on a single computer and / or distributed across two or more computers, and may be run on various computer-readable media. Elements can also communicate via local and / or remote processes based on signals having one or more data packets, for example, signals from data interacting with another element in a local system, a distributed system, and / or interacting with other systems via signals over a network of the Internet.

[0052] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising" or "including" include not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0053] The speed planning method in this embodiment of the invention can be applied to a speed planning device, enabling a user or autonomous driving device to obtain a planned speed curve over a future period of time, and to control the autonomous driving device to perform corresponding driving actions according to the speed curve. These speed planning devices include, but are not limited to, planners on autonomous driving devices, smart tablets, personal PCs, computers, cloud servers, etc. In particular, the speed planning method in this embodiment of the invention can also be applied to autonomous driving vehicles (e.g., passenger cars, buses, minibuses, public buses, trucks, sanitation vehicles, sweepers, floor scrubbers, etc.), robotic vacuum cleaners, motorcycles, electric vehicles, and other autonomous driving devices; this invention does not limit its application to these applications.

[0054] Figure 2The illustration schematically depicts a speed planning method according to an embodiment of the present invention. The execution entity of this method can be a planner or controller on an autonomous vehicle, a processor of a speed planning device such as a smart tablet, personal PC, computer, or cloud server, or a processor of an autonomous driving device such as an unmanned cleaning vehicle, unmanned sweeping vehicle, sweeping robot, or autonomous vehicle. This embodiment of the present invention does not limit this. The present invention will be described in detail using a planner on an autonomous vehicle as the execution entity. Figure 2 As shown, the method of this embodiment of the invention includes:

[0055] Step S100: Obtain the preset obstacle ST risk field and velocity decision results, wherein the obstacle ST risk field is pre-constructed based on multi-frame prediction information of the obstacle;

[0056] Step S200: Determine the constraint boundary of the selected velocity planning model based on the obstacle ST risk field and velocity decision results;

[0057] Step S300: Determine the objective optimization function of the velocity planning model based on the cost function;

[0058] Step S400: Solve the velocity planning model according to the determined constraint boundary and objective optimization function to obtain the velocity planning ST curve.

[0059] In step S100, both the obstacle ST risk field and the speed decision result can be obtained through a front-end module, such as from a controller connected to the planner on the autonomous vehicle, or from another sub-module or another processor of the planner, or from memory. The speed decision result is received from the decision unit connected to the planner on the autonomous vehicle after being processed by the decision unit, so that the planner can perform speed planning based on the pre-built obstacle ST risk field and the speed decision result output by the decision unit.

[0060] In a preferred embodiment of the present invention, the obstacle ST risk field is an ST probability grid map marked with the occupancy probability of each grid point being occupied by an obstacle. The ST probability grid map can be identified by (s, t, P(s,t)), where (s, t) represents a grid point in the ST risk field, and P(s,t) represents the occupancy probability of grid point (s,t) in the ST risk field being occupied by an obstacle. Thus, the obstacle ST risk field of the present invention can effectively describe the relationship between each grid point (s,t) of a dynamic obstacle in the obstacle ST calculation result area and the occupancy probability P. Furthermore, by displaying multi-frame prediction information as the probability of obstacles occupying grid points, the concept of occupancy probability is introduced. This allows the obstacle ST risk field to provide both the accurate location information of the grid points occupied by obstacles and the probability that each occupied grid point is occupied by an obstacle. This achieves the fuzzification of obstacle uncertainty and eliminates the reliance on the accurate predicted trajectory of obstacles to project a single-frame obstacle ST map. It frees speed planning from dependence on the accuracy of the perception and recognition module, improving the stability and comfort of speed planning. The speed planning ST curve planned based on this can effectively avoid problems such as sudden braking caused by unreasonable prediction of obstacle trajectories.

[0061] For example, the obstacle ST risk field in this embodiment of the invention can be specifically implemented as an obstacle ST probability grid map based on the Frenet coordinate system, that is, it is implemented using a grid method based on the Frenet coordinate system (also known as the road coordinate system). Therefore, the obstacle ST risk field in this embodiment of the invention specifically uses a grid to represent the spatial position of obstacles in the grid map, that is, their position in the structured road. Specifically, the obstacle risk field in this embodiment of the invention can represent the relationship between the spatial position of each grid point in the grid map and its corresponding occupancy probability. The specific implementation method for constructing the obstacle ST risk field will be described in detail below.

[0062] Figure 3 This is a schematic diagram of an obstacle risk field environment modeling method according to an embodiment of the present invention, such as... Figure 3 As shown, the method includes:

[0063] S11. Establish the obstacle grid map of the current frame in the Frenet coordinate system with the current position of the vehicle as the origin;

[0064] In this embodiment, the origin of the Frenet coordinate system is set as the current position of the vehicle, that is, the projection point of the vehicle on the center line of the structured road.

[0065] In some implementations, obstacles may include static and dynamic obstacles, and the type of obstacle is included in the perception data. The perception data is output by a known upstream perception module, and a current frame obstacle raster map is established in the Frenet coordinate system with the vehicle's current position as the origin, including:

[0066] Establish a static obstacle SL raster map corresponding to the current frame in the Frenet coordinate system, with the vehicle's current position as the origin; and,

[0067] A current frame ST raster map corresponding to the dynamic obstacle is established in the Frenet coordinate system with the vehicle's current position as the origin.

[0068] Specifically, in the Frenet coordinate system, the centerline of the structured road is used as the reference line (i.e., the vehicle's reference line). Static obstacles are described using S and L parameters to establish an SL grid map, where S is the longitudinal distance (i.e., radial distance) of the static obstacle projected relative to the reference line, and L is the lateral distance (i.e., normal distance) of the static obstacle projected relative to the reference line. Figure 4A The diagram shows the SL coordinate system: 1 represents the vehicle, 2 represents the projection point of a static obstacle onto the reference line (the longitudinal distance of this projection point from the vehicle is s), 3 represents the static obstacle, and 4 represents the reference line. For dynamic obstacles, an ST grid diagram can be created, where T is the obstacle time relative to the current time, and S is the longitudinal distance projected relative to the reference line at time T. Figure 5A The diagram shows the ST coordinate system: 1 represents the vehicle, 2 represents the dynamic obstacle, 3 represents the collision point between the dynamic obstacle and the vehicle, 4 represents the reference line, and 5 represents the predicted trajectory of the dynamic obstacle. This method combines perceived obstacles with a high-precision map.

[0069] S12. Determine the area occupied by the obstacle in the current frame of the obstacle grid map based on the obstacle's current frame perception data;

[0070] The perception data is output from the upstream perception module. The perception module needs to acquire a large amount of environmental information through various sensors, including the vehicle's status, traffic flow information, road conditions, traffic signs, etc. These sensors mainly include: LiDAR, camera, millimeter wave radar, etc.

[0071] Taking static obstacles as an example, Figure 4B This shows the area occupied by a static obstacle in the obstacle raster map of the previous frame, as determined by the perception data of the previous frame. Figure 4C It shows the area occupied by the obstacle in the current frame obstacle raster map, as determined by the current frame perception data of the static obstacle.

[0072] S13. Calculate and update the current frame occupancy probability of each grid cell in the current frame occupancy area of ​​each obstacle in the current frame obstacle grid map to obtain the current frame obstacle risk field.

[0073] For example, the risk fields for static obstacles and dynamic obstacles are respectively represented as SL risk field and ST risk field. The SL risk field mainly describes the relationship between each occupied grid point (s,l) of a static obstacle and its occupancy probability P, denoted as P(s,l). The ST risk field mainly describes the relationship between each occupied grid point (s,t) of a dynamic obstacle and its occupancy probability P, denoted as P(s,t). The static obstacle risk field (i.e., the SL risk field) can be represented as {s,l,P(s,l)}, and the dynamic obstacle risk field (i.e., the ST risk field) can be represented as {s,t,P(s,t)}.

[0074] The embodiments of the present invention establish an obstacle grid map of the current frame based on the Frenet coordinate system. After determining the occupied area of ​​the obstacle in the obstacle grid map of the current frame, the occupancy probability of each grid in the occupied area is calculated. That is, the obstacle risk field established by the technical solution of this application is gridded and refined to the point that each grid has a corresponding occupancy probability, making the obstacle risk field more refined and accurate, and enhancing the effectiveness of the obstacle risk field.

[0075] Before elaborating on the invention in more detail, providing definitions for certain terms used herein may help in understanding the invention.

[0076] In a grid diagram, the area occupied by a static obstacle in the SL grid diagram contains multiple grid points. Uppercase SL represents the set of grid points contained in the area occupied by a single static obstacle, with the subscript representing the obstacle's ID. For example, the area occupied by a static obstacle with ID 5 in the SL grid diagram can be represented as (5S, 5L). Lowercase sl represents a single grid point in SL, with the superscript indicating the grid point number. For example, the 0th grid point in the area occupied by a static obstacle with ID 5 can be represented as (5S, 5L). 0 5s 0 5l). The area occupied by a dynamic obstacle in the ST grid diagram contains multiple grid points. Uppercase ST represents the set of grid points contained in the area occupied by a single dynamic obstacle, with the subscript representing the obstacle's ID. For example, the area occupied by a dynamic obstacle with ID 5 in the ST grid diagram can be represented as (5S, 5T). Lowercase st represents a specific grid point in ST, with the superscript indicating the grid point number. For example, the 0th grid point in the area occupied by a dynamic obstacle with ID 5 can be represented as ( 0 5s 0 5t).

[0077] Since both SL and ST are based on the Frenet coordinate system, with their origin being the projection point of the vehicle on the reference centerline of the structured road, the historical risk field longitudinal coordinate data needs to be continuously updated based on the vehicle's motion. The subscript 'm' represents the time when the Frenet coordinate system for the current frame is established, and the superscript 'n' represents the time when the data for the current frame's coordinate system is established. The data refers to the corresponding 's' and 't' coordinate values ​​of dynamic obstacles calculated based on the current frame's perception data and the vehicle's lane centerline. When the superscript is omitted, it indicates that the time for both coordinate systems is unified; for example, 's2'. 1 s1 represents the projection of the data at time t1 onto the Frenet coordinate system established at time t2, and s2 represents the projection of the data at time t2 onto the Frenet coordinate system established at time t2.

[0078] According to the above definition, the area occupied by a static obstacle with an ID of 5 in the current frame can be represented as (5S) in the Frenet coordinate system of the current frame. m,5 L m ), where the 0th grid point can be represented as ( 0 5s m , 0 5l m ).

[0079] For static obstacles, the maximum value of s is represented as max_s, and the minimum value of s is represented as min_s. The maximum value of l is represented as max_l, and the minimum value of l is represented as min_l.

[0080] For dynamic obstacles, the maximum value of t is denoted as max_t, and the minimum value of t is denoted as min_t. The range of s (min_s, max_s) corresponding to min_t and max_t is different for the same dynamic obstacle in the same time coordinate system. Therefore, they are distinguished by superscript "'", for example, {(min_s, max_s), min_t} and {(min_s', max_s'), max_t}.

[0081] Figure 6 and Figure 7 The implementation methods for determining the currently occupied area of ​​an obstacle in the obstacle grid map of the current frame are illustrated schematically, when the obstacle is a static obstacle and a dynamic obstacle, respectively.

[0082] like Figure 6 As shown, the method for determining the currently occupied area of ​​a static obstacle in the obstacle grid map of the current frame includes:

[0083] S1211: Determine the coverage area of ​​the static obstacle based on its location point and size in the current frame perception data;

[0084] S1212: Project the covered area onto the current frame SL grid to obtain the current frame occupied area of ​​the static obstacle in the current frame SL grid.

[0085] The occupied area of ​​a static obstacle in the current frame's SL grid can be represented as (min_s, max_s) and (min_l, max_l). The occupied obstacle Id of each grid cell is marked in the SL grid. Since the actual positions of obstacles do not overlap, the occupied obstacle Id of each grid cell is theoretically unique.

[0086] It should be noted that for static obstacles, their changes within a structured road are minimal, therefore historical frame information can be disregarded. For example... Figures 4B to 4C As shown, Figure 4B This represents the area occupied by a static obstacle in the previous frame's SL raster map. Figure 4C The area occupied by the static obstacle in the current frame SL raster map.

[0087] like Figure 7 As shown, the method for determining the currently occupied area of ​​a dynamic obstacle in the obstacle grid map of the current frame includes:

[0088] S1221: Determine the ST region where the dynamic obstacle and the vehicle collide based on the current frame predicted trajectory of the dynamic obstacle and the vehicle reference line;

[0089] Based on the current frame predicted trajectory of the dynamic obstacle and the vehicle reference line, the ST region where the dynamic obstacle and the vehicle collide can be represented as {(min_s,max_s),min_t} and {(min_s',max_s'),max_t}. The obstacle Id[] of the occupied grid is marked in the ST grid map. Since the predicted trajectory of the obstacle may overlap, the obstacle Id of each grid is theoretically not unique, so it is denoted as Id[].

[0090] The predicted trajectory of the dynamic obstacle in the current frame is output by a known upstream prediction module. Like the perception module, the prediction module is also an upstream module for environmental modeling in this invention, and this invention does not limit its scope.

[0091] It should be noted that the ST region where the dynamic obstacle collides with the vehicle can be determined according to methods known in the prior art, and the present invention does not limit this. For example, the collision start time t (i.e., min_t) and s (i.e., min_s), and the collision end time t (i.e., max_t) and s (i.e., max_s) are determined based on the current frame predicted trajectory of the dynamic obstacle, the size of the dynamic obstacle, the vehicle's reference line, and the vehicle's size. The ST region where the dynamic obstacle collides with the vehicle is determined based on (min_t, min_s) and (max_t, max_s), and this ST region is not rasterized. For example, a first virtual bounding box representing the dynamic obstacle is generated based on the size of the dynamic obstacle, and a second virtual bounding box representing the vehicle is generated based on the size of the vehicle. Each trajectory point in the predicted trajectory of the dynamic obstacle is used as the center point of the first virtual bounding box, and each waypoint on the reference line of the vehicle is used as the center point of the second virtual bounding box. The first virtual bounding box is simulated to move along the predicted trajectory, and the second virtual bounding box is simulated to move along the reference route. The time points when the collision begins and ends and their corresponding s-coordinate values ​​are recorded.

[0092] S1222: Discretize the ST region raster into the current frame ST raster map to obtain the first estimated occupied region;

[0093] like Figure 5C As shown, the first estimated occupied area is obtained by discretizing the ST region raster into the ST raster map of the current frame.

[0094] S1223: Project the area occupied by the dynamic obstacle in the previous frame onto the current frame ST grid to obtain the second estimated area.

[0095] For example, the area occupied by the dynamic obstacle in the previous frame is displaced according to the time change and the change in the vehicle position between the previous and current frames, so as to obtain the second estimated area of ​​the dynamic obstacle in the ST grid of the current frame.

[0096] For example, the coordinates of the second estimated occupied region obtained by projecting the occupied region of the previous frame onto the ST raster map of the current frame can be expressed as:

[0097] s m =s m m-1 -△s;

[0098] t m =t m m-1 -△t;

[0099] Here, s m s represents the ordinate of the current frame at the corresponding time point. mm-1 It is the calculation result of the previous frame, that is, the vertical coordinate of the corresponding time in the previous frame. △s is the distance the vehicle moves along the path in the current frame and the previous frame, which can be calculated from the vehicle's positioning change. △t is the time difference between the two frames, for example, 0.1 seconds.

[0100] In some implementations, the time interval (i.e., update cycle) between the current frame and the previous frame can be set as needed, for example, determined based on the frequency of hardware such as sensing sensors. This invention does not limit this.

[0101] S1224: Determine the current frame occupied area of ​​the dynamic obstacle in the current frame ST grid map based on the first estimated occupied area and the second estimated occupied area.

[0102] For example, the first and second estimated occupied regions can be filtered to obtain the current frame occupied region of the dynamic obstacle in the current frame ST raster. Filtering can be performed using common filtering methods, such as Bayesian filtering and Kalman filtering. This invention does not limit this approach. Figure 5B and 5C As shown, Figure 5B This represents the area occupied by the dynamic obstacle in the previous frame's ST raster graph. Figure 5C This represents the area occupied by the dynamic obstacle in the current frame's ST raster map. Figure 5B The area occupied by the previous frame is projected onto Figure 5C In the current frame ST raster image, and compare the projected area with... Figure 5C The region occupied by the dynamic obstacle in the current frame is filtered. For simplicity, the region occupied by the dynamic obstacle in the current frame ST raster is not shown in the figure after filtering.

[0103] Alternatively, the first estimated occupied area and the second estimated occupied area can be directly superimposed, that is, the union of the first estimated occupied area and the second estimated occupied area can be taken.

[0104] In some implementations, the method for implementing the risk field of the current frame SL for static obstacles includes:

[0105] For each static obstacle in the current frame occupied area of ​​the current frame SL grid map, the obstacle position confidence in the current frame perception data of the grid is used as the current frame occupancy probability of the grid.

[0106] Therefore, for static obstacles, the occupancy probability of each grid in the current frame can be determined directly based on the perception data.

[0107] In some implementations, the method for determining the ST risk field of the current frame for dynamic obstacles, such as Figure 8 As shown, it includes:

[0108] S131. Calculate and update the current frame predicted trajectory probability of each dynamic obstacle in the current frame occupied area of ​​the current frame ST grid map;

[0109] S132. Determine the current frame occupancy probability of each occupied grid in the current frame ST grid based on the current frame predicted trajectory probability of each dynamic obstacle in the current frame occupied area of ​​the current frame ST grid, so as to obtain the current frame ST risk field.

[0110] For example, further refer to Figure 9 Step S132 can be specifically implemented as follows:

[0111] For each occupied grid cell in the current frame's ST grid map, perform the following steps:

[0112] S1321. Based on the current frame occupied area of ​​each dynamic obstacle in the current frame ST grid, determine at least one target dynamic obstacle corresponding to each occupied grid.

[0113] Because each dynamic obstacle corresponds to a predicted trajectory, and these trajectories may intersect or overlap, some occupied grids in the current frame ST grid may be occupied by multiple dynamic obstacles simultaneously. Therefore, when calculating the current frame estimated occupancy probability of these occupied grids, it is necessary to consider the current frame predicted trajectory probabilities of the corresponding multiple dynamic obstacles in that occupied grid. Thus, it is first necessary to determine which target dynamic obstacles occupy each occupied grid. For example, it can be determined whether each dynamic obstacle contains the occupied grid in the current frame occupied area of ​​the current frame ST grid; if so, the dynamic obstacle is determined to be the target dynamic obstacle occupying the occupied grid.

[0114] S1322. Determine the estimated occupancy probability of the current frame of the occupied grid based on the predicted trajectory probability of each target dynamic obstacle occupying the occupied grid in the current frame and the preset maximum occupancy probability.

[0115] For example, the dynamic obstacle ID is represented by i, the occupied grid is represented by j, and the current frame predicted trajectory probability of a single target dynamic obstacle occupying the occupied grid j is represented as: i P m = i k m The estimated occupancy probability of the current frame for the occupied grid j can be expressed as:

[0116] P m =min(max_p,∑ i k m )

[0117] in i k m The predicted trajectory probability of a target dynamic obstacle with value i in Id[] in the occupied grid j, and max_p is the preset maximum occupancy probability.

[0118] S1323. Calculate the current frame occupancy probability of the occupied grid based on the current frame estimated occupancy probability of the occupied grid and the previous frame occupancy probability of the occupied grid.

[0119] For example, the current frame occupancy probability of an occupied grid can be calculated using a Bayesian filtering formula. The current frame occupancy probability of the occupied grid can be calculated using the following Bayesian filtering formula.

[0120]

[0121]

[0122]

[0123] Where L is an intermediate variable, and according to the physical meaning represented by the obstacle st, when At that time, the obstacle P(S,T) risk field is cleared.

[0124] This allows for the calculation and updating of the risk field P(s,t) for each occupied grid point. A high probability value for P(s,t) indicates high stability of the prediction result in the time dimension and high repeatability in the spatial dimension. This can be used to define a threshold range and determine the risk level of the obstacle's intent and the corresponding predicted trajectory.

[0125] For example, Bayesian filtering can also be replaced by other filtering algorithms, such as Kalman filtering.

[0126] The obstacle risk field calculated and updated by the method of this invention includes two parts: the SL risk field and the ST risk field. It provides accurate location information—the location of the occupied grid point—as well as the probability values ​​P(s,l) and P(s,t) for each grid point. This realizes the establishment of the obstacle risk field environment model, and the autonomous driving system can make decisions and plans based on the obstacle risk fields {s,l,P(s,l)} and {s,t,P(s,t)}.

[0127] Furthermore, from a single-frame perspective, allowing for a certain degree of error within the sensor's accuracy range and transforming the original "there is / is no danger" result description into a probabilistic "0-1" risk description improves the data's applicability. Simultaneously, the superposition of multiple consecutive frames enhances data stability, avoids the impact of single-frame jumps, and improves the accuracy of obstacle risk assessment.

[0128] The technical solution of this invention, when determining the occupied area of ​​a dynamic obstacle in the ST grid, combines the ST area where the dynamic obstacle collides with the vehicle in the current frame and the occupied area of ​​the dynamic obstacle in the previous frame ST grid to determine the current frame occupied area of ​​the dynamic obstacle in the current frame ST grid; and, when calculating the current frame occupied probability of each occupied grid in the current frame ST grid, combines the current frame predicted trajectory probability of multiple target dynamic obstacles occupying that occupied grid and the previous frame occupied probability of the occupied grid to determine the current frame occupied probability of the occupied grid. In other words, the technical solution of this invention can effectively combine the temporal and spatial dimensions of dynamic obstacles simultaneously, improving the stability and accuracy of the dynamic obstacle risk field, and providing accurate environmental model support for more reasonable decision-making and planning.

[0129] It should be noted that the speed decision in step S100 refers to the decision-making process of making speed adjustment or behavioral decisions for specific obstacles or locations by combining historical and traffic regulation information. For example, it refers to making decisions such as "overtaking the obstacle" or "yielding to the obstacle" during vehicle operation. Speed ​​decision provides a clear planning objective for speed planning to improve the smoothness and rationality of vehicle performance. Therefore, in this embodiment of the invention, speed planning is preferably performed based on the speed decision result. In autonomous vehicles, speed decision and speed planning can be implemented by different modules, such as a decision-maker and a planner respectively, or by the same integrated module such as a controller. This embodiment of the invention does not impose any restrictions on this. Preferably, the speed decision result in this embodiment is formed based on the obstacle ST risk field described above. For example, the speed decision result can be a speed decision ST curve composed of multiple discrete ST points. This embodiment of the invention preferably directly utilizes the speed decision result output by the front-end speed decision module. Therefore, the specific implementation method for generating the speed decision result by the front-end speed decision module can refer to relevant existing technologies. The specific implementation method for using the aforementioned obstacle ST risk field to make speed decisions to form the speed decision result in the preferred embodiment of the present invention can be referred to the following description.

[0130] Please refer to Figure 10 This illustrates a velocity decision method based on a probabilistic grid map according to an embodiment of the present invention, the method comprising the following steps:

[0131] Step 101: Establish an ST probability grid map (i.e., ST risk field) in the Frenet coordinate system with the vehicle's position as the origin;

[0132] Step 102: In the ST probability grid map, take the grid point where the vehicle is located as the parent node, and expand multiple grid points that meet the preset conditions according to the search step size; take each expanded grid point as the parent node and continue to expand multiple expanded grid points that meet the preset conditions according to the search step size; search round by round until the preset termination condition is met to obtain multiple paths and their corresponding speed decision results.

[0133] Step 103: For each path, calculate the total cost of the path based on the occupancy probability of each grid point contained in the path and its relative distance to obstacles;

[0134] Step 104: Determine multiple candidate speed decision results based on the total cost corresponding to the multiple paths;

[0135] Step 105: Select the target speed decision result from the multiple candidate speed decision results.

[0136] In this embodiment, for step 101, it is necessary to establish an ST probability grid map in the Frenet coordinate system with the vehicle position as the origin. In the ST probability grid map, the vertical coordinate S is the longitudinal distance of the obstacle relative to the reference path after projection at time T.

[0137] In some embodiments, the aforementioned search step size can be set to, for example, 1 second. The preset conditions can be a series of conditions that ensure the vehicle does not violate traffic rules or engage in risky driving, such as the vehicle not exceeding the speed limit of the current road segment or running red lights. This application does not impose any restrictions on these conditions. The preset termination conditions can be reaching the preset search step size or the search time reaching a preset duration. This application does not impose any restrictions on these conditions. Searching under these preset conditions can ensure safe driving while reducing the number of expanded grid points and limiting the number of paths obtained. Each path corresponds to a speed decision result, which can be, for example, overtaking obstacle A and obstacle B, yielding to obstacle A and obstacle B, or overtaking obstacle A and yielding to obstacle B. The number of obstacles can also be one or more, and the speed decision results can also be other possible combinations. This application does not impose any restrictions on these combinations.

[0138] Then, for step 103, for each path, the total cost of the path is calculated based on the occupancy probability of each grid point in the path and its relative distance to obstacles. Each grid point on each path has an occupancy probability of being occupied by one or more obstacles and a relative distance to each obstacle. The total cost of the path can be calculated based on this occupancy probability and relative distance. For example, for grid point 1, the occupancy probability of being occupied by obstacle A in one search step is 0.4, and the occupancy probability of being occupied by another obstacle B is 0. Similarly, the occupancy probability of each grid point on the path being occupied by an obstacle can be obtained. The total cost can be calculated based on the occupancy probability and relative distance using a potential function, such as a repulsive potential function or an attractive potential function, which is not limited in this application. By introducing occupancy probability and relative distance into the speed decision process, the target speed decision result can avoid grid points with high occupancy probability, stay away from grid points with occupancy probability, and ensure that the vehicle always maintains a safe distance from obstacles.

[0139] Next, in step 104, multiple candidate speed decisions are determined based on the total cost corresponding to the multiple paths. Since each path corresponds to one candidate speed decision, there will be multiple candidate speed decisions determined by multiple total costs. Finally, in step 105, the target speed decision is selected from the multiple candidate speed decisions. The final target speed decision can be selected from the multiple candidate decisions using some filtering methods.

[0140] The method in this embodiment establishes an ST probability grid map, and then uses the occupancy probability and the relative distance between the vehicle and the obstacle in the ST probability grid map during the calculation of the total cost in the expansion process. This allows the speed decision result to better avoid grids with a high occupancy probability and stay as far away from grids with an occupancy probability as possible, so that the vehicle always maintains a safe relative distance from the moving obstacle, thereby ensuring the safety of the final target speed decision result.

[0141] In some optional embodiments, the method further includes: during the round-by-round expansion process, if at least two parent nodes expand to the same expanded grid point, retaining the parent node with the lowest cost. This allows for the removal of some schemes with higher expansion costs during the grid point expansion process, while retaining schemes with lower expansion costs, significantly reducing computational load and improving computational efficiency.

[0142] In some optional embodiments, the preset conditions include speed constraints and collision constraints. This can effectively prevent speeding and collisions with other obstacles. Further optionally, the speed constraint is a maximum speed constraint and / or a maximum acceleration constraint; thus, by constraining the vehicle's maximum speed and maximum acceleration, driving safety can be better guaranteed. For example, the maximum speed constraint and / or maximum acceleration constraint can be set according to the current road speed limit regulations. In other optional embodiments, the collision constraint is prohibiting crossing of risk grid points, where a risk grid point is a grid point with an occupancy probability greater than a preset threshold. For example, the preset threshold can be set to 0.2; when the occupancy probability of a certain grid point is greater than or equal to 0.2, then that grid point is a risk grid point. The preset threshold can also be set to other values, which are not limited in this application.

[0143] Please continue to refer to this. Figure 11 This document illustrates a flowchart of another velocity decision-making method based on a probabilistic grid map, provided by an embodiment of the present invention. This flowchart is primarily aimed at… Figure 10 The flowchart further defines step 103, "Calculate the total cost of the path based on the occupancy probability of each grid point contained in the path and its relative distance to the obstacle".

[0144] like Figure 11 As shown:

[0145] Step 201: Calculate the cost value of each expanded grid point obtained in each round of expansion according to the preset first cost function;

[0146] Step 202: Determine the cost of the last grid point in the path as the total cost of the path.

[0147] In this embodiment, after each round of expansion, the cost of the expanded grid point can be calculated according to a preset first cost function. The first cost function is configured such that the cost of the expanded grid point is the sum of the maximum value of the repulsive potential function of each grid point traversed from the parent node of the expanded grid point to the expanded grid point, and the cost of the parent node of the expanded grid point. The repulsive potential function is related to the occupancy probability of the grid point and its relative distance to obstacles. The cost of subsequent grid points is obtained by superimposing the costs of previous grid points, so that the cost of the last grid point can represent the total cost of the path. Therefore, the cost of each currently expanded grid point can directly represent the total cost of the current path. Specifically, after each round of expansion, the following steps are performed for each expanded grid point in that round: calculating the repulsive potential function value of each grid point traversed between the expanded grid point and its parent node; and using the sum of the maximum repulsive potential function value and the cost value of the parent node of the expanded grid point as the cost of the expanded grid point.

[0148] Continue to refer to Figure 12 This document illustrates a flowchart of another velocity decision-making method based on a probabilistic grid map, provided by an embodiment of the present invention. This flowchart is primarily aimed at… Figure 10 The flowchart further defines step 105, "selecting the target velocity decision result from the plurality of candidate velocity decision results," and includes:

[0149] Step 301: For each candidate speed decision result, calculate the penalty value of the candidate speed decision result according to a preset penalty function, wherein the penalty function is related to whether the candidate speed decision result violates the right-of-way of obstacles and / or disrupts the consistency of historical speed decision results; for example, the more times the right-of-way of obstacles is violated, the higher the penalty value is; the more times the consistency of historical speed decision results is disrupted, the higher the penalty value is.

[0150] Step 302: Select the target speed decision result from the multiple candidate speed decision results based on the penalty value of the multiple candidate speed decision results and the magnitude of s of the corresponding path projection in the ST probability grid.

[0151] In this embodiment, after the speed decision-making device based on the probabilistic grid map calculates the cost of expanding the grid points according to the aforementioned repulsive potential function and obtains multiple candidate speed decision results based on the continuous expansion results, it is still necessary to select the target speed decision result from the multiple feasible decision results. Since the calculation process of the aforementioned multiple candidate speed decision results has already taken into account driving safety, this embodiment mainly considers driving stability and decision consistency, and can conform to safe and civilized driving habits. Therefore, this embodiment mainly calculates the penalty value of the candidate speed decision result according to a preset penalty function, wherein the penalty function is related to whether the candidate speed decision result violates the right-of-way of obstacles and / or disrupts the consistency of historical speed decision results. Finally, by comprehensively considering the penalty value and the magnitude of the vertical coordinate s corresponding to multiple paths in the ST probabilistic grid map of the feasible speed decision result, the final speed decision result is determined.

[0152] In this context, right-of-way is primarily determined by traffic rules requiring vehicles to yield to obstacles with higher right-of-way as much as possible. Consistency determination mainly involves judging whether the current speed decision is consistent with historical speed decisions in the decision pool (which consists of previously executed speed decisions or previous behaviors). For example, for the same obstacle, the decision must be consistent with the most recent speed decision to avoid hesitation and instability in the vehicle's overall performance. Alternatively, it can be set to be consistent with the most recent N decision results for the same obstacle. For instance, if the decision pool shows that the most recent three decisions for obstacle A were to yield, then if the current decision is to yield, it is consistent with previous decisions; if it is to overtake, it is inconsistent with previous decisions. In such cases, a penalty function needs to be used to calculate the penalty value. This application does not impose any restrictions on this.

[0153] In a specific example, regarding consistency, if a historical speed decision is to yield to both obstacles, then if the current decision is to yield to one and then overtake the other, this can be recorded as a violation of consistency, with a penalty value of 1. Overtaking two obstacles can be recorded as a violation of consistency twice. Each change in the decision outcome for each obstacle can be recorded as a violation of consistency, and this applies to multiple obstacles; further details will not be elaborated here. For right-of-way, a violation of right-of-way has a penalty value of 1, which will also not be elaborated here. Right-of-way and consistency can also have corresponding coefficients to make the calculation results more reliable; the determination of these coefficients will not be elaborated here.

[0154] For further options, please refer to Figure 13 It shows a flowchart of a step that further defines the above step 302, "selecting a target speed decision result from the multiple candidate speed decision results based on the penalty value of the multiple candidate speed decision results and the magnitude of s of the corresponding path projected in the ST probability grid."

[0155] like Figure 13 As shown, it includes:

[0156] Step 401: Select the candidate speed decision result with the lowest penalty value;

[0157] Step 402: If there is only one candidate speed decision result with the lowest penalty value, then the candidate speed decision result with the lowest penalty value is determined as the target speed decision result.

[0158] Step 403: If there are multiple candidate speed decision results with the lowest penalty value, select the candidate speed decision result corresponding to the path with the largest vertical coordinate s in the ST probability grid and determine it as the target speed decision result.

[0159] The method in this embodiment mainly represents how to obtain the final target speed decision result based on the penalty value. When there is only one minimum penalty value, the candidate speed decision result corresponding to the minimum penalty value can be directly determined as the final target decision result. When there are multiple minimum penalty values, it is necessary to further consider the size of the vertical axis s in the ST probability grid. The larger the vertical axis s is, the higher the driving efficiency is, and the higher the driving efficiency is, the safer it is.

[0160] Please refer to Figure 14 This document illustrates a flowchart of another velocity decision method based on a probabilistic grid map, provided by an embodiment of the present invention. This flowchart is primarily aimed at… Figure 10 The flowchart further defines step 103, "Calculate the total cost of the path based on the occupancy probability of each grid point contained in the path and its relative distance to obstacles." (See flowchart for example.) Figure 14 As shown, it includes:

[0161] Step 501: Calculate the cost of each expanded grid point obtained in each round of expansion according to a preset second cost function; wherein, the second cost function is configured as a cost function obtained by comprehensively weighting the repulsion potential function, the penalty function for violating the right-of-way of obstacles and / or disrupting consistency, and the vehicle acceleration; the repulsion potential function is related to the occupancy probability of the grid point and its relative distance to obstacles. That is, after each round of expansion is completed, for each expanded grid point in that round of expansion, the following steps are performed: calculate the repulsion potential function value of each grid point traversed between the expanded grid point and its parent node; calculate the penalty value for violating the right-of-way of obstacles and / or disrupting consistency on the path where the expanded grid point is located according to the preset penalty function; and use the maximum value of the repulsion potential function, the penalty value, and the weighted sum of the vehicle acceleration as the cost of the expanded grid point.

[0162] Step 502: Determine the cost of the last grid point on the path as the total cost of the path.

[0163] The second cost function in this embodiment differs from the first cost function in that it is a cost function obtained by comprehensively weighting the repulsive potential function, the right-of-way function for violating obstacles and / or the penalty function for disrupting consistency, and the vehicle acceleration. Therefore, it eliminates the need for separate repulsive potential functions and penalty functions, and instead integrates both functions into the second cost function. The final speed decision result can be obtained using a single cost function, resulting in higher computational efficiency.

[0164] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of combined actions. However, those skilled in the art should understand that the present invention is not limited to the described order of actions, as some steps can be performed in other orders or simultaneously according to the present invention. Secondly, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to the present invention. In the above embodiments, the descriptions of each embodiment have their own emphasis; for parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0165] Furthermore, one formula for calculating the repulsive potential function mentioned in the above embodiments is as follows:

[0166]

[0167] Among them, U req (q) represents the repulsive potential of grid point q in the ST probability grid diagram caused by surrounding obstacles, where η is the repulsive gain, and Q * The effective distance threshold for the obstacle, the effective distance threshold Q * Related to the vehicle's speed, p represents the occupancy probability of grid point q, and d(q) represents the relative distance of grid point q to the vehicle. Therefore, the repulsive potential function is improved by incorporating the probabilities in the ST probability grid. When the relative distance is greater than the effective distance threshold, it indicates that the vehicle maintains a safe driving distance from the obstacle, and the repulsive potential is 0. When the relative distance is less than or equal to the effective distance threshold, a repulsive potential is generated. This repulsive potential is negatively correlated with the relative distance, meaning the smaller the relative distance, the greater the repulsive potential. It is also positively correlated with the occupancy probability, meaning the higher the occupancy probability, the greater the repulsive potential. This allows for better characterization and handling of the repulsive forces between moving obstacles, enabling paths with expansion potential to avoid grids with high occupancy probabilities and to move as far away as possible from grids with occupancy probabilities, thus selecting a safer path for the final speed decision.

[0168] It should be noted that when calculating the repulsive potential function of each grid point, since a grid point may be occupied by different obstacles, meaning the calculated repulsive potential function for that grid point will differ depending on the obstacle, the maximum value of the repulsive potential function at that grid point can be taken when calculating the repulsive potential function. Thus, the value of that grid point as an extended grid point is the maximum value of the repulsive potential function of all obstacles, thereby better characterizing the danger of the grid point and allowing the vehicle to better avoid potentially dangerous grid points. In other solutions, multiple obstacles (all satisfying a relative distance greater than Q) can also be considered. * The value of the repulsive potential function of a grid point is obtained by weighted calculation of the repulsive potential functions of each grid point. This application does not limit this.

[0169] To help those skilled in the art better understand the improvements to the repulsive potential function in this application, the original repulsive potential function is provided for reference: The basic idea of ​​the artificial potential field method is to simulate the repulsive force generated by obstacles on the vehicle's motion. Guided by the potential field, the vehicle avoids the obstacles. The traditional repulsive potential function is:

[0170]

[0171] Where D(q) is the distance between point q and its nearest obstacle, η is the repulsive force gain, and Q * This is the threshold distance at which obstacles exert their influence; obstacles beyond this distance will not have a repulsive effect.

[0172] In the process of developing this application, the inventors discovered that since the probabilistic grid does not provide the precise location of obstacles, but does provide the occupancy probability, the repulsive potential function of the grid point can be calculated from the relative distance to other grid points and the occupancy probability.

[0173] Please refer to Figure 15 This diagram illustrates a speed decision result in an ST probability grid diagram according to an embodiment of the present invention. The following describes the process of calculating the penalty value based on consistency and right-of-way, and determining the final speed decision result, based on this diagram. Figure 15 The horizontal axis in the graph represents time T, and the vertical axis represents time S. The values ​​of 0.2, 0.4, and 0.6 marked within the grid points in the graph represent the probability that an obstacle occupies that grid point in the ST probability grid.

[0174] The multiple feasible decision results generated through the aforementioned expansion steps may contain various topologies, meaning that there are multiple permutations and combinations of decision results for each obstacle. For example... Figure 15 There are three topology choices: ① overtake two obstacles, ② and ③ overtake the first obstacle and yield to the second, and ④ yield to both obstacles. During driving, vehicles typically make the same decision about the same obstacle; otherwise, the vehicle's overall performance will be hesitant and unstable. Simultaneously, according to traffic rules, it's necessary to yield to obstacles with higher right-of-way. Therefore, from all feasible solutions, we need to select the speed decision that best matches historical decisions and aligns with right-of-way requirements. A penalty term for the decision result is designed as follows:

[0175] E = αm + βn,

[0176] Where m and n represent the number of times consistency is broken and right-of-way is violated, respectively, and α and β are the coefficients of the two penalty terms. For example, if the first obstacle has a higher right-of-way than the vehicle and the second obstacle has a lower right-of-way, the decision at the previous time step would be to overtake the first obstacle and yield to the second obstacle. With α = 2 and β = 1, the penalty value for feasible solution ① is 3, the penalty values ​​for ② and ③ are 1, and the penalty value for ④ is 2. Therefore, the topologies corresponding to ② and ③ are chosen. Considering that a larger topology ultimately leads to higher driving efficiency, feasible solution ② is chosen as the final speed decision.

[0177] As a variation, the cost function for calculating the total cost in this embodiment of the invention can be configured to be related to more parameters. For example, it can be related to one or more parameters such as the probability that each grid point is occupied by an obstacle at each time, the relative distance between the vehicle and the obstacle, the vehicle's speed acceleration, whether the vehicle's current speed decision result is consistent with the historical decision results in the decision pool, and the violation of right-of-way. It is understood that the more parameters it is related to, the faster the final speed decision result can be obtained.

[0178] In some optional embodiments, the cost function is a weighted sum of the repulsive potential function, the penalty function for violating obstacle rights and / or consistency, and acceleration, wherein the cost function is positively correlated with the repulsive potential function, the penalty function, and the acceleration. This allows various parameters to be weighted together to form the cost function, which is then used to derive the final decision result.

[0179] Continue to refer to Figure 16 It shows a flowchart of the overall feasible solution search based on a probabilistic grid graph provided by an embodiment of the present invention.

[0180] The ST probability grid reflects the probability that each location on the planned path of the vehicle is occupied at each time point. Grid points with an occupancy probability greater than a preset threshold are considered risky grid points that are prohibited from being traversed. Speed ​​decisions need to avoid all risky grid points while staying as far away as possible from grid points with an occupancy probability. Therefore, the aforementioned potential function is used as the cost function to search for feasible solutions for speed decisions.

[0181] The specific process for searching feasible solutions based on speed decision is as follows: Figure 16As shown. The search step size is a fixed time, such as 1 second. The initial grid point, i.e., the parent node, is (0, 0). Expanding grid points must satisfy the constraints of speed limit and maximum acceleration, and cannot pass through prohibited risk grid points with the parent node. The cost of expanding a grid point is the maximum value of the repulsive potential function among the grid points traversed. After completing the first step of the search, the grid points that can be reached in 1 second can be obtained. Then, these are successively used as parent nodes to continue expanding, which also needs to satisfy kinematic constraints and collision constraints. At this time, the total cost of expanding a grid point is the sum of the cost of the parent node and the maximum value of the potential function among the grid points traversed. If there are other parent nodes that expand to the same grid point and have a smaller total cost, the parent node and total cost of the expanding grid point are updated. This process continues until the search length reaches the planned time, at which point the search stops, and several feasible search results are obtained by reverse searching for parent nodes.

[0182] In developing this application, the inventors discovered that since speed decisions and planning heavily rely on environmental perception and prediction, it is necessary to obtain the true and stable distribution and intentions of other traffic participants. Probabilistic grid maps can effectively combine temporal and spatial information. Therefore, converting the measured location of obstacles into the probability of each location being occupied mitigates the impact of input disturbances to some extent. Consequently, the inventors proposed a speed decision-making method based on probabilistic grid maps. This method searches for several feasible decision results in the probabilistic grid map based on a repulsive field function, reducing the impact of uncertainties in perception and prediction information. Simultaneously, by referencing historical decision results and relative right-of-way relationships, the stability and rationality of the decision results are ensured.

[0183] It should be noted that the ST probability grid diagram established above is the obstacle ST risk field mentioned earlier. The specific method for establishing it can be referred to the previous description, and will not be repeated here.

[0184] As a preferred implementation example, in step S100, determining the constraint boundary of the pre-selected speed planning model based on the obstacle ST risk field and the input speed decision result is implemented by determining the constraint boundary of the pre-selected speed planning model based on the S-axis coordinate value in the obstacle ST risk field, the obstacle occupancy probability of each grid point, and the discrete st points in the speed decision ST curve. Figure 17 The illustration shows the specific implementation method of determining the constraint boundaries of the pre-selected velocity planning model in step S100, such as... Figure 17 As shown, this method can be implemented by including:

[0185] Step S30: Determine the probability threshold p based on the obstacle occupancy probability of each grid point in the obstacle ST risk field. max ;

[0186] Step S31: Starting from the origin of the T-axis of the velocity decision ST curve, and using the directional resolution δt as the time step, traverse the velocity decision ST curve to determine the S-axis coordinate s of the st point corresponding to each time t on the velocity decision ST curve. ref ;

[0187] Step S32: In the obstacle ST risk field, based on the S-axis coordinate value s of the st point corresponding to each time t. ref The S-axis coordinates in the obstacle ST risk field, the obstacle occupancy probability of each grid point, and the probability threshold p. max Determine the constraint boundaries of the pre-selected velocity planning model.

[0188] Since the pre-constructed obstacle ST probability grid map reflects the probability that each grid point on the vehicle's planned path is occupied at each time point, and the higher the probability value of the occupancy probability P(s,t), the higher the stability of the prediction result in the time dimension and the higher the repeatability in the spatial dimension, considering the basic planning goal of the vehicle to avoid collisions with obstacles, when the occupancy probability is greater than a certain threshold, the grid point is considered to be prohibited from crossing. That is, the speed decision needs to avoid all prohibited grids and stay as far away as possible from grids with an occupancy probability. Therefore, the threshold range can be defined according to the probability value of the occupancy probability, and the occupancy probability of prohibited crossing can be determined as the search boundary target for determining the constraint boundary in this embodiment of the invention, that is, the probability threshold p in step S30. max To determine the probability threshold p max The criterion for determining which grid points are prohibited from being traversed is that grid points with an occupation probability less than a probability threshold are considered traversable grid points, while grid points with an occupation probability reaching the probability threshold are considered grid points that are strictly prohibited from being traversed.

[0189] Taking the pre-selected velocity planning model as a polynomial function as an example, the specific processing procedures of steps S31 and S32 will be illustrated below.

[0190] As a preferred implementation example, a fifth-order polynomial can be chosen for velocity planning curve modeling. For example, the velocity planning model is determined by the function s = f(t) = a⁵t. 5 +a4t 4 +a3t 3 +a2t 2 +a1t+a0 represents the equation, where s is the distance traveled along the path in meters (m), t is the travel time in seconds (s), and a0-a5 are the variables to be solved. Accordingly, the constraint boundary determined for this polynomial velocity programming model is exemplarily set to include the first constraint boundary s. max Second constraint boundary s minIf the constraint on this fifth-order polynomial is expressed as s min ≤s≤s max , (0≤t≤T) max The discrete point 'st' in the velocity decision result is labeled as P. ref_i (t i s ref_i ), t i Let s be the T-axis coordinate value of the i-th point s. ref_i Taking the S-axis coordinate of the i-th point as an example, Figure 18 The schematic diagram illustrates the specific processing steps S31 and 32 in this example, as follows: Figure 18 As shown, the speed decision ST curve is traversed to determine the S-axis coordinate s of the st point corresponding to each time t on the speed decision ST curve. ref In the obstacle ST risk field, based on the S-axis coordinate value s of the st point corresponding to each time t... ref The S-axis coordinates in the obstacle ST risk field, the obstacle occupancy probability of each grid point, and the probability threshold p. max The preferred method for determining the constraint boundaries of the pre-selected velocity planning model is as follows:

[0191] In the velocity decision ST curve, starting from the origin of the T-axis (t=0, i=0), the velocity decision curve is traversed using δt as the time step to obtain the t at each time point. i The corresponding point P ref_i (t i S ref_i At each time point ti, the S-axis coordinate s of the corresponding point st is obtained. ref_i In the obstacle ST risk field, the S-axis coordinate value s = s ref_i Starting from the position, search in the positive S-axis direction to determine if there is an obstacle with a probability greater than the probability threshold p. max If a grid point exists, then set the S-axis coordinate value corresponding to that grid point to the current time t, i.e., time t. i The corresponding first constraint boundary s max If it does not exist, then the first constraint boundary s corresponding to the current time t will be... max Set to search path length L pathThe search path length is set to the length of the path output by the previous level planner, i.e., the path planner. It can be directly obtained from the path planner. In this embodiment of the invention, it is preferred to obtain the search path length output by the existing path planner and use it directly. Therefore, the specific implementation method of the path planner generating the search path length can be implemented with reference to the relevant existing technology. The specific implementation method of this embodiment of the invention will not be described in detail.

[0192] After determining the current time t, i.e., time t i After the corresponding first constraint boundary, the S-axis coordinate value s = s in the obstacle ST risk field. ref Starting from the position, search in the negative S-axis direction to determine if there is an obstacle with a probability greater than the probability threshold p. max If a grid point exists, the S-axis coordinate value corresponding to that grid point is set to the second constraint boundary value corresponding to the current time t; otherwise, the second constraint boundary value is set to 0.

[0193] Therefore, by performing the above search process on the obstacle ST risk field for each time point obtained by traversing the velocity decision curve, the first and second constraint boundaries corresponding to each time point can be determined, that is, the constraint boundaries s of the velocity planning model can be determined. min ≤s≤s max , (0≤t≤T) max ).

[0194] It should be noted that, in Figure 18 In the processing shown, for ease of description, the probability of occupancy is greater than the probability threshold p. max The S-axis coordinate value corresponding to the grid point is found in this embodiment of the invention by searching for the occupancy probability greater than the probability threshold p. max The S-axis coordinate value corresponding to the grid point is marked as s map .

[0195] As a preferred implementation, when optimizing the speed planning model, the present invention preferably aims to keep the speed planning ST curve as far away from obstacles as possible. Therefore, in step S300, the selected cost function includes at least a first cost function that keeps the speed planning ST curve defined by the speed planning model as far away from obstacles as possible.

[0196] Preferably, the first cost function is constructed based on the coordinates of each grid point in the obstacle ST risk field and the obstacle occupancy probability. This introduces the concept of occupancy probability into the optimization objective, blurring the precise location of obstacles. This ensures that the planned speed planning ST curve remains consistent across frames, unaffected by fluctuations in obstacle prediction results, thus improving the stability of the planned speed planning ST curve and avoiding adverse effects such as sudden braking due to unreasonable prediction results. For example, the first cost function can be constructed as defined by the following mathematical expression:

[0197] Cost_2=w4*p*∑(f(t)-s i ) 2 ;

[0198] Among them, s i Let be the S-axis coordinate of each grid point in the obstacle risk field, p be the obstacle occupancy probability of the grid point, f(t) be the selected velocity planning model, and w4 be the corresponding weight coefficient.

[0199] In a more preferred implementation, this embodiment of the invention further incorporates an optimization objective of making the speed planning ST curve as smooth as possible when optimizing the speed planning model. Therefore, in step S300, the selected cost function may further include a second cost function that makes the speed planning ST curve defined by the speed planning model as smooth as possible. Exemplarily, the second cost function can be constructed as defined by the following mathematical expression:

[0200] Cost_1 = w1 * ∑f'(t) 2 +w2*∑ f (t) 2 +w3*∑f”'(t) 2 ,

[0201] Where f(t) is the selected velocity planning model, f'(t) is the first derivative of the selected velocity planning model, f"'(t) is the second derivative of the selected velocity planning model, f"'(t) is the third derivative of the selected velocity planning model, and w1-w3 are the corresponding weight coefficients.

[0202] As a preferred implementation, the objective optimization function determined by the cost function in this embodiment of the invention is implemented by simultaneously including a first cost function and a second cost function, so that the optimization objective can simultaneously ensure that the speed planning ST curve is as flat as possible, and that the planned curve is as far away from grid points with a high probability of occupying obstacles as possible. For example, the objective optimization function can be defined by the following mathematical expression:

[0203]

[0204] After determining the constraint boundaries and objective optimization function through the above processing, the pre-selected speed planning model can be solved using the constraint boundaries and objective optimization function. For example, by minimizing the cost of the objective optimization function based on the constraint boundaries, the objective optimization function can be optimized to obtain the values ​​of the variables to be solved in the speed planning model, such as the polynomial a0-a5. Thus, a speed planning model that satisfies the constraint boundaries and optimization objective can be obtained. Based on this model, the corresponding speed planning ST curve can be obtained.

[0205] Since the embodiments of the present invention use the constructed obstacle ST risk field to determine the constraint boundary, and add the optimization objective of staying as far away from the obstacle as possible to the objective optimization function, the dependence of the entire speed planning scheme on the recognition accuracy of the perception module is greatly reduced. The planned speed curve will not be affected by the fluctuation of the obstacle prediction result. The stability and comfort of the planned speed curve are superior. Using the speed planning result to control the driving of the autonomous vehicle can effectively avoid the sudden braking phenomenon caused by the fluctuation of the prediction result and improve the comfort of autonomous driving.

[0206] Figure 19 The speed planning device of one embodiment of the present invention is illustrated schematically, such as... Figure 19 As shown, the device includes:

[0207] The data acquisition module 53 is used to acquire a preset obstacle ST risk field and speed decision results, wherein the obstacle ST risk field is pre-constructed based on multi-frame prediction information of the obstacle, preferably constructed as an ST probability grid map marked with the obstacle occupancy probability of each grid point;

[0208] The constraint determination module 50 is used to determine the constraint boundaries of the pre-selected velocity planning model based on the obstacle ST risk field and velocity decision results;

[0209] The optimization objective determination module 51 is used to determine the objective optimization function of the pre-selected velocity planning model based on the cost function; and

[0210] The curve generation module 52 is used to solve the velocity planning model based on the determined constraint boundaries and objective optimization function to obtain the velocity planning ST curve output.

[0211] In a preferred implementation, the cost function constructed by the optimization objective determination module 51 includes at least a first cost function that makes the speed planning ST curve defined by the speed planning model as far away from the obstacle as possible.

[0212] Preferably, the first cost function is constructed based on the coordinates and occupancy probabilities of each grid point in the obstacle ST risk field.

[0213] For example, the first cost function is constructed as defined by the following mathematical expression:

[0214] Cost_2=w4*p*∑(f(t)-s i ) 2 ;

[0215] Among them, s i Let be the S-axis coordinate of each grid point in the obstacle risk field, p be the obstacle occupancy probability of the grid point, and w4 be the corresponding weight coefficient.

[0216] As another preferred implementation, the cost function constructed by the optimization objective determination module 51 also includes a second cost function that makes the speed planning ST curve defined by the speed planning model as smooth as possible.

[0217] Preferably, the data acquisition module 53 acquires the obstacle ST risk field and velocity decision results from the front-end module or memory. The acquired velocity decision results are a velocity decision ST curve composed of multiple discrete ST points generated based on the obstacle ST risk field.

[0218] The constraint determination module 50 is specifically used to determine the constraint boundary of the pre-selected speed planning model based on the S-axis coordinate value in the obstacle ST risk field, the obstacle occupancy probability of each grid point, and the discrete st points in the speed decision ST curve.

[0219] As a preferred implementation example, the constraint boundaries determined for the planning velocity model include a first constraint boundary and a second constraint boundary.

[0220] It should be noted that the specific implementation process of each module of the speed planning device in this embodiment of the invention can be referred to the description in the method section above, and therefore will not be repeated here.

[0221] Figure 20 A speed planning device according to another embodiment of the present invention is schematically shown in the figure, which is implemented by including:

[0222] Memory 60 is used to store executable instructions; and

[0223] Processor 61 is configured to execute executable instructions stored in memory, which, when executed by the processor, implement the steps of the speed planning method described in any of the preceding embodiments.

[0224] In practice, the speed planning device described above can be applied to autonomous driving equipment such as autonomous vehicles, unmanned cleaners, unmanned sweepers, and robots to perform speed planning for these devices, thereby improving the comfort of autonomous driving. Specifically, the speed planning device can be implemented as a controller or planner in autonomous driving equipment such as autonomous vehicles.

[0225] Figure 21 The diagram schematically illustrates a control device according to one embodiment of the present invention, such as... Figure 21 As shown, the control device includes:

[0226] Planner 70, configured to perform speed planning according to the method of any of the above embodiments; and

[0227] The controller 71 is used to control the speed of the vehicle according to the speed planning ST curve determined by the planner 70.

[0228] In other implementations, the planner and controller in the control device can be integrated into a single control module. Optionally, in practical applications, the control device may also include a perception and recognition module and other planning and control modules, such as a path planning controller, a low-level controller, etc., and the embodiments of the present invention do not impose any limitations on this.

[0229] In other embodiments, the present invention Figure 21 The control device shown can also be applied to vehicles, such as autonomous vehicles, so that the vehicle can have the functions described in the embodiments of the present invention.

[0230] In some embodiments, the present invention provides a non-volatile computer-readable storage medium storing one or more programs including execution instructions, which can be read and executed by an electronic device (including but not limited to a computer, server, or network device, etc.) to perform the speed planning method of any of the above embodiments of the present invention.

[0231] In some embodiments, the present invention also provides a computer program product, the computer program product including a computer program stored on a non-volatile computer-readable storage medium, the computer program including program instructions, which, when executed by a computer, cause the computer to perform the speed planning method of any of the above embodiments.

[0232] In some embodiments, the present invention also provides an electronic device comprising: at least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the speed planning method of any of the above embodiments.

[0233] In some embodiments, the present invention also provides a storage medium having a computer program stored thereon, which, when executed by a processor, implements the speed planning method of any of the above embodiments.

[0234] Figure 22 This is a schematic diagram of the hardware structure of a speed planning device according to another embodiment of this application. The speed planning device can be implemented using the structure shown in the diagram, such as... Figure 22 As shown, the speed planning device includes:

[0235] One or more processors 610 and memory 620, Figure 22 Take the 610 processor as an example.

[0236] The speed planning device may also include an input device 630 and an output device 640.

[0237] The processor 610, memory 620, input device 630, and output device 640 can be connected via a bus or other means. Figure 22 Taking the example of a connection between China and Israel via a bus.

[0238] The memory 620, as a non-volatile computer-readable storage medium, can be used to store non-volatile software programs, non-volatile computer-executable programs, and modules, such as the program instructions / modules corresponding to the speed planning method in the embodiments of this application. The processor 610 executes various functional applications and data processing of the server by running the non-volatile software programs, instructions, and modules stored in the memory 620, thereby implementing the speed planning method of the above-described method embodiments.

[0239] The memory 620 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created according to the speed planning method. Furthermore, the memory 620 may include high-speed random access memory and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other non-volatile solid-state storage device. In some embodiments, the memory 620 may optionally include memory remotely located relative to the processor 610, and these remote memories can be connected to the electronic device via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0240] Input device 630 can receive input digital or character information and generate signals related to user settings and function control of the image processing device. Output device 640 may include a display device such as a display screen.

[0241] The one or more modules are stored in the memory 620, and when executed by the one or more processors 610, they execute the speed planning method in any of the above method embodiments.

[0242] The above-described product can perform the methods provided in the embodiments of this application, and has the corresponding functional modules and beneficial effects for performing the methods. Technical details not described in detail in this embodiment can be found in the methods provided in the embodiments of this application.

[0243] The electronic devices in this application embodiments exist in various forms, including but not limited to:

[0244] (1) Mobile communication devices: These devices are characterized by their mobile communication capabilities and primarily aim to provide voice and data communication. These terminals include: smartphones (e.g., iPhones), multimedia phones, feature phones, and low-end phones, etc.

[0245] (2) Ultra-mobile personal computer devices: These devices fall under the category of personal computers, possessing computing and processing capabilities, and generally also have mobile internet access features. These terminals include PDAs, MIDs, and UMPCs, such as the iPad.

[0246] (3) Portable entertainment devices: These devices can display and play multimedia content. This category includes: audio and video players (such as iPods), handheld game consoles, e-books, as well as smart toys and portable car navigation devices.

[0247] (4) Server: A device that provides computing services. The components of a server include a processor, hard disk, memory, system bus, etc. Servers are similar to general computer architectures, but because they need to provide highly reliable services, they have higher requirements in terms of processing power, stability, reliability, security, scalability, and manageability.

[0248] (5) Other electronic devices with data interaction functions.

[0249] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0250] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented using software plus a general-purpose hardware platform, or of course, using hardware. Based on this understanding, the above technical solutions, in essence or the parts that contribute to the related technology, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0251] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A speed planning method, characterized in that, The method includes: Obtain a preset obstacle ST risk field and velocity decision results, wherein the obstacle ST risk field is an ST probability grid map pre-constructed based on multi-frame prediction information of obstacles; The constraint boundaries of the selected speed planning model are determined based on the obstacle ST risk field and the speed decision results; The objective optimization function of the velocity planning model is determined based on the cost function. The velocity planning model is solved based on the constraint boundary and objective optimization function to obtain the velocity planning ST curve.

2. The method according to claim 1, characterized in that, The cost function includes at least a first cost function that makes the speed planning ST curve defined by the speed planning model as far away from the obstacle as possible.

3. The method according to claim 2, characterized in that, The obstacle ST risk field is constructed as an ST probability grid map marked with the obstacle occupancy probability of each grid point. The first cost function is constructed based on the coordinates of each grid point in the obstacle ST risk field and the obstacle occupancy probability.

4. The method according to claim 2, characterized in that, The cost function also includes a second cost function that makes the speed planning ST curve defined by the speed planning model as smooth as possible.

5. The method according to any one of claims 1 to 4, characterized in that, The speed decision result is a speed decision ST curve composed of multiple discrete ST points, generated based on the obstacle ST risk field. The step of determining the constraint boundary of the selected speed planning model based on the obstacle ST risk field and the speed decision result specifically includes: determining the constraint boundary of the speed planning model based on the S-axis coordinate value in the obstacle ST risk field, the obstacle occupancy probability of each grid point, and the discrete st points in the speed decision ST curve.

6. The method according to claim 5, characterized in that, The step of determining the constraint boundary of the speed planning model based on the S-axis coordinates in the obstacle ST risk field, the obstacle occupancy probability of each grid point, and the discrete st points in the speed decision ST curve specifically includes: The probability threshold p is determined based on the obstacle occupancy probability of each grid point in the obstacle ST risk field. max ; Starting from the origin of the T-axis of the velocity decision ST curve and using the directional resolution δt as the time step, the S-axis coordinate s of the st point corresponding to each time t on the velocity decision ST curve is determined. ref ; In the obstacle ST risk field, based on the S-axis coordinate value s of the st point corresponding to each time t... ref The S-axis coordinates in the obstacle ST risk field, the obstacle occupancy probability of each grid point, and the probability threshold p. max Determine the constraint boundaries of the velocity planning model.

7. The method according to claim 6, characterized in that, The constraint boundary includes a first constraint boundary and a second constraint boundary, and the S-axis coordinate value s of the point st corresponding to each time t is used. ref The S-axis coordinates in the obstacle ST risk field, the obstacle occupancy probability of each grid point, and the probability threshold p. max Determining the constraint boundaries of the velocity planning model specifically includes: The S-axis coordinate value s=s in the obstacle ST risk field ref Starting from the position, search in the positive S-axis direction to determine if there is an obstacle with a probability greater than the probability threshold p. max If a grid point exists, its corresponding S-axis coordinate is set as the first constraint boundary at the current time t; otherwise, the first constraint boundary at the current time t is set as the search path length L. path ; After determining the first constraint boundary corresponding to the current time t, the S-axis coordinate value s = s in the risk field of the obstacle ST. ref Starting from the position, search in the negative S-axis direction to determine if there is an obstacle with a probability greater than the probability threshold p. max If a grid point exists, the S-axis coordinate value corresponding to that grid point is set to the second constraint boundary value corresponding to the current time t; otherwise, the second constraint boundary value is set to 0.

8. A speed planning device, characterized in that, The device includes: The data acquisition module is used to acquire the preset obstacle ST risk field and velocity decision results, wherein the obstacle ST risk field is an ST probability grid map pre-constructed based on multi-frame prediction information of the obstacle; The constraint determination module is used to determine the constraint boundary of the selected speed planning model based on the obstacle ST risk field and the speed decision result; The optimization objective determination module is used to determine the objective optimization function of the velocity planning model based on the cost function; and The curve generation module is used to solve the velocity planning model based on the constraint boundary and the objective optimization function to obtain the velocity planning ST curve output.

9. A speed planning device, characterized in that, include: Memory, used to store executable instructions; as well as A processor for executing executable instructions stored in memory, which, when executed by the processor, implement the steps of the method according to any one of claims 1 to 7.

10. A control device, characterized in that, include: A planner is configured to perform speed planning according to any one of claims 1 to 7 and output a speed planning ST curve to the controller; A controller is used to control the speed of the vehicle according to the speed planning ST curve.

11. A vehicle, characterized in that, Includes the control device as described in claim 10.

12. A storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the program implements the steps of the method described in any one of claims 1-7.

13. A computer program product comprising a computer program stored on a non-volatile computer-readable storage medium, the computer program comprising program instructions that, when executed by a computer, cause the computer to perform the method of any one of claims 1-7.

Images