Method for Planning Driving Route of Autonomous Driving Vehicle Using High-Precision Map

Abstract

This application relates to the technical field of route planning, specifically to a method for planning the driving route of an autonomous vehicle using a high-precision map. The method includes: obtaining high-precision map data at each moment and a preset number of moments before, including parameter data of each node and section on the high-precision map at each moment; obtaining the obstacle influence cost of each section according to the total number of obstacles in each section and the position distribution characteristics between the obstacles; obtaining the special road condition cost of each section according to the distribution characteristics of the traffic flow data before each moment; obtaining the passing cost when the autonomous vehicle passes through each section, combining the distance from each node of the autonomous vehicle to the vehicle driving end point, obtaining the cost function of the A* algorithm, and using the A* algorithm to obtain the final path planning result. This application aims to make the driving planned route of the autonomous vehicle reach the optimal state.

Description

Technical Field

[0001] This application relates to the technical field of route planning, and particularly to a method for planning the driving route of an autonomous vehicle using a high-precision map. Background Art

[0002] In recent years, as an integrated application system that deeply integrates cutting-edge technologies such as communication technology, control technology, and artificial intelligence, the intelligent transportation system has become a research focus that has received wide attention. Among them, autonomous driving, as an important part of the intelligent transportation system, related technologies have become new research hotspots. Among them, route planning, as a part of the decision-making module of autonomous driving, has an important impact on the driving safety and driving efficiency of autonomous vehicles.

[0003] The literature "Chen Tao, Li Chengxin, Zhao Chongyang. Research on Route Planning Strategies for Autonomous Vehicles Considering Safety and Efficiency [J]. Journal of Transportation Science and Engineering, 2024, 40(06): 127-134" mentions that by determining the starting point and the ending point to generate a target route and introducing different cost conditions, the path search algorithm can simulate different types of routes. However, the routes formulated by a single cost condition often fail to reach the optimal state. Therefore, many researchers have established a multi-cost-based path planning method by analyzing factors such as the shortest route, vehicle density on the road, and probability of failure, and then calculating the route cost.

[0004] However, there are still deficiencies in the planning and adjustment of the route of autonomous vehicles using the above methods. Currently, the range of sensor data acquisition on autonomous vehicles is limited, and thus the range of route adjustment is limited. Therefore, the current route planning of autonomous vehicles mainly relies on high-precision maps. However, due to the certain lag of the data on high-precision maps, the road conditions may change significantly after the route is planned, resulting in the route being difficult to reach the optimal state. Summary of the Invention

[0005] In view of the above, it is necessary to provide a method for planning the driving route of an autonomous vehicle using a high-precision map to solve the above problems.

[0006] An embodiment of this application provides a method for planning the driving route of an autonomous vehicle using a high-precision map, and the method includes:

[0007] Obtain high-precision map data at each moment and within a preset number of moments before, including parameter data of each node and section on the high-precision map at each moment; wherein, the parameter data includes the number of obstacles, the positions of obstacles, and traffic flow data for each section.

[0008] According to the total number of obstacles in each section and the positional distribution characteristics between obstacles when the autonomous driving vehicle starts to pass through each section, the obstacle impact cost when the autonomous driving vehicle passes through each section is obtained; according to the distribution characteristics of the traffic flow data of each section before each moment, the special road condition cost of each section is obtained.

[0009] Based on the obstacle impact cost and the special road condition cost of each section, the passing cost when the autonomous driving vehicle passes through each section is obtained; based on the passing cost of the sections between the driving starting point and each node of the autonomous driving vehicle, combined with the distances from each node to the vehicle driving end point, the cost function of the Astar algorithm is obtained, and the final path planning result is obtained by using the Astar algorithm.

[0010] Among them, an intersection on the high-precision map at each moment where the number of selectable paths is greater than or equal to the preset value is called a node.

[0011] Among them, the only path between one node and another node is used as a section.

[0012] Among them, the specific method for obtaining the obstacle impact cost when the autonomous driving vehicle passes through each section is as follows:

[0013] For each section, calculate the degree of dispersion of the distance metrics between all pairwise combinations of obstacles in each section;

[0014] The positive fusion of the negative correlation mapping result of the degree of dispersion and the total number of obstacles in each section is performed to obtain the obstacle impact cost when the autonomous driving vehicle passes through each section.

[0015] Among them, the specific method for obtaining the special road condition cost of each section is as follows:

[0016] Calculate the mean value and the extreme difference value of all traffic flow data of each section before each moment, and use the negative correlation mapping result after the positive fusion of the mean value and the extreme difference value as the special road condition cost of each section at each moment.

[0017] Among them, the mathematical relationship of the special road condition cost is: ; where is the special road condition cost of the i-th section, is the mean value of the traffic flow data of the i-th section before the Z-th moment, is the extreme difference value of all traffic flow data of the i-th section before the Z-th moment.

[0018] Among them, the passing cost when the autonomous driving vehicle passes through each section is determined by the positive fusion result of the obstacle impact cost and the special road condition cost of the corresponding section.

[0019] Among them, the formula for obtaining the cost function of the A* algorithm is as follows: ; In the formula, represents the cost from the driving starting point to the j-th node; represents the actual cost from the driving starting point to the j-th node; represents the estimated cost from the j-th node to the driving end point.

[0020] Among them, the actual cost is specifically the sum of the passing costs of all road sections passed by the autonomous driving vehicle from the driving starting point to the j-th node.

[0021] Among them, the estimated cost is specifically the Euclidean distance from the j-th node to the driving end point.

[0022] The present application has at least the following beneficial effects:

[0023] By analyzing the number and position changes of obstacles in a single road section during the route planning process of an autonomous driving vehicle using a high-precision map, the present application obtains the obstacle impact cost of each road section, which helps to determine whether the obstacles in each road section will affect the normal operation of the autonomous driving vehicle; further, considering that there is a certain lag in the update of the road conditions of the high-precision map, according to the distribution characteristics of the traffic flow data of each road section before each moment, the special road condition cost of each road section is obtained. Through in-depth analysis of historical traffic flow data, the traffic patterns of certain road sections during specific time periods can be identified; based on the obstacle impact cost and the special road condition cost of each road section, the passing cost of the autonomous driving vehicle when passing through each road section is obtained. The beneficial effect is that it comprehensively considers the driving costs of each road section in various situations; based on the passing costs of the road sections between the driving starting point and each node of the autonomous driving vehicle, combined with the distances from each node to the driving end point of the vehicle, the cost function of the A* algorithm is obtained, thereby enabling the route planned by the autonomous driving vehicle to reach the optimal state. This solves the problem that when an autonomous driving vehicle uses a high-precision map for path planning, due to the certain lag of the data on the high-precision map, the road conditions may change greatly after the route planning is completed, resulting in the route being difficult to reach the optimal state, and can greatly improve the adaptability of the autonomous driving vehicle in the road environment, so that the path planning always maintains the optimal state. Description of the Drawings

[0024] Figure 1 is a flowchart of the method for planning the driving route of an autonomous driving vehicle using a high-precision map provided by the present application;

[0025] Figure 2 is a schematic diagram of nodes and road sections provided by the present application;

[0026] Figure 3Schematic diagram for obtaining the cost function provided by this application. Detailed implementation manners

[0027] In the description of the embodiments of this application, words such as "exemplary", "or", "for example", etc. are used to indicate examples, illustrations or explanations. Any embodiment or design solution described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or having more advantages than other embodiments or design solutions. Exactly speaking, using words such as "exemplary", "or", "for example", etc. aims to present relevant concepts in a specific way.

[0028] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which this application belongs. The terms used in the description of this application are only for the purpose of describing specific embodiments and are not intended to limit this application.

[0029] In addition, it should be noted that the terms "first" and "second" in this application and its drawings are used to distinguish similar objects and are not used to describe a specific order or sequence. For the methods disclosed in the embodiments of this application or the methods shown in the flowcharts, including one or more steps for implementing the methods, without departing from the scope of protection of this application, the execution order of multiple steps can be interchanged with each other, and some steps can also be deleted.

[0030] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which this application belongs.

[0031] This application proposes a method for planning the driving route of an autonomous vehicle using a high-precision map, which is applied to the technical field of route planning. Referring to the attached Figure 1 , the method includes the following steps:

[0032] S1: Obtain high-precision map data at each moment and within a preset number of moments before, including parameter data of each node and section on the high-precision map at each moment; wherein, the parameter data includes data such as the number of obstacles, obstacle positions, and traffic flow on each section.

[0033] At time Z, the autonomous vehicle is ready to go to the destination. The autonomous vehicle obtains the map data at each update within the past T hours including time Z through the route planning system, and obtains the obstacle number data, position data, and traffic flow data of each section therein, and preprocesses the data.

[0034] In this application, T is a preset quantity, and its value range is 24 to 168. In this embodiment, T is set to 168, and the implementer can select according to the actual situation. The data preprocessing adopts the method of mean filling, and the implementer can adopt other data preprocessing algorithms on the premise of ensuring data integrity.

[0035] It should be noted that the way of dividing road segments in the high-precision map is as follows: An intersection with the number of selectable paths greater than or equal to the preset value at a single intersection is called a node; The only direct connection path from one node to another node is used as a road segment. In this embodiment, the preset value is set to 3, and the selectable path refers to the path corresponding to different driving directions that can be selected starting from an intersection. Among them, the schematic diagrams of nodes and road segments are as Figure 2 shown Figure 2 In, the number of selectable paths corresponding to points P and Q are 3 and 4 respectively. Therefore, both P and Q are nodes, and the path from node P to node Q is used as road segment PQ.

[0036] S2: According to the total number of obstacles and the position distribution characteristics of obstacles on each road segment when the autonomous vehicle starts to travel on each road segment, obtain the obstacle influence cost when the autonomous vehicle travels on each road segment; According to the distribution characteristics of traffic flow data on each road segment before each moment, obtain the special road condition cost of each road segment.

[0037] Under normal circumstances, there are fewer obstacles on the urban roads where vehicles travel. Even if there are obstacles, they are usually fixed and immovable obstacles, and such obstacles usually have less impact on route planning. However, the route that the vehicle travels is not always an urban road, and it may also be a block road. There are many obstacles such as small stalls, parked non-motor vehicles and motor vehicles on the block road. Generally speaking, the more obstacles there are, the greater the impact on the vehicle's passing efficiency. Therefore, when planning through traditional route planning algorithms, the road cost calculated for road segments with more obstacles is usually higher, and thus this road will not be selected as the route planning road.

[0038] However, in fact, it is not the case that the more obstacles there are, the greater the impact on the vehicle's passing efficiency. If the positions of the obstacles are messy and the regularity is low, then the impact on the vehicle's passing efficiency is greater. When the regularity degree of the obstacle positions is relatively high, such as parked vehicles are all within the specified parking spaces, and small stalls and street sundries are all placed on both sides of the road, in this case, even if there are many obstacles, in fact, the impact on the passing efficiency of passing vehicles is relatively low.

[0039] In summary, if the obstacles in a road section are distributed in a chaotic manner and the distance between obstacles varies greatly, it indicates that the obstacles in this road section have a greater impact on the vehicle passing efficiency. On the contrary, if the obstacle distribution is relatively regular and the distance change between obstacles is small, the impact on the vehicle passing efficiency is smaller, and the cost incurred by the autonomous vehicle when passing through this road section is lower. In order to reflect the impact of the above obstacle distribution on the vehicle passing efficiency, the obstacle impact cost when the vehicle passes through each road section is now obtained: for each road section, calculate the dispersion degree of the distance metrics between all pairwise combinations of obstacles in each road section; positively fuse the negative correlation mapping result of the dispersion degree with the total number of obstacles in each road section to obtain the obstacle impact cost of the vehicle in each road section.

[0040] Taking the i-th road section as an example in this embodiment, the mathematical relationship of the obstacle impact cost is as follows: ; where is the obstacle impact cost when the autonomous vehicle passes through the i-th road section, is the total number of obstacles in the i-th road section, is the variance of the Euclidean distance between obstacles in the i-th road section, which is used to measure the dispersion degree of the distance between obstacles; represents a preset parameter greater than zero, which is used to prevent the denominator from being 0, and its value is 0.01 in this embodiment.

[0041] Generally, during the route planning process of a driverless vehicle, the cost of road sections with low traffic flow is usually small, so these road sections will be selected for route planning. However, during actual driving, it is not necessarily the case that the vehicle passing efficiency of road sections with low traffic flow is good. It may also be because there are sudden situations in this road section, such as vehicle accidents, temporary road construction, etc., or it may be because the road surface condition of this road is poor, such as there being a lot of sand and gravel on the road surface, or the road surface being uneven, etc., which leads to most vehicles not choosing to pass through this road. However, due to the certain lag in the update of the road condition information in the high-precision map, it cannot be guaranteed that the above special road condition information can be obtained and updated immediately during route planning, and thus it is considered that the traffic flow of this road section is small, making it think that the passing efficiency of this road is high during the route planning process, and thus planning it into the route, resulting in the finally planned route still not being able to reach the optimal state. Therefore, in order to avoid this situation, further analysis is required.

[0042] For a normal road, the volume of traffic has certain regularity, such as rush hours during commuting times and peak travel periods on holidays. Therefore, whether the traffic volume on a normal road is high or low, it will change after only maintaining for a period of time, and the overall traffic volume data changes greatly. For a road with special road conditions, vehicles cannot pass or can only pass with difficulty, which leads to extremely low traffic volume on this road for a long time and small changes, and the overall traffic volume data changes little. Based on this, according to the distribution characteristics of the traffic volume data of each road section before each moment, the special road condition cost of each road section is obtained: calculate the mean value and the range value of all traffic volume data of each road section before each moment, and take the negative correlation mapping result after the positive fusion of the mean value and the range value as the special road condition cost of each road section at each moment.

[0043] In this embodiment, taking the i-th road section as an example, the mathematical relationship of its special road condition cost is: ; where is the special road condition cost of the i-th road section, is the mean value of the traffic volume data of the i-th road section before the Z-th moment, is the range value of all traffic volume data of the i-th road section before the Z-th moment.

[0044] S3: Based on the obstacle impact cost and the special road condition cost of each road section, obtain the passing cost when the autonomous vehicle passes through each road section; based on the passing cost of the road sections between the driving starting point and each node of the autonomous vehicle, combined with the distances from each node to the vehicle driving end point of the autonomous vehicle, obtain the cost function of the Astar algorithm, and use the Astar algorithm to obtain the final path planning result.

[0045] The passing cost of the autonomous vehicle on each road section can be determined by the positive fusion result of the obstacle impact cost and the special road condition cost of the corresponding road section. In this embodiment, the positive fusion between multiple variables is calculated by using the sum value method.

[0046] It should be understood that when the autonomous vehicle conducts path planning, if the number of obstacles on this road section is larger, the distribution is more chaotic, or the traffic volume on this road section is extremely low for a long time, it means that the passing efficiency of the vehicle on this road section is lower, the passing cost required is greater, and this road section should not be planned into the driving route.

[0047] In the route planning system of the autonomous vehicle, first, according to the starting point and the end point determined by the user, each road section in the high-precision map is divided, then the passing cost of each road section is calculated, and the cost function of the Astar algorithm is improved by using the passing cost. The improved cost function is: ; In the formula, Denote the cost from the vehicle driving starting point to the j-th node; Denote the actual cost from the vehicle driving starting point to the j-th node, and its value is specifically the sum of the passing costs of all road sections passed by the autonomous vehicle from the driving starting point to the j-th node; Denote the estimated cost from the j-th node to the vehicle driving end point, which is specifically the Euclidean distance from the j-th node to the driving end point. Finally, the Astar algorithm is used for route planning, and the optimal route including road section information is output; among them, the specific steps of the Astar algorithm are existing well-known technologies, and the present application will not elaborate on this.

[0048] Among them, the schematic diagram for obtaining the cost function is as Figure 3 shown.

[0049] The present application builds a simulation road network according to the high-precision map in a certain urban area through SUMO software, and a total of 50 test nodes and 62 test road sections are set. The road network contains the number information, position information and traffic flow information of each road obstacle. Under the same road network environment, simulation tests are respectively carried out on the traditional route planning strategy (set as Route 1 in this scheme, giving priority to passing efficiency) and the route planning strategy considering the influence of obstacle position rules and the cost of special road conditions (set as Route 2 in this scheme, considering both the influence of obstacle position rules and special road conditions). The costs used in the route planning simulation results are shown in the following table:

[0050] Table 1: Actual cost of the route

[0051]

[0052] In the above table, the passing time cost is specifically the ratio of the length of each road section to its corresponding maximum speed limit of the road section; the comprehensive passing cost is specifically the mean value of the obstacle influence cost, the special road condition cost and the passing time cost.

[0053] It can be concluded from the simulation results of the route planning that there are significant differences in the route costs between the starting point and the destination. Compared with the traditional route, the time-consuming cost of the route with the minimum comprehensive passing cost is reduced by 18.49%. The simulation results further prove the effectiveness of the route planning strategy proposed by the present application, which not only avoids the lag in route planning using high-precision maps, but also can plan a route with low time-consuming cost and high passing efficiency.

[0054] The present application provides a method for planning a driving route of an autonomous vehicle using a high-precision map. The method includes: by analyzing the number and position changes of obstacles in a single section during the route planning process of the autonomous vehicle using the high-precision map, the obstacle impact cost of each section is obtained, which helps to determine whether the obstacles in each section will affect the normal operation of the autonomous vehicle; further, considering that there is a certain lag in the update of the road conditions of the high-precision map, according to the distribution characteristics of the traffic flow data of each section before each moment, the special road condition cost of each section is obtained. Through in-depth analysis of historical traffic flow data, the traffic patterns of certain sections during specific time periods can be identified; based on the obstacle impact cost and the special road condition cost of each section, the passing cost of the autonomous vehicle when passing through each section is obtained. The beneficial effect is that the driving costs of each section in various situations are integrated; based on the passing costs of the sections between the driving starting point of the autonomous vehicle and each node, combined with the distances from each node to the vehicle driving end point, the cost function of the A* algorithm is obtained, so that the route planned by the autonomous vehicle reaches the optimal state. This solves the problem that when the autonomous vehicle uses the high-precision map for path planning, due to the certain lag of the data on the high-precision map, the road conditions may change greatly after the route planning is completed, resulting in the route being difficult to reach the optimal state. It can greatly improve the adaptability of the autonomous vehicle in the road environment, so that the path planning always remains in the optimal state.

[0055] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architectures, functions, and operations of systems, methods, and computer program products according to embodiments of the present application. In this regard, each block in the flowchart or block diagram may represent a module, a program segment, or a part of code, and the module, program segment, or part of code contains one or more executable instructions for implementing the specified logical function. In some alternative implementations, the functions marked in the block may occur in a different order than marked in the accompanying drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in the reverse order, depending on the functions involved. In the descriptions corresponding to the flowcharts and block diagrams in the accompanying drawings, the operations or steps corresponding to different blocks may also occur in a different order than disclosed in the description. Sometimes, there is no specific order between different operations or steps. For example, two consecutive operations or steps may actually be executed substantially in parallel, and they may sometimes be executed in the reverse order, depending on the functions involved. Each block in the block diagram and / or flowchart, and the combination of blocks in the block diagram and / or flowchart, can be implemented by a dedicated hardware-based system for performing the specified functions or actions, or can be implemented by a combination of dedicated hardware and computer instructions.

[0056] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the various embodiments of the present application, and should all be included within the protection scope of the present application.

Claims

1. A method for planning a driving route for an autonomous driving vehicle using a high-precision map, characterized in that: The method comprises the following steps: Obtaining high-precision map data at each time and a preset number of times before, including parameter data of each node and road section on the high-precision map at each time; wherein the parameter data includes the number of obstacles, obstacle locations and traffic flow data of each road section; According to the negative correlation mapping result of the discrete degree of distance measurement between all the obstacles in each road section when the autonomous driving vehicle starts to pass through each road section, the total number of obstacles in each road section is positively fused to obtain the obstacle impact cost of the autonomous driving vehicle when passing through each road section; according to the mean and range value of all traffic flow data of each road section before each moment, the negative correlation mapping result after positive fusion of the mean and the range value is used as the special road condition cost of each road section at each moment; Based on the obstacle impact cost and the special road condition cost of each road section, the travel cost of the autonomous driving vehicle when passing through each road section is obtained; based on the travel cost of the road section from the starting point of the autonomous driving vehicle to each node, combined with the distance from each node to the end point of the vehicle, the cost function of the Astar algorithm is obtained, and the Astar algorithm is used to obtain the final path planning result.

2. The method for planning a driving route for an autonomous driving vehicle using a high-precision map according to claim 1, wherein: An intersection on the high-precision map at each moment where the selectable path at a single intersection is greater than or equal to a preset value is called a node.

3. The method for planning a driving route for an autonomous driving vehicle using a high-precision map according to claim 1, wherein: The only path from one node to another is called a road segment.

4. The method for planning a driving route for an autonomous driving vehicle using a high-precision map according to claim 1, wherein: The mathematical relationship of the special road condition cost is: ;in, is the special road condition cost of the i-th road section, is the mean value of traffic flow data of the ith road section before the Zth moment, is the extreme value of all traffic flow data of the ith road section before the Zth moment.

5. The method for autonomous driving vehicle route planning using a high-precision map according to claim 1, wherein: The travel cost of the autonomous driving vehicle when passing through each road section is determined by the forward fusion result of the obstacle impact cost and the special road condition cost of the corresponding road section.

6. The method for planning a route for an autonomous driving vehicle using a high-precision map according to claim 1, wherein: The formula for obtaining the cost function of the Astar algorithm is: ; In the formula, represents the cost from the starting point to the jth node; represents the actual cost from the starting point to the jth node; Represents the estimated cost from the jth node to the destination.

7. The method for planning a driving route for an autonomous driving vehicle using a high-precision map according to claim 6, wherein: The actual cost is specifically the sum of the travel costs of all road sections that the autonomous driving vehicle passes through from the starting point to the jth node.

8. The method for planning a driving route for an autonomous driving vehicle using a high-precision map according to claim 6, wherein: The estimated cost is specifically the Euclidean distance from the jth node to the travel end point.

Images