Mobile robot continuous space path planning method for narrow roadway

Abstract

The invention discloses a narrow roadway-oriented mobile robot continuous space path planning method, and belongs to the field of mobile robot path planning. Framing a feasible range of the narrow roadway based on the model, wherein the range is a quadrangle; determining two short edges according to the four vertexes of the quadrangle to generate a central axis, setting four transition points between the central axis and the long edges on the short edges, and calculating an inclination angle according to the two transition points on the same side of the central axis; judging the relationship between the yaw angle and the inclination angle when entering the narrow roadway based on a driving rule, and inserting two transition points corresponding to the inclination angle into a task sequence when the condition of entering the narrow roadway is met based on the relationship; a conflict-free path which passes through two transition points in a task sequence from an initial position and finally reaches a target position is planned based on a multi-mobile-robot path planning algorithm of a continuous space.

Description

Technical Field

[0001] The invention belongs to the field of mobile robot path planning, and in particular relates to a mobile robot continuous space path planning method for narrow lanes. Background Art

[0002] With the rapid development of artificial intelligence technology, robots have shown strong advantages in various fields. Mobile robots have become important transportation tools in logistics warehousing, flexible manufacturing, container terminals and other fields with high efficiency, low cost, high stability, accuracy and flexibility, and are widely favored by enterprises. Path planning, as a key underlying technology for controlling robots from the starting point to the target position, has prompted researchers to conduct in-depth research on the Multi-Agent Path Finding (MAPF) problem.

[0003] The task of the MAPF problem is to plan paths for multiple agents. The key constraint is that all agents can follow the paths obtained by the strategy at the same time without colliding with each other. The early research on the MAPF algorithm mainly focused on the improvement of classic path planning algorithms such as Dijkstra algorithm, A* algorithm, D*Lite algorithm, artificial potential field algorithm, heuristic swarm optimization algorithm, and reinforcement learning. With the deepening of research, optimal solvers specifically for MAPF problems have been developed, including optimal solvers based on A*, based on reduction, based on cost growth tree, and based on conflict search. At the same time, many suboptimal solvers with higher solving efficiency have also been extended, including enhanced CBS algorithm (ECBS), explicit estimated CBS algorithm (EECBS), priority search-based solver (PBS), and multi-agent path planning algorithm based on large neighborhood search (MAPF-LNS2). These algorithms sacrifice optimality in exchange for solving speed, which can better meet the real-time requirements of actual industrial scenarios, and have better solving effects when facing some special scenarios. As one of the mainstreams, the priority search-based suboptimal solver (PBS) is a two-layer search algorithm. The high-level search is responsible for conflict detection, building a priority binary tree and executing the best priority local search strategy. The low-level search uses a single-agent path planning algorithm to plan the shortest path that meets the priority for each agent. In order to meet the multi-agent path planning in continuous space, the kinematic constraints and geometric shape constraints of the agent need to be considered. Someone proposed a Car-like CBS algorithm in continuous space. This algorithm solves the conflict-free path planning of agents that conform to the Ackerman model in free space, taking into account the shape characteristics of the agent, and the planned path trajectory does not need to be optimized twice, and can be directly applied to actual scenarios. Some people replaced the CBS algorithm with the PBS algorithm based on the above algorithm to improve the overall solution speed of the algorithm. The Ackerman model is a geometric model used to describe the steering characteristics of a car. It is based on the Ackerman steering geometry principle, that is, when a car turns, the turning radius of the inner tire is smaller than the turning radius of the outer tire to achieve more stable driving.

[0004] Logistics warehousing is one of the most important application scenarios of mobile robots. Shelves are usually placed according to the optimized layout of the storage area. In order to maximize the use of limited space, shelves are often arranged densely, which greatly reduces the range of feasible paths for mobile robots and often faces conflicts with opposite directions in narrow lanes. The above algorithms are prone to deadlock when facing this special problem. The motion units of mobile robots in the Ackerman kinematic model are usually fixed to seven types: forward left turn, forward straight, forward right turn, backward left turn, backward straight, backward right turn and waiting in place. However, in narrow lanes, the space available for robot movement is extremely limited, and these predefined motion units are often unable to be fully executed, resulting in the inability to avoid oncoming vehicles. In actual planning, this limitation often leads to conflicts between two mobile robots in the middle of narrow lanes. The solution strategy of some algorithms is to make one of the robots back out of the lane to make way for the other, while other algorithms cannot provide effective solutions due to the deadlock problem. Someone analyzed the "selfish" behavior of the PBS algorithm when resolving conflicts, that is, the low-priority agent unconditionally avoids the high-priority agent, and proposed an adversarial conflict resolution strategy. By scaling the geometric constraints of the agents, a more selfless conflict resolution method between agents can be achieved. This method requires continuous iteration of the scaling factor until the original geometric constraints are restored. The setting of the initial value of the scaling factor needs to be considered, and the number of iterations will also affect the overall solution speed. Summary of the invention

[0005] In order to solve the shortcomings of the prior art, reduce the complexity of path planning, improve the speed of path planning, and ensure the success rate of path planning, the present invention adopts the following technical solutions:

[0006] A continuous space path planning method for a mobile robot in a narrow lane comprises the following steps:

[0007] Step 1: Construct a kinematic model of the mobile robot;

[0008] Step S2: obtaining map information and obstacle information of the mobile robot's motion area, including using the mobile robot's onboard laser radar or onboard depth camera to frame a quadrilateral formed by a feasible range of the narrow alley, obtaining the coordinate positions of the four vertices of the quadrilateral, and inputting them in the form of a YAML configuration file;

[0009] Step S3: According to the coordinate positions of the four vertices, the front and rear short sides in the direction of travel are determined, and the central axis within the feasible range is obtained by connecting the midpoints of the two short sides. According to the distance from the central axis to the two long sides on the left and right sides, four transition points between the central axis and the long sides on both sides are set on the front and rear short sides. The transition points are used to limit the driving trajectory of the mobile robot in the narrow lane, wherein the inclination angle is calculated with the two transition points on the same side of the central axis, and the posture of the four transition points is determined according to the driving rules;

[0010] Step S4: according to the driving trajectory of the mobile robot in the single mobile robot non-conflict path, the relationship between the yaw angle when entering the narrow lane and the tilt angle is determined according to the driving rules, and when the condition for entering the narrow lane is met based on the relationship, two transition points corresponding to the tilt angle are inserted into the task sequence;

[0011] Step S5: Based on the multi-mobile robot path planning algorithm in continuous space, a conflict-free path is planned from the starting position through two transition points in the task sequence and finally reaches the target position.

[0012] Furthermore, in step 1, the mobile robot is modeled according to its Ackerman kinematic model, a global map coordinate system and a robot local coordinate system are established for the movement area of ​​the mobile robot, and the starting posture state and target posture state of each mobile robot are set according to the coordinate system; the number of mobile robots is an integer greater than or equal to 2.

[0013] Furthermore, by discretizing time, the position and posture state of the mobile robot at time t is obtained:

[0014]

[0015] Among them, S t represents the position state of the mobile robot that satisfies the Ackerman kinematic equation at time t, θ represents the yaw angle of the robot, that is, the angle difference between the global map coordinate system and the robot local coordinate system, T s represents the time interval between time t and time t-1, v represents the linear velocity of the robot, L represents the distance between the front and rear wheels of the robot, Lb represents the position from the rear axle to the rear of the vehicle, Lf represents the position from the rear axle to the front of the vehicle, Ф represents the steering angle of the front wheels of the robot, and when the steering angle is fixed at Ф, the angular velocity of the robot is v / L×tanФ.

[0016] Furthermore, in step S2, the coordinate positions of the four vertices are A=(x 1 ,y 1 ), B=(x 2 ,y 2 ), C=(x 3 ,y 3)、D=(x 4 ,y 4 ); In step S3, the coordinate positions of the four transition points are:

[0017]

[0018] The tilt angle is:

[0019]

[0020] If the denominator of the tilt angle formula is 0, the angle is 90°.

[0021] Furthermore, in step S4, according to the rule of driving to the right, if the absolute value of the difference between the yaw angle and the tilt angle is less than 90°, two transition points corresponding to the tilt angle are inserted before the target position in the task sequence.

[0022] Furthermore, in step S5, after obtaining the complete task sequence, the multi-mobile robot path planning algorithm is called to sequentially calculate the path from the starting position to the first transition point, the path from the first transition point to the second transition point, and the path from the second transition point to the target position.

[0023] Furthermore, in step S5, the high-level multi-agent path planner performs conflict detection on the obstacle avoidance path of each mobile robot, and all detected conflicts are added to the conflict set. The high-level multi-agent path planner establishes a priority binary tree PCT and performs a depth-first search on it. Each node on the binary tree PCT stores a priority set, a path cost, and a path of each agent. All conflicts are collected by traversing the path of each agent, wherein the priority set and the path set are inherited from the parent node, and new priorities are added and new paths are covered;

[0024] The strategies for filtering conflicts include partial expansion of sub-nodes, prioritizing the resolution of target position conflicts, and adaptive influence priority. Appropriate conflicts are selected to construct priorities. The priority set planning starts with an empty set. The planned path is the shortest path set for each agent ignoring other agents. Based on the path set, it is determined whether the agent intends to pass through the narrow alley and its driving direction in the narrow alley, thereby filtering the corresponding two transition points and inserting them into the task sequence.

[0025] Furthermore, in step S5, the low-level single-agent path planner adopts a multi-label A* algorithm to process multi-task single-agent path planning. Each task is assigned a corresponding label according to its order in the task sequence, and a label for the current task is set globally. Each time a task is completed, the label will be updated to the label of the next task until all tasks are completed. The low-level path planner can reasonably solve the multi-task path planning problem caused by transition point insertion.

[0026] Furthermore, in step S5, the low-level single-agent path planner adopts a SIPP algorithm based on a multi-label A* algorithm to solve the path planning problem of multiple tasks. The SIPP algorithm is based on discrete space, and before each planning, the safety intervals of all map cells need to be calculated. Therefore, the overall search latitude is expanded from the original four dimensions to five dimensions. In order to adapt to the continuous space, it is necessary to maintain a region list to store the regions with safety intervals, and calculate the safety intervals of neighbor regions according to the region list each time the neighbor regions are expanded. The SIPP algorithm is based on the safety interval and uses a five-dimensional search space (x; y; θ; t; s) to plan the path of a single mobile robot. x, y, θ are continuous, representing the coordinate position (x, y) and posture θ of the robot; t is discrete, representing time; s is a set of time intervals, representing the safe time interval in which the current robot posture will not collide; each task in the task sequence of the agent is assigned a label corresponding to the order, and the task currently being executed and the number of tasks remaining to be executed are identified according to the label.

[0027] Furthermore, in step S5, the actions of the expandable path nodes of the Ackerman model mobile robot supported by the low-level single-agent path planner include: forward left turn, forward straight ahead, forward right turn, backward left turn, backward straight ahead, backward right turn and waiting in place.

[0028] The advantages and beneficial effects of the present invention are:

[0029] The present invention sets corresponding transition points by giving a feasible range of narrow alleys, and uses the transition points to limit the driving trajectory of the mobile robot in the narrow alleys, thereby solving the problem of opposite-direction conflicts in narrow alleys. The overall algorithm architecture is simple, and the path planning success rate and solution speed are high. It can provide conflict-free path planning results for robot clusters in narrow alleys in a relatively short time. The trajectory planned by the algorithm does not need to be optimized twice and can be directly applied to actual industrial scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 is a flow chart of a method in an embodiment of the present invention.

[0031] Figure 2It is a schematic diagram of the movement variation of the mobile robot in the kinematic model based on the Ackerman model in an embodiment of the present invention.

[0032] Figure 3 It is a schematic diagram of the feasible range and transition points framed in the corridor environment in an embodiment of the present invention.

[0033] Figure 4a This is one of the schematic diagrams of the execution of path planning of multiple mobile robots in a simulation environment in an embodiment of the present invention.

[0034] Figure 4b This is the second schematic diagram of the execution of path planning of multiple mobile robots in a simulation environment according to an embodiment of the present invention. DETAILED DESCRIPTION

[0035] The specific implementation of the present invention is described in detail below in conjunction with the accompanying drawings. It should be understood that the specific implementation described here is only used to illustrate and explain the present invention, and is not used to limit the present invention.

[0036] In order to achieve efficient multi-agent path planning in narrow lanes, the embodiments of the present invention provide the conditions that the multi-agent path planning algorithm needs to meet:

[0037] (1) The algorithm has a simple architecture and high solving efficiency. In a 50-square-meter test map with 16 agents, it can give a planning result in less than 30 seconds, and the planning success rate is over 90%.

[0038] (2) The test maps include two types: corridors and mazes. Both involve narrow alleys and conform to the placement habits of logistics warehouse shelves. The width of the narrow alleys is just enough for two mobile robots to pass through in parallel.

[0039] (3) The algorithm needs to support path planning in continuous space, that is, the planned path trajectory does not need to undergo secondary optimization and can be directly applied to actual industrial scenarios.

[0040] Based on the above conditions, the present invention provides a continuous space path planning method for a mobile robot in narrow lanes, such as Figure 1 As shown, the following steps are included:

[0041] Step S1: Model the mobile robot according to its Ackerman kinematic model. Establish a global map coordinate system and a robot local coordinate system for the mobile robot's motion area, and set the starting posture state and target posture state of each mobile robot according to the coordinate system; the number of mobile robots is an integer greater than or equal to 2.

[0042] like Figure 2As shown in the figure, the global map coordinate system is established with the lower left corner of the test map as the origin, the east direction as the X axis, and the north direction as the Y axis. The robot local coordinate system is established with the center of the rear axle of the mobile robot as the origin, the front direction of the mobile robot as the X' axis, and the left direction of the mobile robot as the Y' axis. The angle difference between the global map coordinate system and the robot local coordinate system is θ.

[0043] By discretizing time, the position and posture state of the robot at time t can be obtained.

[0044]

[0045] Among them, S t is the robot's position state at time t that satisfies the Ackerman kinematic equation, θ is the robot's yaw angle (the angle difference between the global map coordinate system and the robot's local coordinate system), T s is the time interval between time t and time t-1, v is the linear velocity of the robot, L is the distance between the front and rear wheels of the robot, Lb is the position from the rear axle to the rear of the vehicle, Lf is the position from the rear axle to the front of the vehicle, Ф is the steering angle of the front wheels of the robot, when the steering angle is fixed at Ф, the angular velocity of the robot is v / L×tanФ.

[0046] Step S2: Obtain map information and obstacle information of the mobile robot's motion area. The acquisition method includes using the mobile robot's on-board laser radar or on-board depth camera to frame the feasible range of the narrow alley, and the range is a quadrilateral. The coordinate positions of the four vertices of the feasible range need to be obtained and input in the form of a YAML configuration file. The coordinate positions of the four vertices A = (x 1 ,y 1 ), B=(x 2 ,y 2 ), C=(x 3 ,y 3 )、D=(x 4 ,y 4 ),like Figure 3 shown.

[0047] Step S3: Four transition points are set according to the feasible range of the narrow alley to limit the driving trajectory of the mobile robot in the narrow alley.

[0048] According to the four coordinate positions of the feasible range of the given narrow lane, the two short sides can be determined. By connecting the midpoints of the two short sides, a central axis within the feasible range can be obtained, that is, the central axis connects the midpoints of the two short sides of the quadrilateral. Then, according to the distance from the central axis to the two long sides, four transition points between the central axis and the long sides are obtained, and the four transition points must all be located on the short sides. The coordinate positions of the four transition points can be calculated as follows:

[0049]

[0050] Calculate the inclination angle using two transition points on the same side of the central axis:

[0051]

[0052] If the denominator is 0, the angle is 90°. According to the traffic rule of driving on the right, determine the positions of the four transition points.

[0053] The present invention can efficiently solve the problem of opposite-direction conflicts in narrow alleys whose width only allows two mobile robots to safely pass in parallel. According to the traffic rule of driving on the right, the feasible range of narrow alleys is framed, and corresponding transition points are set. The number of transition points is 4. The feasible range is defined as a quadrilateral, and its position is uniquely determined on the map by the coordinates of the four vertices. The central axis connecting the midpoints of the two short sides of the quadrilateral can be obtained through the coordinates of the four vertices, and the 4 transition points evenly distributed on both sides of the central axis can be obtained. The role of the transition point is to enable the intelligent body to travel according to the established route in the narrow alley, avoiding opposite-direction conflicts in the alley, thereby improving the efficiency of the solver in dealing with narrow alleys and greatly reducing the probability of deadlock.

[0054] Step S4: According to the driving trajectory of the mobile robot in the single mobile robot conflict-free path, two transition points are selected and inserted into the task sequence.

[0055] Each mobile robot calls the low-level single-agent path planning algorithm to calculate a conflict-free path while ignoring other mobile robots. Two transition points are selected according to its driving trajectory when passing through narrow alleys and inserted into the task sequence of the mobile robot.

[0056] If the conflict-free path passes through the feasible range of the narrow alley, it is necessary to determine the relationship between the yaw angle when entering the narrow alley and the inclination angle of the transition point. If the absolute value of the difference between the yaw angle and one of the inclination angles is less than 90°, the corresponding two transition points are inserted before the target position in the task sequence.

[0057] Step S5: Utilize the multi-mobile robot path planning algorithm based on continuous space to plan a conflict-free path from the starting position through two transition points and finally to the target position.

[0058] Specifically, after obtaining the complete task sequence, the multi-mobile robot path planning algorithm is called to sequentially calculate the path from the starting position to the first transition point, the path from the first transition point to the second transition point, and the path from the second transition point to the target position.

[0059] The high-level multi-agent path planner performs conflict detection on the obstacle avoidance paths of each mobile robot and adds all detected conflicts to the conflict set. The high-level multi-agent path planner establishes a priority binary tree (PCT) and performs a depth-first search on it. Each node on the PCT stores the priority set, path cost, and path of each agent. All conflicts are collected by traversing the path of each agent, where the priority set and path set are inherited from the parent node and new priorities are added and new paths are covered. The strategies for filtering conflicts include partial expansion of child nodes, priority resolution of target position conflicts, and adaptive influence priority. The present invention selects appropriate conflicts according to the strategy to construct priorities. The priority set planning starts with an empty set, so the planned path is the shortest path set for each agent ignoring other agents. According to the path set, it can be judged whether the agent intends to pass through the narrow alley and determine its driving direction in the narrow alley, so as to reasonably select the corresponding two transition points and insert them into the task sequence.

[0060] The multi-mobile robot path planning algorithm is based on priority search. Priority is used to resolve conflicts between mobile robots. The Ackerman model mobile robot supported by the low-level single-agent path planner can expand the path node actions including: forward left turn, forward straight, forward right turn, backward left turn, backward straight, backward right turn and waiting in place. The SIPP algorithm based on the multi-label A* algorithm is used to solve the multi-task path planning problem. The SIPP algorithm is based on discrete space, and the safety interval of all map cells needs to be calculated before each planning. Therefore, the overall search latitude is expanded from the original four dimensions to five dimensions. In order to adapt to the continuous space, it is necessary to maintain a region list to store the regions with safety intervals. The safety interval of the neighboring region is calculated according to the region list each time the neighboring region is expanded. The algorithm is based on the safety interval and uses a five-dimensional search space (x; y; θ; t; s) to plan the path for a single mobile robot. x, y, θ are continuous, representing the position and posture of the robot; t is discrete, representing time; s is a set of time intervals, representing the safe time interval in which the current robot posture will not collide. The tasks in each agent's task sequence are assigned a label corresponding to the order, and the label is used to identify the currently executed task and the number of tasks remaining to be executed.

[0061] The multi-label A* algorithm can handle multi-task single-agent path planning problems. Each task is assigned a corresponding label according to its order in the task sequence. The algorithm sets a label for the current task globally. After each task is completed, the label is updated to the label of the next task until all tasks are completed. This low-level path planner can reasonably solve the multi-task path planning problem caused by transition point insertion.

[0062] The present invention solves the path planning problems of multiple mobile robots that meet the Ackerman kinematic model in narrow lanes. Each mobile robot has an independent starting position and target position. The proposed narrow lane feasible range framing and transition point setting guidance can efficiently and quickly solve the conflict problem in the narrow lane, and the planned path trajectory conforms to the Reeds-Shepp curve, that is, it can realize the reverse parking of vehicles.

[0063] The narrow lane conflict resolution strategy proposed in this invention mainly relies on the definition of the feasible range and the calculation of the transition point, such as Figure 2 As shown in the figure, the central axis is determined according to the coordinates of the four vertices in the feasible range, and the coordinate positions of the four transition points are calculated. The transition points are divided into two groups according to the central axis. The inclination angle of each group of transition points is calculated according to the coordinate position, and the posture of each transition point is finally determined. The low-level single-agent path planning algorithm is used to calculate the conflict-free path of each agent when ignoring other agents. According to the posture state when it enters the feasible range of the narrow alley, the corresponding transition point is inserted before the target position of the task sequence. The transition point plays a role in guiding the driving of the mobile robot in the entire planning process. According to the traffic rules of driving on the right, the driving trajectory of the mobile robot in the narrow alley is regulated, the possibility of internal conflicts in the alley is reduced, and the solution efficiency of the algorithm is improved.

[0064] The present invention has been tested in a simulation environment, and the test results are as follows: Figure 4a , Figure 4b In the simulation environment, the present invention can simultaneously perform path planning for 16 2m×3m Ackerman model agents on a 50m×50m corridor map, with a planning time of less than 30s and a success rate of more than 95%. It can simultaneously perform path planning for 12 2m×3m Ackerman model agents on a 50m×50m maze map, with a planning time of less than 15s and a success rate of more than 95%.

[0065] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some or all of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A continuous space path planning method for a mobile robot in narrow lanes, characterized in that The steps include: Step 1: Construct a kinematic model of the mobile robot; Step S2: Frame a quadrilateral formed by the feasible range of the narrow lane, and obtain four vertices of the quadrilateral; Step S3: Determine the front and rear sides in the direction of travel according to the four vertices, obtain the central axis within the feasible range by connecting the midpoints of the two sides, and set four transition points between the central axis and the two sides on the front and rear sides according to the distance from the central axis to the left and right sides thereof. The transition points are used to limit the driving trajectory of the mobile robot in the narrow lane, wherein the inclination angle is calculated using the two transition points on the same side of the central axis; Step S4: according to the driving trajectory of the mobile robot in the single mobile robot non-conflict path, the relationship between the yaw angle when entering the narrow lane and the tilt angle is determined according to the driving rules, and when the condition for entering the narrow lane is met based on the relationship, two transition points corresponding to the tilt angle are inserted into the task sequence; Step S5: Based on the multi-mobile robot path planning algorithm in continuous space, a conflict-free path is planned from the starting position through two transition points in the task sequence and finally reaches the target position.

2. The method for continuous spatial path planning of a mobile robot facing a narrow lane according to claim 1, characterized in that: In the step 1, a mobile robot is modeled according to its Ackerman kinematic model, a global map coordinate system and a robot local coordinate system are established for the movement area of ​​the mobile robot, and a starting posture state and a target posture state of each mobile robot are set according to the coordinate system; the number of mobile robots is an integer greater than or equal to 2.

3. The method for continuous spatial path planning of a mobile robot facing a narrow lane according to claim 2, characterized in that: By discretizing time, the position and posture state of the mobile robot at time t is obtained: Among them, S t represents the position state of the mobile robot that satisfies the Ackerman kinematic equation at time t, θ represents the yaw angle of the robot, that is, the angle difference between the global map coordinate system and the robot local coordinate system, T s represents the time interval between time t and time t-1, v represents the linear velocity of the robot, L represents the distance between the front and rear wheels of the robot, Lb represents the position from the rear axle to the rear of the vehicle, Lf represents the position from the rear axle to the front of the vehicle, Ф represents the steering angle of the front wheels of the robot, and when the steering angle is fixed at Ф, the angular velocity of the robot is v / L×tanФ.

4. The method for continuous spatial path planning of a mobile robot facing a narrow lane according to claim 1, characterized in that: In step S2, the coordinate positions of the four vertices are A=(x1, y1), B=(x2, y2), C=(x3, y3), and D=(x4, y4); In step S3, the coordinate positions of the four transition points are: The tilt angle is: If the denominator of the tilt angle formula is 0, the angle is 90°.

5. The method for continuous spatial path planning of a mobile robot facing a narrow lane according to claim 1, characterized in that: In step S4, according to the rule of driving to the right, if the absolute value of the difference between the yaw angle and the tilt angle is less than 90°, two transition points corresponding to the tilt angle are inserted before the target position in the task sequence.

6. The method for continuous spatial path planning of a mobile robot facing a narrow lane according to claim 1, characterized in that: In step S5, after obtaining the complete task sequence, the multi-mobile robot path planning algorithm is called to sequentially calculate the path from the starting position to the first transition point, the path from the first transition point to the second transition point, and the path from the second transition point to the target position.

7. The method for continuous spatial path planning of a mobile robot facing a narrow lane according to claim 1, characterized in that: In step S5, the high-level multi-agent path planner performs conflict detection on the obstacle avoidance path of each mobile robot, and adds all detected conflicts to the conflict set. The high-level multi-agent path planner establishes a priority binary tree and performs a depth-first search on it. Each node on the binary tree stores a priority set, a path cost, and a path of each agent. All conflicts are collected by traversing the path of each agent, wherein the priority set and the path set are inherited from the parent node, and new priorities are added and new paths are covered; The strategies for filtering conflicts include partial expansion of sub-nodes, prioritizing the resolution of target position conflicts, and adaptive influence priority. Appropriate conflicts are selected to construct priorities. The priority set planning starts with an empty set. The planned path is the shortest path set for each agent ignoring other agents. Based on the path set, it is determined whether the agent intends to pass through the narrow alley and its driving direction in the narrow alley, thereby filtering the corresponding two transition points and inserting them into the task sequence.

8. The method for continuous spatial path planning of a mobile robot facing a narrow lane according to claim 1, characterized in that: In step S5, the low-level single-agent path planner uses a multi-label A* algorithm to process multi-task single-agent path planning. Each task is assigned a corresponding label according to its order in the task sequence. A label for the current task is set globally. Each time a task is completed, the label will be updated to the label of the next task until all tasks are completed.

9. The method for continuous spatial path planning of a mobile robot facing a narrow lane according to claim 8, characterized in that: In step S5, the low-level single-agent path planner adopts the SIPP algorithm based on the multi-label A* algorithm; in order to adapt to the continuous space, it is necessary to maintain an area list to store the areas with safe intervals, and calculate the safe intervals of the neighboring areas according to the area list each time the neighboring areas are expanded; the SIPP algorithm is based on the safe interval and uses a five-dimensional search space (x; y; θ; t; s) to plan the path of a single mobile robot, where x, y, and θ are continuous, representing the robot's coordinate position (x, y) and posture θ; t is discrete, representing time; s is a set of time intervals, representing the safe time interval in which the current robot posture will not collide; each task in the task sequence of the agent is assigned a label corresponding to the order, and the task currently being executed and the number of tasks remaining to be executed are identified based on the label.

10. The method for continuous spatial path planning of a mobile robot facing a narrow lane according to claim 1, characterized in that: In step S5, the actions of the expandable path nodes of the Ackerman model mobile robot supported by the low-level single-agent path planner include: forward left turn, forward straight ahead, forward right turn, backward left turn, backward straight ahead, backward right turn and waiting in place.

Images