Robot path planning method and system based on avoidance strategy

Abstract

The invention relates to the technical field of path planning, and discloses a robot path planning method and system based on an avoidance strategy, and the robot path planning method based on the avoidance strategy comprises the following steps: a scheduling system generates a motion path of a robot; the scheduling system calculates whether the sum of the lengths of the road sections occupied by the current robot is smaller than planDist, if yes, whether the conditions are met is judged, and if yes, the passing permission is given to the robot; after the robot obtains the passing authority, the robot occupies the next road section and the terminal point of the next road section and advances towards the terminal point of the next road section, and when the robot travels for a specified distance on the road section, the starting point of the road section is released; when the robot runs to the end point of the road section, the road section is released; and each robot repeats the above steps until all robots reach the end point of the task. According to the method, the possibility of deadlock occurrence is reduced, and the deadlock and the speed adjustment strategy are quickly relieved, so that the overall waybill efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of path planning, and more specifically, to a robot path planning method and system based on an avoidance strategy. Background Art

[0002] In automated warehousing and manufacturing scenarios, multiple robots are often required to work together to complete material handling tasks. In order for the robots to complete tasks efficiently, safely, and without collision, the robot's motion path needs to be reasonably planned and scheduled. When multiple robots are assigned to carry goods from the pickup point to the delivery point in a limited space, the exact location of the robots cannot be confirmed in the case of poor network conditions.

[0003] When the network is good, robots that may collide need to wait for execution commands before executing them. This waiting process increases the time consumption of the entire task.

[0004] Since waybills are unpredictable, the execution efficiency of different waybill strategies is also different. Summary of the invention

[0005] The present invention provides a robot path planning method and system based on an avoidance strategy, which solves the technical problems in the related art such as poor network, inability to confirm the accurate position of the robot, increased possibility of deadlock, difficulty in releasing the deadlock and low execution efficiency.

[0006] The present invention provides a robot path planning method based on an avoidance strategy, comprising the following steps:

[0007] Step 100, the scheduling system generates a movement route according to the current station of the robot and the task target station;

[0008] Step 200, the dispatching system calculates whether the sum of the lengths of the sections currently occupied by the robot is less than planDist based on the generated motion route. If so, it determines whether steps 201-204 are satisfied, and grants the robot corresponding access rights based on whether steps 201-204 are satisfied.

[0009] Step 201, perform deadlock detection to determine whether the current robot will be deadlocked with other robots when driving to the end of the next road segment; if not, proceed to step 202; if it occurs and can be resolved, generate a waiting instruction for the current station; if it occurs and cannot be resolved, adopt an avoidance strategy;

[0010] Step 202, the dispatching system determines whether the next segment and the end point of the next segment are occupied by other robots; if the next segment or the end point of the next segment is occupied by other robots, the dispatching system instructs the robot to wait at the current station; if the robot has occupied the next segment and the end point of the next segment, the dispatching system instructs the robot to move to the end point of the next segment; if the next segment and the end point of the next segment are not occupied, the judgment step 203 is performed;

[0011] Step 203, the dispatching system uses a collision detection method to determine whether the robot will collide with other robots when traveling to the end of the next road segment according to the length, width and driving direction of the robot; if there is a collision, the dispatching system adopts a speed adjustment strategy for the robot; if there is no collision, the judgment step 204 is performed;

[0012] Step 204, determine whether the next segment and the end point of the next segment are in the mutually exclusive area; if there are other robots in the mutually exclusive area, the dispatching system instructs the robot to wait at the current station; if there are no other robots in the mutually exclusive area, the robot obtains the right of passage;

[0013] Step 300, after the robot obtains the right of passage, the robot occupies the next road section and the end point of the next road section. The dispatching system instructs the robot to move to the end point of the next road section. When the robot travels a specified distance on the road section, the starting point of the road section is released. The specified distance is set to the maximum length of the robot plus the safety margin and the positioning error. When the robot travels to the end point of the road section, the road section is released.

[0014] In step 400, each robot repeats steps 100-300 until all robots reach the end point of the task.

[0015] In a preferred embodiment, the deadlock detection method is as follows:

[0016] If the two robots are about to deadlock, the judgment method is that the remaining path of robot No. 1 is about to pass through the occupied station of robot No. 2 or the associated station of the occupied station, and the remaining path of robot No. 2 is about to pass through the next station of robot No. 1 or its associated station, then robot No. 1 needs to stop at the current station and wait for robot No. 2 to pass through the next station of robot No. 1 or its associated station;

[0017] The associated site of a site means that if there is a robot on the site, there cannot be other robots on its associated site at the same time, otherwise a collision will occur. These sites are called the associated sites of the site; the remaining path of the robot refers to the path of the robot from the current site to the end point of the current task, excluding the path the robot has already traveled.

[0018] In a preferred embodiment, the avoidance strategy is as follows:

[0019] If there are fixed avoidance points in the scene, select a deadlocked robot and look for the nearest avoidance point that is not occupied by other robots and will not deadlock with other robots when reaching the avoidance point. If an avoidance point that meets this condition is found, occupy the avoidance point and update the path to this avoidance point; if not found, the path update with the lowest cost in the event of deadlock will be selected for path update; if there are no fixed avoidance points in the scene, select an idle storage site or an idle stop point as a temporary avoidance point; if the deadlock is still not resolved after executing the above avoidance strategy, add robots that need to execute the avoidance strategy one by one and repeat the above avoidance strategy until the deadlock is resolved.

[0020] In a preferred embodiment, the collision detection method is: based on the two rectangular bounding boxes formed by the length, width, driving direction angle, safety margin and positioning error of the two robots, determine whether the two bounding boxes intersect. If they intersect, the two robots will collide, otherwise there will be no collision; the default value of the safety margin is 300mm, and the default value of the positioning error is 20mm.

[0021] In a preferred embodiment, the scheduling system has an input parameter planDist, which is the distance that the scheduling system plans the robot to walk in advance each time, and the default value is 3 meters; the scheduling system calculates whether the sum of the lengths of the occupied sections is less than planDist; if it is less than planDist, then continue to determine whether the robot can walk to the next section and the end of the next section according to steps 201-204; the section occupied by the robot, that is, the section allocated to the robot by the scheduling system is the route that the robot must travel; if it is greater than planDist, then the current robot in this round of planning is judged to be over, and continue to plan the next robot until all robots in this round of planning are judged to be over.

[0022] In a preferred embodiment, there can be at most one robot in the mutually exclusive zone.

[0023] In a preferred embodiment, the speed adjustment strategy is as follows:

[0024] Step 1, obtain the status information of the current robot and other robots in the scene at the current time step;

[0025] Step 2: construct a directed graph G(V,E) for the state information of multiple robots obtained at the current time step, where each robot is a node, the edge represents the influence relationship of each robot, and the node feature is the state information feature of the robot;

[0026] Step 3, constructing a speed adjustment model;

[0027] Step 4: The model outputs the adjustment amount of the current robot speed.

[0028] In a preferred embodiment, the loss function of the speed adjustment model is:

[0029] ;

[0030] ;

[0031] ;

[0032] ;

[0033] : Safety loss item;

[0034] : Smoothness loss term;

[0035] : efficiency loss term;

[0036] : safety distance threshold;

[0037] : The Euclidean distance between the i-th robot and the j-th robot;

[0038] : The new acceleration of the i-th robot after adjustment;

[0039] : The new speed of the i-th robot after adjustment;

[0040] : expected speed;

[0041] , , : The first, second and third weight coefficients;

[0042] : Total reward function.

[0043] The present invention also provides a robot path planning system based on an avoidance strategy, which is used to execute the above-mentioned robot path planning method based on an avoidance strategy.

[0044] The beneficial effects of the present invention are as follows: the present invention implements an efficient waybill strategy, and during the waybill process, the robot occupies not a spatial position in the scheduling system, but a route, a group of mutually exclusive groups, and a detour area, thereby reducing the possibility of deadlock. Even if deadlock occurs, the deadlock can be resolved through an avoidance strategy, and a speed adjustment strategy is performed when a collision is about to occur, thereby improving the overall waybill efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 is a flow chart of a robot path planning method based on an avoidance strategy of the present invention;

[0046] Figure 2 It is a flow chart for judging traffic conditions in a robot path planning method based on an avoidance strategy of the present invention. DETAILED DESCRIPTION

[0047] The subject matter described herein will now be discussed with reference to example implementations. It should be understood that the discussion of these implementations is only to enable those skilled in the art to better understand and implement the subject matter described herein, and the functions and arrangements of the elements discussed may be changed without departing from the scope of protection of the present specification. Various examples may omit, replace, or add various processes or components as needed. In addition, the features described in some examples may also be combined in other examples.

[0048] At least one embodiment of the present invention discloses a robot path planning method based on an avoidance strategy, such as Figure 1 As shown, the following steps are included:

[0049] Step 100, the scheduling system generates a movement route according to the current station of the robot and the task target station;

[0050] In one embodiment of the present invention, the dispatching system is a comprehensive system that integrates the back-end Core (development), front-end software (including implementation tools RoboShop Pro and RDS (development), etc.), simulation robots, and simulation equipment. It realizes the communication connection between various components through multiple protocols such as HTTP, TCP, and ModbusTCP. It can reasonably dispatch multiple robots to complete the task of transporting goods from the pickup point to the delivery point in a limited space according to the orders issued by users or WMS (Warehouse Management System), and has complex traffic control, task management, equipment monitoring and operation functions, aiming to improve logistics transportation efficiency, reduce implementation costs, and ensure the efficient and stable operation of the entire logistics operation process.

[0051] Step 200, the dispatching system calculates whether the sum of the lengths of the sections currently occupied by the robot is less than planDist based on the generated motion route. If so, it determines whether steps 201-204 are satisfied, and grants the robot corresponding access rights based on whether steps 201-204 are satisfied.

[0052] The steps of determining whether steps 201-204 are satisfied are as follows: Figure 2 As shown, including:

[0053] Step 201, perform deadlock detection to determine whether the current robot will be deadlocked with other robots when driving to the end of the next road segment; if not, proceed to step 202; if it occurs and can be resolved, generate a waiting instruction for the current station; if it occurs and cannot be resolved, adopt an avoidance strategy;

[0054] In one embodiment of the present invention, the deadlock detection method is as follows:

[0055] If the remaining path of Robot No. 1 is about to pass through the occupied site of Robot No. 2 or the associated site of the occupied site, and the remaining path of Robot No. 2 is also about to pass through the next site of Robot No. 1 or its associated site, it is judged that a deadlock is about to occur. The method to be adopted is: Robot No. 1 needs to stop at the current site and wait for Robot No. 2 to pass through the next site of Robot No. 1 or its associated site.

[0056] In one embodiment of the present invention, if two robots have deadlocked, the robot with the smallest updated path cost is selected to go to the avoidance point, that is, the robot that is closer to the nearest avoidance point and does not deadlock with other robots.

[0057] In one embodiment of the present invention, if a loop deadlock occurs among three robots, the judgment method is that robot No. 1 is about to pass through the occupied site of robot No. 2 or its associated site, and robot No. 2 is about to pass through the occupied site of robot No. 3 or its associated site, and robot No. 3 or No. 2 is also about to pass through the current site of robot No. 1 or its associated site, and a closed loop needs to be formed, that is, the loop occupies each other, before it can be judged as a loop deadlock; if a loop deadlock occurs among more than three robots, the judgment method is similar. When a loop deadlock occurs, firstly, all robots in the deadlock loop need to be detected according to the above method, and then the robot with the smallest path update cost is selected to update the path to the nearest avoidance point. If a robot deadlocks with other robots in the process of going to the avoidance point, the robot with the smallest path update cost will delete the original avoidance point plan, re-update the path to the new nearest avoidance point or restore the original path.

[0058] It should be noted that the associated site of a site means that if there is a robot on the site, there cannot be other robots on its associated site at the same time, otherwise a collision will occur. These sites are called the associated sites of the site. The remaining path of the robot refers to the path of the robot from the current site to the end point of the current task, excluding the path that the robot has already traveled.

[0059] In one embodiment of the present invention, the avoidance strategy is as follows:

[0060] If there is a fixed avoidance point in the scene, select a deadlocked robot and look for the nearest avoidance point that is not occupied by other robots and will not cause deadlock with other robots when reaching the avoidance point. If an avoidance point that meets this condition is found, occupy the avoidance point and update the path to this avoidance point; if not found, choose the path with the lowest cost in the event of deadlock to update the path; if there is no fixed avoidance point in the scene, select an idle warehouse site or an idle stop point as a temporary avoidance point; if the deadlock is still not resolved after executing the above avoidance strategy, add robots that need to execute the avoidance strategy one by one and repeat the above avoidance strategy until the deadlock is resolved;

[0061] Updating the path with the minimum cost means that the robot chooses the path with the highest comprehensive score when planning the path; the comprehensive score includes short path, many avoidance points and no deadlock with other robots when updating the path.

[0062] Step 202, the dispatching system determines whether the next segment and the end point of the next segment are occupied by other robots; if the next segment or the end point of the next segment is occupied by other robots, the dispatching system instructs the robot to wait at the current station; if the robot has occupied the next segment and the end point of the next segment, the dispatching system instructs the robot to move to the end point of the next segment; if the next segment and the end point of the next segment are not occupied, the judgment step 203 is performed;

[0063] Step 203, the dispatching system uses a collision detection method to determine whether the robot will collide with other robots when traveling to the end of the next road segment according to the length, width and driving direction of the robot; if there is a collision, the dispatching system adopts a speed adjustment strategy for the robot; if there is no collision, the judgment step 204 is performed;

[0064] In one embodiment of the present invention, the collision detection method is: based on the rectangular bounding box formed by the length, width, driving direction angle, safety margin and positioning error of the robot, determine whether the two bounding boxes intersect. If they intersect, the two robots will collide, otherwise, no collision will occur; the default value of the safety margin is 300mm, and the default value of the positioning error is 20mm.

[0065] In one embodiment of the present invention, the speed adjustment strategy is as follows:

[0066] Step 1, obtain the status information of the current robot and other robots in the scene at the current time step;

[0067] The robot's status information includes: the robot's current speed, acceleration, driving direction angle, position coordinates, length, and width;

[0068] Step 2: construct a directed graph G(V,E) for the state information of multiple robots obtained at the current time step, where each robot is a node, the edge represents the influence relationship of each robot, and the node feature is the state information feature of the robot;

[0069] Step 3, constructing a speed adjustment model;

[0070] In one embodiment of the present invention, MLP, GRU, and GNN are used to perform step 3. The specific formula is as follows:

[0071] Node feature initialization:

[0072] ;

[0073] : Initialization node features of the i-th robot;

[0074] : Multilayer Perceptron;

[0075] : Node features of the i-th robot;

[0076] The edge features between the i-th robot and the j-th robot are: ;

[0077] ;

[0078] ;

[0079] ;

[0080] ;

[0081] , : The position coordinates of the i-th and j-th robots in the two-dimensional plane;

[0082] , : The speed of the i-th and j-th robots;

[0083] , : acceleration of the ith and jth robots;

[0084] , : The driving direction angle of the i-th and j-th robots;

[0085] : The Euclidean distance between the i-th robot and the j-th robot;

[0086] : The speed difference between the i-th robot and the j-th robot;

[0087] : The acceleration difference between the ith robot and the jth robot;

[0088] : The minimum angle difference between the i-th robot and the j-th robot;

[0089] : The edge feature between the i-th robot and the j-th robot;

[0090] Edge attention calculation:

[0091] ;

[0092] : Attention weight coefficient, indicating the influence of node j on node i;

[0093] : The hidden state of node i in layer k;

[0094] : The hidden state of node j in layer k;

[0095] : The hidden state of node k in the kth layer;

[0096] : The edge feature between the i-th robot and the j-th robot;

[0097] : The edge feature between the i-th robot and the k-th robot;

[0098] : vector concatenation operator;

[0099] : The set of neighbor nodes of node i;

[0100] : Learnable attention weight matrix;

[0101] : Activation function;

[0102] Messaging:

[0103] ;

[0104] : Attention weight coefficient, indicating the influence of node j on node i;

[0105] : The hidden state of node j in layer k;

[0106] : vector concatenation operator;

[0107] : The set of neighbor nodes of node i;

[0108] : The aggregated message received by node i in layer k;

[0109] : learnable aggregation weight matrix;

[0110] Status Update:

[0111] ;

[0112] : The hidden state of node i in the k+1th layer;

[0113] : The hidden state of node i in layer k;

[0114] : The aggregated message received by node i in layer k;

[0115] : Gated recurrent unit;

[0116] Step 4, the model outputs the adjustment amount of the current robot speed;

[0117] In one embodiment of the present invention, MLP is applied to perform model output, and the specific formula is as follows:

[0118] ;

[0119] ;

[0120] : The speed adjustment predicted by the ith robot;

[0121] : Multilayer Perceptron;

[0122] : The hidden state of node i in the Kth layer after propagation through K layers;

[0123] : The new speed of the i-th robot after adjustment;

[0124] : The speed of the ith robot;

[0125] In one embodiment of the present invention, there are physical and dynamic constraints on speed and acceleration, and the specific constraints are:

[0126] ;

[0127] ;

[0128] ;

[0129] ;

[0130] : Speed ​​lower limit;

[0131] : upper speed limit;

[0132] : The new speed of the i-th robot after adjustment;

[0133] : The new speed of the jth robot after adjustment;

[0134] : The new acceleration of the i-th robot after adjustment;

[0135] : Maximum permissible acceleration;

[0136] : The Euclidean distance between the i-th robot and the j-th robot;

[0137] : safety distance threshold;

[0138] : The speed of the ith robot;

[0139] : time step;

[0140] In one embodiment of the present invention, the aforementioned MLP, GRU, and GNN all need to be trained to update parameters. They can be jointly trained with the MLP executing step 4. In one embodiment of the present invention, the joint training is performed by an unsupervised learning method, and a loss function is designed, such as:

[0141] ;

[0142] ;

[0143] ;

[0144] ;

[0145] : Safety loss item;

[0146] : Smoothness loss term;

[0147] : efficiency loss term;

[0148] : safety distance threshold;

[0149] : The Euclidean distance between the i-th robot and the j-th robot;

[0150] : The new acceleration of the i-th robot after adjustment;

[0151] : The new speed of the i-th robot after adjustment;

[0152] : expected speed;

[0153] , , : The first, second and third weight coefficients;

[0154] : total reward function;

[0155] Step 204, determine whether the next segment and the end point of the next segment are in the mutually exclusive area; if there are other robots in the mutually exclusive area, the dispatching system instructs the robot to wait at the current station; if there are no other robots in the mutually exclusive area, the robot obtains the right of passage;

[0156] In one embodiment of the present invention, there can be at most one robot in the mutually exclusive zone.

[0157] Step 300: After the robot obtains the right of passage, it occupies the next road segment and the end point of the next road segment. The dispatching system instructs the robot to move to the end point of the next road segment. When the robot travels a specified distance on the road segment, the starting point of the road segment is released. The specified distance is set to the maximum length of the robot plus the safety margin and positioning error. When the robot reaches the end point of the road segment, the road segment is released;

[0158] In step 400, each robot repeats steps 100-300 until all robots reach the end point of the task.

[0159] In one embodiment of the present invention, the scheduling system has an input parameter planDist, which is the distance that the scheduling system plans the robot to walk in advance each time, and the default value is 3 meters. The scheduling system calculates whether the sum of the lengths of the occupied sections is less than planDist. If it is less than planDist, continue to judge whether the robot can walk to the next section and the end of the next section according to the above steps 201-204. The section occupied by the robot, that is, the section that the scheduling system has allocated to the robot is the route that the robot must travel. If it is greater than planDist, the current robot in this round of planning is judged to be over, and the next robot is planned until all robots in this round of planning are judged to be over.

[0160] In one embodiment of the present invention, a robot path planning system based on an avoidance strategy is used to execute the above-mentioned robot path planning method based on an avoidance strategy.

[0161] The above describes an embodiment of the present invention, but this embodiment is not limited to the above-mentioned specific implementation mode. The above-mentioned specific implementation mode is merely illustrative and not restrictive. Under the guidance of this embodiment, ordinary technicians in this field can also make more forms of equivalent embodiments, all of which are within the protection of this embodiment.

Claims

1. A robot path planning method based on avoidance strategy, characterized in that: The following steps are involved: Step 100, the scheduling system generates a movement route according to the current station of the robot and the task target station; Step 200, the dispatching system calculates whether the sum of the lengths of the sections currently occupied by the robot is less than planDist based on the generated motion route. If so, it determines whether steps 201-204 are satisfied, and grants the robot corresponding access rights based on whether steps 201-204 are satisfied. Steps 201-204 are as follows: Step 201, perform deadlock detection to determine whether the current robot will be deadlocked with other robots when driving to the end of the next road segment; if not, proceed to step 202; if it occurs and can be resolved, generate a waiting instruction for the current station; if it occurs and cannot be resolved, adopt an avoidance strategy; Step 202, the dispatching system determines whether the next segment and the end point of the next segment are occupied by other robots; if the next segment or the end point of the next segment is occupied by other robots, the dispatching system instructs the robot to wait at the current station; if the robot has occupied the next segment and the end point of the next segment, the dispatching system instructs the robot to move to the end point of the next segment; if the next segment and the end point of the next segment are not occupied, the judgment step 203 is performed; Step 203, the dispatching system uses a collision detection method to determine whether the robot will collide with other robots when traveling to the end of the next road segment according to the length, width and driving direction of the robot; if a collision occurs, the dispatching system adopts a speed adjustment strategy for the robot; if no collision occurs, the judgment step 204 is performed; Step 204, determine whether the next segment and the end point of the next segment are in the mutually exclusive area; if there are other robots in the mutually exclusive area, the dispatching system instructs the robot to wait at the current station; if there are no other robots in the mutually exclusive area, the robot obtains the right of passage; Step 300, after the robot obtains the right of passage, the robot occupies the next road section and the end point of the next road section. The dispatching system instructs the robot to move to the end point of the next road section. When the robot travels a specified distance on the road section, the starting point of the road section is released. The specified distance is set to the maximum length of the robot plus the safety margin and the positioning error. When the robot travels to the end point of the road section, the road section is released. In step 400, each robot repeats steps 100-300 until all robots reach the end point of the task.

2. A robot path planning method based on avoidance strategy according to claim 1, characterized in that: The deadlock detection method is as follows: If the two robots are about to deadlock, the judgment method is that the remaining path of robot No. 1 is about to pass through the occupied station of robot No. 2 or the associated station of the occupied station, and the remaining path of robot No. 2 is about to pass through the next station of robot No. 1 or its associated station, then robot No. 1 needs to stop at the current station and wait for robot No. 2 to pass through the next station of robot No. 1 or its associated station; The associated site of a site means that if there is a robot on the site, there cannot be other robots on its associated site at the same time, otherwise a collision will occur. These sites are called the associated sites of the site; the remaining path of the robot refers to the path of the robot from the current site to the end point of the current task, excluding the path the robot has already traveled.

3. A robot path planning method based on avoidance strategy according to claim 1, characterized in that: The avoidance strategy is as follows: If there is a fixed avoidance point in the scene, select a deadlocked robot and look for the nearest avoidance point that is not occupied by other robots and will not cause deadlock with other robots when reaching the avoidance point. If an avoidance point that meets this condition is found, occupy the avoidance point and update the path to this avoidance point; if not found, choose the path with the minimum cost in the event of deadlock to update the path; if there is no fixed avoidance point in the scene, select an idle storage site or an idle stop point as a temporary avoidance point; If the deadlock is still not resolved after executing the above avoidance strategy, add robots that need to execute the avoidance strategy one by one and repeat the above avoidance strategy until the deadlock is resolved.

4. A robot path planning method based on avoidance strategy according to claim 1, characterized in that: The collision detection method is: based on the two rectangular bounding boxes formed by the length, width, driving direction angle, safety margin and positioning error of the two robots, determine whether the two bounding boxes intersect. If they intersect, the two robots will collide, otherwise there will be no collision; the default value of the safety margin is 300mm, and the default value of the positioning error is 20mm.

5. A robot path planning method based on avoidance strategy according to claim 1, characterized in that: The scheduling system has an input parameter planDist, which is the distance that the scheduling system plans the robot to walk in advance each time, and the default value is 3 meters; the scheduling system calculates whether the sum of the lengths of the occupied sections is less than planDist; if it is less than planDist, it continues to determine whether the robot can walk to the next section and the end of the next section according to steps 201-204; the section occupied by the robot, that is, the section allocated to the robot by the scheduling system is the route that the robot must travel; if it is greater than planDist, the current robot in this round of planning is judged to be over, and the next robot is planned until all robots in this round of planning are judged to be over.

6. A robot path planning method based on avoidance strategy according to claim 1, characterized in that: There can be at most one robot in the mutually exclusive zone.

7. A robot path planning method based on avoidance strategy according to claim 1, characterized in that: The speed adjustment strategy is as follows: Step 1, obtain the status information of the current robot and other robots in the scene at the current time step; Step 2: construct a directed graph G(V,E) for the state information of multiple robots obtained at the current time step, where each robot is a node, the edge represents the influence relationship of each robot, and the node feature is the state information feature of the robot; Step 3, constructing a speed adjustment model; Step 4: The model outputs the adjustment amount of the current robot speed.

8. A robot path planning method based on avoidance strategy according to claim 7, characterized in that: The loss function of the speed adjustment model is: ； ； ； ； : Safety loss item; : Smoothness loss term; : efficiency loss term; : safety distance threshold; : The Euclidean distance between the i-th robot and the j-th robot; : The new acceleration of the i-th robot after adjustment; : The new speed of the i-th robot after adjustment; : expected speed; , , : The first, second and third weight coefficients; : Total reward function.

9. A robot path planning system based on avoidance strategy, characterized in that: It is used to execute a robot path planning method based on an avoidance strategy as described in any one of claims 1-8.

Images