Multi-Dubins robot collaborative coverage path planning method based on competitive game

Through the path planning method based on competition game, the coverage path path planning method of multi-Dubins robots is decomposed and adjusted, and the problem of path planning efficiency and balance in the existing technology is solved, and efficient and balanced coverage path planning is achieved in complex environments.

CN118936481APending Publication Date: 2024-11-12MILITARY INTELLIGENCE RES INST OF THE CHINESE PEOPLES LIBERATION ARMY ACAD OF MILITARY SCI
View PDF 0 Cites 2 Cited by

Patent Information

Application Number
CN202411401011.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Priority Date
2024-10-08
Filing Date
2024-10-09
Publication Date
2024-11-12

AI Technical Summary

Technical Problem

Existing multi-Dubins robot collaborative coverage path planning algorithms are difficult to plan balanced, efficient and smooth coverage paths in a short period of time, especially in complex environments, and existing algorithms usually need to search the entire solution space or adopt heuristics to reduce Calculate time, but this can affect the accuracy or lead to uneven coverage results.

Method used

Using the multi-Dubins robot coverage path planning method based on competitive game, the decomposition target area is approximately equilibrium by constructing a system including area decomposition module, initial division module, credit model-based task fine adjustment module and game strategy-based task fine adjustment module. and adjust the path through game strategy to achieve balanced coverage.

Benefits of technology

A nearly optimal coverage path is planned for multiple Dubins robots within a limited time, ensuring smooth and continuous paths, meeting the robot's kinematic constraints, and improving the allocation balance and solution accuracy of coverage tasks.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118936481A_ABST
    Figure CN118936481A_ABST
Patent Text Reader

Abstract

The invention discloses a multi-Dubins robot collaborative coverage path planning method based on a competitive game. The multi-Dubins robot collaborative coverage path planning method comprises the step of constructing a multi-Dubins robot coverage path planning system composed of a server, a client and K robots. The client is responsible for collecting information of a target area. The initial division module divides a target area into K partitions, generates a partition set P, and calculates the cost of each partition. And a task fine adjustment module based on the credit model adjusts the partition set P to obtain a fine adjustment partition set P '. The task fine adjustment module based on the game strategy executes a path planning algorithm in parallel on the K partitions corresponding to the P ', K Dubins coverage paths are generated, a path set RD is formed, and RD' is obtained through further adjustment. And the K robots move in parallel according to the coverage path to generate K motion tracks so as to realize the coverage of the target area. By adopting the method, an approximately optimal coverage path can be planned, and the kinematic mechanics constraint of the robot is met.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the technical field of multi-robots, and in particular to a multi-Dubins robot path planning method that can be used for area coverage applications. Background Art

[0002] In application scenarios that require area coverage, such as commercial mapping, agricultural plant protection, and floor cleaning, task allocation and path planning of multi-robot systems are key technologies for solving practical problems. Area coverage aims to determine a path for a mobile robot that passes through all points in the target area through reasonable task allocation and path planning algorithms. The complexity of area coverage is affected by many factors, of which the two most important factors are the target environment and the robot's dynamic constraints. Generally speaking, coverage path planning in obstacle scenarios is more complex than coverage path planning in open, obstacle-free environments; coverage path planning for robots with non-holonomic dynamic constraints is more challenging than coverage path planning for robots with complete constraints. Many robots in real applications are non-holonomically constrained robots, such as unmanned vehicles, fixed-wing drones, and unmanned underwater vehicles. These robots can usually be modeled as Dubins robots. The minimum turning radius of Dubins robots is limited, and their motion paths can only be composed of straight lines and curves. Many practical applications, such as environmental search, rescue detection, and agricultural automation, widely use Dubins robots. These Dubins robots need to perform efficient path planning in complex environments to achieve the task of fully covering the target area.

[0003] The collaborative coverage path planning problem of multiple Dubins robots has been proven to be an NP-hard problem. Existing multi-robot collaborative coverage path planning algorithms can be divided into two categories: exact algorithms and heuristic algorithms. Both types of algorithms usually consider dividing the area first and then planning the path. The algorithm that divides the area exactly (such as the BCPq method, Zhou X, Wang H, Ding B, et al. Balanced connected task allocations for multi-robot systems: An exact flow-based integer program and an approximate tree-based genetic algorithm [J]. Expert Systems with Applications, 2019, translated as Zhou X, Wang H, DingB, etc., Balanced connected task allocations for multi-robot systems: An exact flow-based integer program and an approximate tree-based genetic algorithm, 2019 Expert Systems and Applications) divides the target area by searching the entire solution space. This type of method uses brute force solutions such as linear programming to obtain balanced area division results. However, the exact method needs to search the entire solution space, and its corresponding solution speed is slow, which is usually suitable for small-scale scenarios. Heuristic region partitioning algorithms (such as the CDM algorithm, i.e., Li L, Shi D, Jin S, et al. Exact and heuristic multi-robot dubins coveragepath planning for known environments[J]. Sensors, 2023, translated as Li L, Shi D, JinS, etc., Exact and heuristic multi-robot dubins coveragepath planning for known environments, 2023 Sensors) abstract the target area into a graph or tree structure, and then compress the solution space by cutting some edges that are unlikely to appear in the optimal solution, thereby reducing the calculation time. Heuristic methods can solve quickly, but will affect the accuracy of the solution. They are generally used to solve approximate optimal results in large-scale scenarios. In addition, some exact algorithms and heuristic algorithms only consider how to evenly divide the area and then plan the path. However, the balanced area partition result is not equivalent to a balanced coverage path.

[0004] Based on the current problems in the coverage path planning of multiple Dubins robots, how to plan a balanced, efficient and smooth coverage path for multiple Dubins robots subject to dynamic constraints in a short period of time is a technical problem that urgently needs to be solved in the field of collaborative coverage of multiple Dubins robots. Summary of the invention

[0005] In view of the problems existing in the prior art, the purpose of the present invention is to provide a multi-Dubins robot coverage path planning method based on competitive game. The method can plan a balanced, efficient and robot kinematically constrained coverage path for multiple Dubins robots within a limited time.

[0006] To achieve the above object, the present invention provides a multi-Dubins robot collaborative coverage path planning method based on competitive game, the method comprising the following steps:

[0007] The first step is to build a multi-Dubins robot coverage path planning system, which includes a client, a server and K Dubins robots, where K is a positive integer; the server includes a region decomposition module, an initial division module, a task fine-tuning module based on a credit model, and a task fine-tuning module based on a game strategy;

[0008] In the second step, the client collects target area information through detection sensors and generates a corresponding two-dimensional binary unit map MAP, wherein the target area information includes the area, obstacles and / or boundaries of the target area; the client sends the two-dimensional binary unit map MAP and the sensor coverage width width to the region decomposition module, sends the starting point SP of the K robots to the initial division module, sends the maximum number of iterations of the first algorithm Iter1 (the default value is 200) to the task fine-tuning module based on the credit model, and sends the maximum number of iterations of the second algorithm Iter2 (the default value is 50) and Time, the starting point SP of the K robots, the forward speed speed, and the turning radius radius to the task fine-tuning module based on the game strategy; MAP contains units, each unit corresponds to a small square in the target area, with a width of width. A unit with a value of 0 indicates a unit with obstacles, and a value of 1 indicates a unit to be covered. The algorithm information includes the first maximum number of iterations Iter1, the second maximum number of iterations Iter2, and the time Time for a single Dubins path solution required by the server. The client sends the robot information and , the target area information MAP and algorithm information Iter1, Iter2, Time are sent to the server;

[0009] The first maximum number of iterations iter1 and the second maximum number of iterations iter2 are used in the task fine-tuning module based on the credit model and the task fine-tuning module based on the game strategy, respectively. Both values ​​are positive integers, which are set by the user and can be different.

[0010] In the third step, the area decomposition module on the server side receives the two-dimensional binary unit map MAP and the coverage width width of the robot sensor sent by the client, and decomposes the target area into multiple rectangular units according to the Semi-BCD decomposition algorithm. These units form the initial coverage unit set C, and the initial coverage unit set C is sent to the initial division module on the server side;

[0011] In the fourth step, the initial partitioning module receives the initial coverage unit set C sent by the client, receives the starting points SP of the K robots from the area decomposition module, and divides the target area into K initial partitions . Get the partition set P = Get the cost set Cost corresponding to P, and send P and Cost to the task fine-tuning module based on the credit model;

[0012] Step 5: The task fine-tuning module based on the credit model receives the information sent by the initial division module. and , receiving the client's , according to the CDM algorithm, the partition set Adjust to get the adjusted partition set ,Will Send it to the task fine-tuning module based on game strategy;

[0013] Step 6: The task fine-tuning module based on the game strategy receives the Iter2 and Time sent by the client, and receives the credit model-based task fine-tuning module ,for In The partitions generate the corresponding Paths, get the path set , and Make adjustments to obtain the adjusted path set ,Will Each path is sent to Dubins robots;

[0014] In the seventh step, after the K robots receive their respective Dubins coverage paths from the game strategy-based task fine-tuning module, they move along these paths in parallel, perform tasks at the same time, and generate K motion trajectories; the K motion trajectories can completely cover the target area.

[0015] Furthermore, in the first step, the client stores three types of information: robot information, target area information and algorithm information; the robot information includes the starting position SP, forward speed speed, turning radius radius and coverage width width of the task sensor of K robots; the target area information includes a two-dimensional binary unit map MAP of the target area; the algorithm information includes the first maximum number of iterations Iter1, the second maximum number of iterations Iter2 and the time Time for a single solution of the Dubins path required by the server; the client calculates the binary map MAP by obtaining the area, obstacles and boundary information of the target area through the detection sensor, and sends it to the server; the detection sensor includes a lidar and / or a binocular camera.

[0016] Furthermore, in the second step, the two-dimensional binary unit map MAP contains The value of the unit with obstacles is 0, and the value of the unit to be covered is 1.

[0017] Furthermore, in the third step, the region decomposition module decomposes the target region into multiple rectangular units, and these units form an initial covering unit set C. The specific method is:

[0018] According to the two-dimensional binary unit map MAP, the target area is decomposed into Z rectangular units using the Semi-BCD decomposition algorithm. , Z is a positive integer. The width of a single rectangular unit is equal to the robot's coverage area , the length is determined by the boundary of the target area and the boundary of the obstacle. These Z rectangular units constitute an initial coverage set , , where 1≤z≤Z, each covering unit in the initial covering unit set corresponds to a covering task, represent The zth task in the process; the regional decomposition module sends the initial covering unit set C to the initial partitioning module.

[0019] Z covering units can also be called Z rectangular units, which means that after the regional decomposition stage, a total of Z rectangular units are generated. Each rectangular unit is equivalent to a unit to be covered, that is, a task to be assigned. The starting unit is the first rectangular unit assigned to each region. Covering units and rectangular units have the same meaning. Rectangular units are used to emphasize that the units generated after regional decomposition are long rectangular strips.

[0020] Further, the specific implementation process of the fourth step is as follows:

[0021] S4.1 Perform graph abstraction on the initial cover set C to generate a connected graph , ,in is a set of vertices, ,in Corresponding to the zth covering unit in C ,| Representing Vertex The weight of the unit is equal to area; It is a collection of edges. , representing the vertex and The common boundary of , and 1≤i≤Z, 1≤j≤Z, ;

[0022] S4.2 Divide the target area into K partitions to form a partition set P. ,in Indicates partitions, and 1≤k≤K.

[0023] Further, the specific implementation process of step S4.2 is as follows:

[0024] S4.2.1 Order is empty, let the number of allocated units be num, let num = 0;

[0025] S4.2.2 assigns a starting unit to each partition in P and calculates the cost of the robot visiting these K partitions. The cost is recorded as ;

[0026] S4.2.3 Allocate unallocated units to , and update K robots to access The cost.

[0027] Further, the specific implementation process of step S4.2.2 is as follows:

[0028] S4.2.2.1 Let the sequence number of the initial partition be k = 1;

[0029] S4.2.2.2 Calculation of unit distances for equally spaced partitions , let the selected unit number be , +1, among which , The initial unit of a partition is , let the number of allocated units decrease by 1, that is, num = num-1, Round down;

[0030] S4.2.2.3 Calculating the number of robot visits to the kth partition The price ;

[0031] S4.2.2.4 If k = K, go to S4.2.3; otherwise, let k = k+1 and go to S4.2.2.2.

[0032] Further, the specific implementation process of step S4.2.3 is as follows:

[0033] S4.2.3.1 Order collection Indicates whether the partition is optional. Indicates that the partition is optional. Indicates that the partition is not selectable; All elements in are 1, so , 1 K;

[0034] S4.2.3.2 Find the partition with the lowest current cost and adjacent coverage units from the current cost set Cost , the cost of which is ;

[0035] S4.2.3.3 Compute the unassigned vertices in the set V that are Adjacent covering unit set ;

[0036] 4.2.3.4 Randomly select a set of adjacent covering cells Select a unit , and Assigned to , , let the number of allocated coverage units , let partition The number of covered units included ;

[0037] 4.2.3.5 Update robot access partition The price ,in Indicates the coverage unit The area will Add to the cost set Cost;

[0038] 4.2.3.6 Determine whether the number of allocated units num is equal to Z; if , P and Send to the task fine-tuning module based on the credit model and go to step 5; if num < Z, go to S4.2.3.2.

[0039] Furthermore, the specific implementation process of the fifth step is: using the tree partitioning strategy based on the credit model to initialize the partition set sent by the sub-module Adjust to obtain a set of partitions with approximately balanced areas , , ;Will Sent to the task fine-tuning module based on game strategy.

[0040] Furthermore, the specific implementation process of the sixth step is:

[0041] 6.1 The single robot Dubins planning algorithm is executed in parallel on the corresponding K partitions to obtain the path set and the cost set of K paths ,in is the length of the kth path; let the adjusted path set ,in ; Let the adjusted path cost set ,in, Corresponding partition Path cost; let the number of iterations L2 = 0;

[0042] 6.2 Based on Select the partitions that participate in the game and determine the partitions that may need to exchange tasks;

[0043] 6.3 Determine the partition and partitions Unit for task exchange ;

[0044] 6.4 Boot Partition and partitions Exchange tasks TC and update the path set and path cost set;

[0045] 6.5 Orders If L1 is equal to the second maximum number of iterations Iter2, the iteration ends and the path set is obtained. ,Will The corresponding K Dubins coverage paths are sent to the K Dubins robots corresponding to P' respectively, and go to step 7; otherwise, go to S6.2.

[0046] The following technical effects can be achieved by adopting the present invention:

[0047] 1. The first step of the present invention is to construct a multi-Dubins robot coverage path planning system, which can plan approximately optimal coverage paths for multiple Dubins robots within a limited time. These paths have the characteristics of smoothness and continuity, satisfying the kinematic constraints of the Dubins robots, thereby ensuring that the Dubins robots can execute smoothly.

[0048] 2. The second to sixth steps of the present invention decompose the target area into K approximately balanced partitions, each of which corresponds to a robot. Then, by adjusting the coverage paths corresponding to each partition, the coverage task time of the K robots is nearly balanced.

[0049] In order to verify the feasibility of the present invention, it is compared with the CDM algorithm. The results show that the present invention not only improves the solution accuracy, but also can achieve a more balanced coverage task allocation. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 It is an overall flow chart of the multi-Dubins robot collaborative coverage path planning method based on competitive game of the present invention;

[0051] Figure 2 The logical structure diagram of the homogeneous multi-robot system task planning system for full coverage application constructed for the first step of the present invention;

[0052] Figure 3 The experimental results of the coverage time comparison and task balance comparison between the present invention and the existing coverage algorithm CDM algorithm are as follows;

[0053] Figure 4 This is the result of the comparison experiment between the coverage time of the present invention and the existing coverage algorithm CDM algorithm;

[0054] Figure 5 This is the experimental result of comparing the task balance between the present invention and the existing coverage algorithm CDM algorithm. DETAILED DESCRIPTION

[0055] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0056] In the description of the present invention, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc., indicating the orientation or positional relationship, are based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", and "third" are used for descriptive purposes only, and cannot be understood as indicating or implying relative importance.

[0057] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0058] The following combination Figure 1-Figure 5 The specific embodiments of the present invention are described in detail. It should be understood that the specific embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0059] like Figure 1 As shown, the multi-Dubins robot collaborative coverage path planning method based on competitive game of the present invention includes the following steps:

[0060] The first step is to build a multi-Dubins robot coverage path planning system. The multi-Dubins robot coverage path planning system consists of a client, a server, and K Dubins robots, where K is a positive integer.

[0061] The client is connected to the server, and the client can be a server, PC or mobile terminal. The client stores three types of information: robot information, target area information and algorithm information. Robot information includes the starting position SP, forward speed speed, turning radius radius and coverage width width of the task sensor of K robots. The target area information includes the two-dimensional binary unit map MAP of the target area. The detection sensors such as laser radar and binocular camera on the client calculate the two-dimensional binary unit map MAP by obtaining the area, obstacles and boundary information of the target area, and send it to the server. MAP contains Units, each unit is a small square with a width of width. Units with a value of 0 represent units with obstacles, and units with a value of 1 represent units to be covered. The algorithm information includes the first maximum number of iterations Iter1, the second maximum number of iterations Iter2, and the time Time for a single solution of the Dubins path. The client sends the robot information SP, speed, radius and width, the target area information MAP, and the algorithm information Iter1, Iter2, and Time to the server.

[0062] The server is connected to the client and K robots. The server is a server, which includes a region decomposition module, an initial division module, a task fine-tuning module based on a credit model, and a task fine-tuning module based on a game strategy.

[0063] The region decomposition module is connected to the client and the initial division module. The region decomposition module is used to receive the two-dimensional binary unit map MAP and the robot coverage width width sent by the client, and decompose the target area. After decomposition, the initial coverage unit set C is obtained, each of which is a rectangle with a width of width and a length determined by the boundary of the target area and the obstacle boundary. Then, the region decomposition module sends the initial coverage unit set C to the initial division module.

[0064] The initial partitioning module is connected to the client, the regional decomposition module, and the task fine-tuning module based on the credit model. The initial partitioning module is used to receive the starting positions SP of the K robots from the client and the initial coverage unit set C from the regional decomposition module. The initial partitioning module divides the initial coverage unit set C into K partitions, each of which contains a number of coverage units, and puts these K partitions into the partition set P, which is expressed as ,in express The , According to the area of ​​the partition, the initial partition module generates the cost set Cost corresponding to each partition, and sends the partition set P and its corresponding Cost to the task fine-tuning module based on the credit model.

[0065] The task fine-tuning module based on the credit model is connected to the client, the initial partitioning module, and the task fine-tuning module based on the game strategy. It receives the partition set P and the cost set Cost corresponding to each partition from the initial partitioning module, receives the maximum number of algorithm iterations Iter1 from the client, and adjusts the partition set P to obtain the adjusted fine-tuning partition set. ,in yes After the partitions are adjusted, the task fine-tuning module based on the credit model will fine-tune the partition set Sent to the task fine-tuning module based on game strategy.

[0066] The task fine-tuning module based on game strategy is connected to the client and the task fine-tuning module based on credit model, receives the maximum number of algorithm iterations Iter2 and the time limit Time for solving a single Dubins covering path algorithm from the client, and receives the fine-tuning partition set from the task fine-tuning module based on credit model. , fine-tune partition set calculations There are K Dubins paths corresponding to the K partitions in the , and the K Dubins paths constitute the path set RD= , and adjust the path set RD to obtain the adjusted path set . .in Yes The adjusted path sets the adjusted path Sent to K robots.

[0067] K isomorphic Dubins robots are connected to a game strategy-based task fine-tuning module. A Dubins robot receives K Dubins paths from the game strategy-based task fine-tuning module and performs coverage in parallel along the planned K paths.

[0068] In the second step, the client collects information about the target area, such as the area, obstacles, and boundaries of the target area through detection sensors, and generates a corresponding two-dimensional binary unit map MAP. Units, the unit value of the unit with obstacles is 0, and the unit value to be covered is 1. The starting point SP of the robot is specified, the speed speed of the robot and the minimum turning radius radius are obtained by querying the robot manual, the sensor coverage width width of the robot is obtained by querying the sensor manual, and the maximum number of iterations of the algorithm Iter1, Iter2 and Time (Iter1, Iter2 and Time are all integers, the default value of Iter1 is set to 200, the default value of Iter2 is set to 50, and the default value of Time is set to 20s) are set by the user. The client sends the two-dimensional binary unit map MAP and the sensor coverage width width to the regional decomposition module, sends the starting point SP of K robots to the initial division module, sends Iter1 to the task fine-tuning module based on the credit model, and sends Iter2 and Time, the starting point SP of K robots, the forward speed speed, and the turning radius radius to the task fine-tuning module based on the credit game strategy.

[0069] In the third step, the region decomposition module receives the two-dimensional binary unit map MAP and the coverage width width of the robot sensor from the client. The task of the region decomposition module is to decompose the target area into multiple rectangular units, and these units form the initial coverage unit set C, and send the initial coverage unit set C to the initial division module. The specific method is:

[0070] First, according to the two-dimensional binary unit map MAP, the target area is decomposed into several rectangular units using the Semi-BCD decomposition algorithm. The width of the rectangular unit is equal to the coverage of the robot. , the length is determined by the boundary of the target area and the boundary of the obstacle. All rectangular units constitute an initial coverage unit set , , where Z is a positive integer and 1≤z≤Z. Each covering unit in the initial covering unit set corresponds to a covering task. represent The region decomposition module sends the cover set C to the initial partitioning module.

[0071] In the fourth step, the initial partitioning module receives the starting points SP and initial coverage unit set C of the K robots from the client and the region decomposition module respectively, and then divides the target area into K initial partitions, represented by the partition set P, and calculates the cost set Cost corresponding to the K partitions. The initial partitioning module sends P and Cost to the task fine-tuning module based on the credit model. The specific method is as follows:

[0072] S4.1 According to the initial coverage unit set C, the corresponding connectivity graph of the target area is given , ,in is a set of vertices, ,in Corresponding to the zth covering unit in C , Representing Vertex The weight of the unit is equal to area. It is a collection of edges. , representing the vertex and The common boundary of , and 1≤i≤Z, 1≤j≤Z, .

[0073] S4.2 Divide the target area into K partitions to form a partition set P. ,in Indicates partitions, and 1≤k≤K. The specific method is as follows:

[0074] S4.2.1 Let the partition set is empty, let the number of allocated units be num, and let num = 0.

[0075] S4.2.2 assigns a starting unit to each partition in the partition set P and calculates the cost of the robot accessing these K partitions. The costs are recorded as , the specific operations are as follows:

[0076] S4.2.2.1 Let the sequence number of the initial partition be k = 1.

[0077] S4.2.2.2 Calculation of unit distances for equally spaced partitions = ⌊ , let the selected unit number be , +1, among which , The initial unit of a partition is , let the number of allocated units decrease by 1, that is, num = num-1, Round down.

[0078] S4.2.2.3 Calculating the number of robot visits to the kth partition The price .

[0079] S4.2.2.4 If k = K, go to S4.2.3; otherwise, let k = k + 1 and go to S4.2.2.2.

[0080] S4.2.3 Allocate unallocated units to , and update K robots to access The specific operation is as follows:

[0081] S4.2.3.1 Order collection Indicates whether the partition is optional. Indicates that the partition is optional. Indicates that the partition is not selectable. All elements in are 1, that is , .

[0082] S4.2.3.2 Find the partition with the lowest current cost and adjacent coverage units from the current cost set Cost , the cost of which is .

[0083] S4.2.3.3 Compute the unassigned vertices in the set V that are Adjacent covering unit set The specific operations are as follows:

[0084] S4.2.3.3.1 Order is empty, let Represents a partition The number of covered units included, let the sequence number .

[0085] S4.2.3.3.2 Compute all unassigned vertices in the set V that are Middle Covering Unit Adjacent units constitute a set of adjacent units . Define the coverage unit and adjacent if and only if the partition There are covering units in ,and and There is an edge between . Where a and b are both positive integers and satisfy .

[0086] S4.2.3.3.3 Order

[0087] S4.2.3.3.4 If ,judge Is it empty? If is empty, let , go to S ; Otherwise, switch to S ;if ,make , turn S .

[0088] S4.2.3.4 Randomly select a set of adjacent covering cells Select a unit , and Assigned to , , let the number of allocated coverage units , let partition The number of covered units included .

[0089] S4.2.3.5 Update robot access partition The price ,in Indicates the coverage unit The area will Add to the Cost collection.

[0090] S4.2.3.6 Determine whether the number of allocated units num is equal to Z (Z represents the total number of coverage units). , P and Send to the task fine-tuning module based on the credit model and go to step 5; if num < Z, go to S4.2.3.2.

[0091] Step 5: The task fine-tuning module based on the credit model receives the and , received from the client , and for the partition set Adjust, get , then Send it to the task fine-tuning module based on game strategy. The specific method is:

[0092] The partition set sent by the initialization module is initialized using a tree partitioning strategy based on the credit model. Adjust to obtain a set of partitions with approximately balanced areas , , .Will Sent to the task fine-tuning module based on game strategy.

[0093] Step 6: Receive Iter2 and Time from the client and receive them from the task fine-tuning module based on the credit model. ,for In The partitions generate the corresponding Paths, get the path set , and Make adjustments to obtain the adjusted path set ,Will Each path is sent to A Dubins robot. The specific method is:

[0094] S6.1 The single robot Dubins planning algorithm is executed in parallel on the corresponding K partitions to obtain a set of paths. and the cost set of K paths ,in is the length of the kth path. Let the adjusted path set ,in Let the adjusted path cost set ,in, Corresponding partition The path cost is . Let the number of iterations L2 = 0.

[0095] S6.2 According to Select the partitions that participate in the game, that is, the partitions that may need to exchange tasks. The specific method is as follows:

[0096] S6.2.1 From Select the partition with the largest path cost As a player participating in the game, the corresponding maximum cost is .

[0097] S6.2.2 Computing Partition Sets middle, and Y adjacent partitions, these Y partitions constitute the adjacent partition set AD={ }, where Y is a positive integer that satisfies . Define the partitions and partitions ( ,) are adjacent if and only if there is at least one pair of covering cells and ,in , ,satisfy and There is an edge between . Using the quick sort algorithm, the Y partitions in AD are sorted in ascending order according to the path cost. After sorting, the partition set ADV={ } and ADV corresponding path cost set ,satisfy ,make .

[0098] S6.2.3 Select the first partition in ADV as The partition that performs negotiation task exchange is recorded as the minimum adjacent partition . It means that the current ADV is neutralized The partition for task coordination. This is the first partition in ADV. Because the partitions in ADV are arranged in ascending order, the first one is the smallest, expressed as .

[0099] S6.3 Determine the partition and partitions Unit for task exchange The specific method is:

[0100] S6.3.1 Order partition The number of covered units included is MG (MG is an integer and satisfies ),Right now ,in For partition Middle Covering units, . Order partition The number of covered units included is NG (NG is an integer and satisfies ),Right now For partition Middle Covering units, . Define the coverage unit With partition adjacent if and only if the partition Storage Unit ,and and There is an edge between .

[0101] S6.3.2 Order partition Included coverage units and partitions The number of adjacent cells is ( ),this The covering cells constitute the adjacent cell set , , RG is a positive integer, .

[0102] S6.3.4 From adjacent unit collection The covering unit TC to be exchanged is calculated in the following way:

[0103] S6.3.4.1 Number of partitions . Among them, y is a variable that records The adjacent partition set ADV is neutralized The number of partitions that have undergone task negotiation. That is, if the first adjacent partition is successfully negotiated, the next cycle can be entered; if all adjacent partitions fail to negotiate, that is, yy=RG, it means balance and the cycle can be ended.

[0104] S6.3.4.2 Select candidate switching unit cc as The first unit of .

[0105] S6.3.4.3 Order partition After the task cc is transferred, the corresponding partition is , the corresponding covering unit set is , the corresponding coverage path is , let partition After receiving the task cc, the corresponding partition is The corresponding set of covering units is , the corresponding coverage path is .

[0106] S6.3.4.4 During the calculation of task exchange, the candidate exchange unit cc is the partition of the task to be released. The effect and , , the candidate switching unit cc is for the partition of the receiving task The effect and , .

[0107] S6.3.4.5 Determine whether the candidate switching unit cc can reduce the longest path cost or at least reduce the path cost of one partition without increasing the longest path cost, that is, after the candidate switching unit cc is exchanged, the partition and partitions The maximum value of the corresponding two path costs decreases, or after exchanging cc and partitions The maximum of the two corresponding path costs remains unchanged, while the other value becomes smaller, the method is:

[0108] S6.3.4.5.1 Judgment Is it greater than or equal to 0? If it is greater than or equal to 0, cc will be exchanged to the partition After that, partition The path cost does not increase, so we need to judge Is it less than or equal to 0? If it is less than or equal to 0, it means that the longest path cost has decreased, so the exchange coverage unit , go to 6.4; otherwise, from the adjacent set Remove cc and go to S6.3.4.5.2.

[0109] if Less than 0, judge Is it less than If yes, judge Is it less than 0? If so, it means the longest path cost is reduced. , go to S6.4; otherwise, from the adjacent unit set Remove cc and go to S6.3.4.5.2.

[0110] S6.3.4.5.2 Judgment Is it equal to RG? If so, determine whether y is equal to the number of adjacent partitions Y. If , indicating that equilibrium is reached. The corresponding K Dubins coverage paths are sent to the K Dubins robots corresponding to P', and then go to step 7; otherwise, remove the partition from ADV , go to S6.2.3; if Not equal to RG, let , go to S6.3.4.2.

[0111] S6.4 Boot Partition and partitions Exchange tasks TC and update the path set and path cost set. The method is:

[0112] S6.4.1 Update and fine-tune the partition set Central Division and partitions The corresponding set of covering units is ,make .

[0113] S6.4.2 Update the path set corresponding to K partitions Central Division and partitions The corresponding path is ,make .

[0114] S6.4.3 Update cost set Central Division and partitions The corresponding path cost is ,make .

[0115] S6.5 Order If L1 is equal to the second maximum number of iterations Iter2, the iteration ends and the path set is obtained. ,Will The corresponding K Dubins coverage paths are sent to the K Dubins robots corresponding to P' respectively, and go to step 7; otherwise, go to S6.2.

[0116] In the seventh step, after receiving their respective Dubins coverage paths from the game strategy-based task fine-tuning module, the K robots move along these paths in parallel, perform tasks simultaneously, and generate K motion trajectories that can completely cover the target area.

[0117] In order to verify the effect of the present invention, the following experiments were performed:

[0118] The experiment was completed on a PC configured as AMD Ryzen 7 7840H with Radeon 780M Graphics 3.80 GHz. The software platform of the experiment is Visual Studio 2019, and MATLAB is used to display the covered path trajectory. In the experiment, the forward speed of the Dubins robot is set to 1m / s, the minimum turning radius is 0.5m, the starting point SP is set to the origin (i.e. (0, 0)), the task sensor coverage width of the robot is set to 1 meter, the first maximum number of iterations is set to 200 by default, the second maximum number of iterations is set to 50 by default, and the time for solving a single Dubins path is set to 20s by default. The appropriate size can also be set according to the scale of the target area. The experimental scene is 13 two-dimensional target areas. The length and width of the four small scenes are 10m and 10m respectively. The length and width of the remaining scenes are (a) 34m, 50m; (b) 34m, 45m; (c) 25m, 38m; (d) 25m, 38m; (e) 20m, 30m; (f) 20m, 30m; (g) 30m, 40m; (h) 30m, 40m; (i) 40m, 50m. These scenes all contain multiple obstacles of inconsistent shapes and sizes. Set the algorithm iteration number Iter1 to 100 and Iter2 to 50. In the scenario of , the number of test robots is 2 and 3, and the number of robots in other scenarios is 3, 6, 9, 12, 15, and 18 respectively. The coverage time (in seconds) and task balance of the present invention and the latest and best CDM algorithm with similar research direction are compared. The experimental results of the four small scenarios are shown in Figure 3 shown. Figure 3 In (a), the horizontal axis is the scene name, the vertical axis is the longest coverage time (in seconds), the solid line is the comparison result when the number of robots is 2, and the dotted line is the comparison result when the number of robots is 3. Figure 4 The experimental results of the other 9 scenarios are shown. Figure 4 (a)-(f) represent the results when the number of robots is 3, 6, 9, 12, 15, and 18, respectively. Figure 4 In (a)-(f), the horizontal axis is the name of the scene, and the vertical axis is the longest coverage time (in seconds). The longest coverage time refers to the time required for all robots to follow the planned path to perform the coverage task. The performance of the robot in performing the coverage task is closely related to its coverage time, and the coverage time is affected by the number of robots and the length of the planned path. Specifically, when the number of robots increases, the area that each robot needs to cover is reduced accordingly, resulting in a reduction in coverage time. On the other hand, the shorter the planned path, the less time the robot needs to perform the coverage task. Therefore, the reduction of coverage time is an important indicator for evaluating the performance improvement of the method. It can be seen from the experimental results that only when the number of robots is 12, the (c) scene and the number of robots are 18, the longest coverage time of scene (b) is the same as that of the CDM algorithm. In other cases, the coverage time of the present invention is shorter than that of the CDM algorithm. Therefore, with respect to the coverage time indicator, the present invention has a great advantage over CDM.

[0119] Figure 3 (b) shows the comparison of task balance between the present invention and the CDM algorithm in four small scenarios when the number of robots is 2 and 3. Figure 3 In (b), the horizontal axis is the scene name, and the vertical axis is the coefficient of variation (the coefficient of variation is the ratio of the standard deviation to the mean, which takes into account the degree of dispersion of the data and performs standardization. The larger the coefficient of variation, the higher the degree of dispersion of the data and the lower the degree of balance). The solid line is the comparison result when the number of robots is 2, and the dotted line is the comparison result when the number of robots is 3. Figure 5 (a)-(f) represent the results when the number of robots is 3, 6, 9, 12, 15, and 18, respectively. Figure 4The horizontal axis of (a)-(f) is the name of the scene, and the vertical axis is the coefficient of variation. In a multi-machine system, the energy and task execution ability of each robot are fixed. For multiple homogeneous robots, by evenly distributing tasks to multiple machines, it can be ensured that each machine is fully utilized, avoiding overload of some machines while other machines are idle, thereby improving the resource utilization of the entire system. Therefore, in a system composed of multiple homogeneous robots, the more balanced the task distribution, the better. It can be seen from the experimental results that, except when the number of robots is 12, the (b) scene and the number of robots is 15, the (d) scene has a lower task balance than the CDM algorithm. In other cases, the task allocation results of the present invention are more balanced than the CDM algorithm. Therefore, in terms of balanced task distribution, the present invention has more advantages than CDM.

[0120] Any process or method description in the flowchart of the present invention or described in other ways herein can be understood as a module, segment or part of a code including one or more executable instructions for implementing the steps of a specific logical function or process, which can be implemented in any computer-readable medium for use by an instruction execution system, device or equipment, and the computer-readable medium can be any medium containing storage, communication, propagation or transmission programs for use by execution systems, devices or equipment, including read-only memories, magnetic disks or optical disks, etc.

[0121] In the description of this specification, the description with reference to the terms "embodiment", "example", etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. In addition, those skilled in the art can combine or combine different embodiments or examples described in this specification and the features therein without causing any contradiction.

[0122] Although the above content has shown and described the embodiments of the present invention, it can be understood that the above embodiments are exemplary and cannot be understood as limitations of the present invention. Ordinary technicians in the field can change, modify, replace, modify and other update operations on the above embodiments within the scope of the present invention.

Claims

1. A multi-Dubins robot collaborative coverage path planning method based on competitive game, characterized in that: The method comprises the following steps: The first step is to build a multi-Dubins robot coverage path planning system, which includes a client, a server and K Dubins robots, where K is a positive integer; the server includes a region decomposition module, an initial division module, a task fine-tuning module based on a credit model, and a task fine-tuning module based on a game strategy; In the second step, the client collects target area information through detection sensors and generates a corresponding two-dimensional binary unit map MAP, wherein the target area information includes the area, obstacles and / or boundaries of the target area; the client sends the two-dimensional binary unit map MAP and the sensor coverage width width to the region decomposition module, sends the starting point SP of the K robots to the initial division module, sends the maximum number of iterations of the first algorithm Iter1 to the task fine-tuning module based on the credit model, and sends the maximum number of iterations of the second algorithm Iter2 and Time, the starting point SP of the K robots, the forward speed speed, and the turning radius radius to the task fine-tuning module based on the game strategy; The third step is to decompose the target area into multiple rectangular units according to the Semi-BCD decomposition algorithm using the two-dimensional binary unit map MAP from the client and the coverage width width of the robot sensor. These units form the initial coverage unit set C, and the initial coverage unit set C is sent to the initial division module of the server. Step 4: The initial partitioning module receives the starting points SP of the K robots from the client and the region decomposition module, divides the target area into K initial partitions, represented by the partition set P, and calculates the cost set Cost corresponding to the K partitions, and sends the partition set P and the cost set Cost to the task fine-tuning module based on the credit model; Step 5: The task fine-tuning module based on the credit model receives the partition set P and the cost set Cost from the initial partition module and receives the , according to the CDM algorithm, the partition set Adjust to get the adjusted partition set , the adjusted partition set Send it to the task fine-tuning module based on game strategy; Step 6: Receive the maximum number of iterations Iter2 and Time of the second algorithm from the client, and receive the adjusted partition set from the task fine-tuning module based on the credit model , is the adjusted partition set In The partitions generate the corresponding Paths, get the path set , and for the path set Make adjustments to obtain the adjusted path set , the adjusted path set Each path is sent to Dubins robots; In the seventh step, after the K robots receive their respective Dubins coverage paths from the game strategy-based task fine-tuning module, they move along these paths in parallel, perform tasks at the same time, and generate K motion trajectories; the K motion trajectories can completely cover the target area.

2. The method for collaborative coverage path planning of multiple Dubins robots based on competitive game according to claim 1 is characterized in that: In the first step, the client stores three types of information: robot information, target area information and algorithm information; the robot information includes the starting position SP, forward speed speed, turning radius radius and coverage width width of the task sensor of K robots; the target area information includes the two-dimensional binary unit map MAP of the target area; the algorithm information includes the first maximum number of iterations Iter1, the second maximum number of iterations Iter2 and the time Time for a single solution of the Dubins path required by the server; the client calculates the binary map MAP by obtaining the area, obstacles and boundary information of the target area through the detection sensor, and sends it to the server; the detection sensor includes a lidar and / or a binocular camera.

3. The method for collaborative coverage path planning of multiple Dubins robots based on competitive game according to claim 1, characterized in that: In the second step, the two-dimensional binary cell map MAP contains The value of the unit with obstacles is 0, and the value of the unit to be covered is 1.

4. The method for collaborative coverage path planning of multiple Dubins robots based on competitive game according to claim 1, characterized in that: In the third step, the region decomposition module decomposes the target region into multiple rectangular units, and these units form the initial coverage unit set C. The specific method is: According to the two-dimensional binary unit map MAP, the target area is decomposed into several rectangular units using the Semi-BCD decomposition algorithm. The width of the rectangular unit is equal to the coverage of the robot. , the length is determined by the boundary of the target area and the boundary of the obstacle; all rectangular units constitute an initial coverage unit set , , where 1≤z≤Z, each unit in the initial coverage unit set corresponds to a coverage task, represent The zth task in .

5. The method for collaborative coverage path planning of multiple Dubins robots based on competitive game according to claim 1, characterized in that: The specific implementation process of the fourth step is as follows: S4.1 According to the initial coverage set C, the corresponding connected graph of the target area is given , ,in is a set of vertices, ,in Corresponding to the zth covering unit in C , Representing Vertex The weight of the unit is equal to area; It is a collection of edges. , representing the vertex and The common boundary of , and 1≤i≤Z, 1≤j≤Z, ; S4.2 Divide the target area into K partitions to form a partition set P. ,in Indicates partitions, and 1≤k≤K.

6. The method for collaborative coverage path planning of multiple Dubins robots based on competitive game according to claim 5, characterized in that: The specific implementation process of step S4.2 is as follows: S4.2.1 Order is empty, let the number of allocated units be num, let num = 0; S4.2.2 assigns a starting unit to each partition in P and calculates the cost of the robot visiting these K partitions. The cost is recorded as ; S4.2.3 Allocate unallocated units to , and update K robots to access The cost.

7. The method for collaborative coverage path planning of multiple Dubins robots based on competitive game according to claim 6, characterized in that: The specific implementation process of step S4.2.2 is as follows: S4.2.2.1 Let the sequence number of the initial partition be k = 1; S4.2.2.2 Calculation of unit distances for equally spaced partitions , let the selected unit number be , +1, among which , Order The initial unit of a partition is , reduce the number of allocated units by 1, so that num = num-1, Round down; S4.2.2.3 Calculating the number of robot visits to the kth partition The cost ; S4.2.2.4 If k = K, go to S4.2.3; otherwise, let k = k+1 and go to S4.2.2.

2.

8. The method for collaborative coverage path planning of multiple Dubins robots based on competitive game according to claim 6, characterized in that: The specific implementation process of step S4.2.3 is as follows: S4.2.3.1 Order collection Indicates whether the partition is optional. Indicates that the partition can be selected. Indicates that the partition cannot be selected; make All elements in are 1, so , ; S4.2.3.2 Find the partition with the lowest current cost and adjacent coverage units from the current cost set Cost , the cost of which is ; S4.2.3.3 Compute the unassigned vertices in the set V that are Adjacent covering unit set ; 4.2.3.4 Randomly select a set of adjacent covering cells Select a unit , and Assigned to , , let the number of allocated coverage units , partition The number of covered cells included ; 4.2.3.5 Update robot access partition The cost ,in Indicates the coverage unit The area will Add to the cost set Cost; 4.2.3.6 Determine whether the number of allocated units num is equal to Z; if , P and Send to the task fine-tuning module based on the credit model and go to step 5; if num < Z, go to S4.2.3.

2.

9. The method for collaborative coverage path planning of multiple Dubins robots based on competitive game according to claim 1, characterized in that: The specific implementation process of the fifth step is: use the tree partitioning strategy based on the credit model to initialize the partition set sent by the sub-module Adjust to obtain a set of partitions with approximately balanced areas , , ;Will Sent to the task fine-tuning module based on game strategy.

10. The method for collaborative coverage path planning of multiple Dubins robots based on competitive game according to claim 1, characterized in that: The specific implementation process of the sixth step is: 6.1 The single robot Dubins planning algorithm is executed in parallel on the corresponding K partitions to obtain the path set and the cost set of K paths ,in is the length of the kth path; Let the adjusted path set ,in ; Let the adjusted path cost set ,in, Corresponding partition Path cost; let the number of iterations L2 = 0; 6.2 Based on Select the partitions participating in the game and determine the partitions that need to exchange tasks; 6.3 Determine the partition and partitions Unit for task exchange ; 6.4 Boot Partition and partitions Exchange tasks TC and update the path set and path cost set; 6.5 Orders If L1 is equal to the second maximum number of iterations Iter2, it means that the iteration is over and the path set is obtained. ,Will The corresponding K Dubins coverage paths are sent to the K Dubins robots corresponding to P' respectively, and go to step 7; otherwise, go to S6.2.

Citation Information

Cited By

  • Multi-Dubins robot collaborative coverage path planning method based on tree division

    CN118409597A

  • Two-vehicle target defense intelligent game playing method based on reachable set

    CN121069999A