Regional coverage path planning method based on dynamic networking under local communication

Abstract

The method comprises the following steps: initializing the positions of N robots which can communicate with each other, and setting a virtual position outside the coverage area; in each iteration period, all robots move to the next target neighbor point synchronously; determining whether all robots have moved, if yes, resetting the movement state of all robots, and ending the iteration period; otherwise, determining whether there is a robot trapped in a "deadlock", if yes, releasing the "deadlock", and ending the iteration period when all robots trapped in the "deadlock" have moved to the target neighbor point. The above robot movement iteration period is repeated until the iteration number reaches a preset value or all environment areas are covered. The method can effectively respond to the change of the environment area to be covered in real time in a communication-limited obstacle-free environment, and realize complete coverage of the area at a minimum cost.

Description

Technical Field

[0001] This invention relates to the field of collaborative execution of area coverage path planning by swarm robots, specifically a method for swarm robots to achieve collaborative area coverage path planning through dynamic networking in the case of local communication and only local communication. Background Technology

[0002] In real-world applications, autonomous navigation robots can perform various tasks, such as environmental exploration, disaster search and rescue, goods handling, and warehouse management. These tasks all require the robot to autonomously plan its movement path, reach the target location according to task requirements, and perform the corresponding operations. Path planning algorithms are essential for the planning of movement paths for autonomous navigation robots. From a task classification perspective, a path planning algorithm covering the entire target area aims to plan the shortest path between the starting point and the target point. Building upon path planning, the focus then shifts to optimizing path costs.

[0003] Coverage path planning is an intelligent optimization algorithm, a derivative of the traveling salesman problem. It utilizes autonomous robot systems, equipped with specific sensors or communication devices, to autonomously navigate and cover areas in both unknown and known environments. Since its inception, it has been the subject of extensive research in academia and widespread application in industry.

[0004] In the area coverage path planning problem for individual robots, current technologies combine various strategies such as the Traveling Salesman Algorithm (TSP), Ant Colony Algorithm (ACO), Genetic Algorithm (GA), and K-Means algorithm to optimize the problem. However, when the search area is large, it is impossible to avoid practical problems such as low coverage efficiency and excessive power consumption. Therefore, cooperative coverage path planning methods for swarm robots have received attention from academia and industry. For example, in 2022, Xie et al. proposed an improved genetic algorithm to solve the energy-constrained CPP problem for multiple UAVs in multiple regions. In the same year, Ma et al. proposed an improved cooperative coverage BA algorithm. * Algorithm (CCIBA) * This paper addresses the collaborative coverage path planning problem for unmanned surface mapping vehicles (USMVs). In 2023, Cheng Xiaoming et al. introduced the collaborative region coverage achievement of swarm robots into the coverage detection problem of large 3D structures, proposing a new method for solving the coverage path planning problem based on an improved genetic algorithm.

[0005] Although the coverage path planning problem for swarm robots has been studied extensively, most existing coverage path planning algorithms do not have the ability to enable swarm robots to achieve complete coverage of a region with minimal path cost even under local communication conditions. Summary of the Invention

[0006] To address the aforementioned issues, this invention provides a method for collaborative regional coverage path planning by a group of robots through dynamic networking in the case of local communication and only local communication.

[0007] Technical principle of this invention:

[0008] Based on the predator-prey theory, in each iteration, all robots in the swarm synchronously move towards their next target neighbor. The selection of the target neighbor is calculated using a total reward function and communication constraint rules. The total reward function consists of three sub-reward functions: the first is related to the robot's movement away from the predator (a virtual point), controlling the robot's movement trend and avoiding inefficient backtracking; the second is related to the robot's continuous linear movement, ensuring it maintains a straight line as much as possible; and the third is related to the robot's ability to cover the boundary, further avoiding backtracking and deadlock. The total reward function acts as a heuristic, guiding the robot's movement direction at each step. Another guiding factor in the selection of the target neighbor is the local communication constraint rules among the swarm robots. Due to the limited communication range between robots, the algorithm constructs a dynamic network model to ensure that robots maintain network communication while exploring the next path target, thereby guaranteeing real-time information sharing among the robot swarm.

[0009] Technical solution of the present invention:

[0010] A method for regional coverage path planning based on dynamic networking under local communication includes the following steps:

[0011] Step 1: Establish the search environment and initialize the swarm of robots;

[0012] Step 2: Calculate the potential neighbor points for each robot;

[0013] Step 3: Based on the communication range of the already moved robot group, obtain selectable neighbor points for robots that have not yet moved;

[0014] Step 4: For each selectable neighbor point obtained in Step 3, calculate its predator avoidance reward;

[0015] Step 5: For the selectable neighbor points obtained in Step 3, calculate their linear motion rewards respectively;

[0016] Step 6: For the selectable neighbor points obtained in Step 3, calculate their boundary coverage rewards respectively;

[0017] Step 7: Calculate the total revenue of neighboring points based on the reward value components obtained in steps 4-6 respectively;

[0018] Step 8: For robots that have not yet moved in this iteration, based on the total revenue of each neighbor point calculated in Step 7, select the neighbor point corresponding to the maximum revenue value as the robot's moving target in the iteration cycle, and set the robot's moving state.

[0019] Step 9: In this iteration cycle, repeat steps 3-8 until all robots have moved or there are robots stuck in a deadlock.

[0020] Step 10: If all robots have moved, the iteration ends and the state of all robots is reset; if there are any deadlocked robots, the deadlock is resolved and the iteration ends.

[0021] Step 11: If the current iteration count reaches the preset value or all environmental areas have been covered, proceed to step 12; otherwise, proceed to step 2.

[0022] Step 12: Output the movement coverage path of each robot.

[0023] Compared with the prior art, the present invention has the following beneficial effects:

[0024] This invention enables swarm robots to achieve collaborative area coverage path planning through dynamic networking, even in situations with limited local communication. In obstacle-free environments with restricted communication, this method can effectively respond in real-time to changes in the area to be covered, achieving complete area coverage at minimal cost. Attached Figure Description

[0025] Figure 1 This is a flowchart illustrating the regional coverage path planning method based on dynamic networking under local communication according to the present invention.

[0026] Figure 2 This is the connected graph of Example 8.

[0027] Figure 3 This is a schematic diagram of the follow-escape mechanism in an embodiment. Detailed Implementation

[0028] This invention proposes a regional coverage path planning method based on dynamic networking under local communication, which can be applied to magnetic spatial measurement problems.

[0029] In large-scale magnetic spaces to be measured, commonly used swarm robots equipped with sensors or communication devices are subject to magnetic field interference, significantly reducing their communication range during measurement tasks. Therefore, collaborative area coverage path planning for swarm robots is limited to real-time information sharing under localized communication conditions. We now assume that the robots are much smaller than the magnetic space and have a limited local sensing range. Furthermore, the robots themselves possess no prior knowledge of the magnetic space and can only obtain relevant information from other robots within the communication network through localized communication and sensors, while ensuring the efficiency of the measurement task and achieving complete area coverage with minimal path cost.

[0030] The technical solutions provided in this application will be further described below with reference to specific embodiments and accompanying drawings. The advantages and features of this application will become clearer from the following description.

[0031] Example

[0032] like Figure 1 As shown, the regional coverage path planning method based on dynamic networking under local communication includes the following steps:

[0033] Step 1: Establish the search environment. Divide the magnetic measurement space using the dynamic grid method. Set a virtual reference position ψ outside the magnetic space. Initialize the algorithm parameters: number of robots N, robot communication distance C, and weight factor ω. p ω b ω s .

[0034] Initialize the robot's position o k ,k∈{1,2,...,N}, and it is guaranteed that they can communicate with each other through AD-HOC.

[0035] Step 2: As Figure 2 As shown, let the current position of any robot k be o. k The neighboring target points are denoted as N8, including the four-connected point N4 (i.e., target regions 2, 4, 5, and 7), and their upper-left, lower-left, upper-right, and lower-right neighbor regions. The candidate neighbor points for each robot are calculated using the following formula:

[0036]

[0037] Where x and y are the current coordinates of robot k.

[0038] Step 3: Obtain selectable neighbor points of the robot that has not yet moved. Let P(o m P(o) represents the location of the already moved robot group, and the neighbor point to be selected is P(o). m The shortest Euclidean distance is The calculation formula is as follows:

[0039]

[0040] Where i∈{-1,0,1}: when i=-1, j∈{-1,0,1}; when i=0, j∈{-1,1}; when i=1, j∈{-1,0,1}, o m For a robot that has already moved. If and only if When the target point is within the communication range of the already moved robot group, it is considered a selectable neighbor point.

[0041] Step 4: For each of the selected neighbor points, calculate its predator avoidance reward. The formula is as follows:

[0042]

[0043] Where r is the number of selectable neighbor points. For the preset maximum reward value component, D(o) j )=||o j -ψ|| is the neighbor's point o j The Euclidean distance to the predator ψ; similarly, D max (o k ) = max j ||o j -ψ|| represents the current point o k The maximum Euclidean distance to the predator among all its neighboring points, and D min (o k ) = min j ||o j -ψ|| represents the current point o k The minimum Euclidean distance to the predator among all its neighbors. The formula for calculating this Euclidean distance is:

[0044]

[0045] Step 5: For the selectable neighbor points obtained in Step 3, calculate their linear motion rewards respectively. The formula is as follows:

[0046]

[0047] Where r is the number of selectable neighbor points. For the preset maximum reward value component, o k-1 For the current point o k The previous historical position; o j Current position o k The j-th selectable neighbor point; ∠o k-1 o k o j ∈(0°,180°] is a vector (ok -o j ) and (o k-1 -o k The oblique angle formed by )

[0048] Step 6: For the selectable neighbor points obtained in Step 3, calculate their boundary cover reward. The formula is as follows:

[0049]

[0050] Where r is the number of selectable neighbor points. This is the preset maximum reward value component. o k The maximum possible number of neighbors of n is easily determined to be 8 based on the properties of an 8-connected graph; while n N (o j ) for o k The j-th neighbor point o j Number of uncovered neighbors.

[0051] Step 7: Calculate the reward value component R based on steps 4-6 respectively. d (o j ), R s (o j ) and R b (o j ), calculate neighbor point o j The total revenue. The formula is as follows:

[0052] R(o j )=ω p R d (o j )+ω s (R s (o j ))+ω b (R b (o j ))

[0053] Where, ω p ω s and ω b These are weighting factors for predator avoidance, linear movement, and boundary coverage rewards, respectively. The value of the weighting factor determines the degree of importance the forager attaches to different rewards when deciding on its next movement direction.

[0054] Step 8: For robots that have not yet moved in this iteration, based on the total revenue of each neighbor point calculated in Step 7, select the neighbor point corresponding to the maximum revenue value as the robot's moving target in the iteration cycle, and set the robot's moving state.

[0055] Step 9: In this iteration cycle, repeat steps 3-8 until all robots have moved or there are robots stuck in a deadlock.

[0056] Step 10: If all robots have moved, the iteration ends, and the state of all robots is reset to 0; if any robots are deadlocked, deadlock resolution is performed, and the iteration ends. The following specific cases are considered for deadlock resolution:

[0057] 1) Scenario 1: All robots have no selectable neighbor points, and all their neighbor points have been covered.

[0058] 2) Scenario 2: Some robots have no selectable neighbor points, and all of their neighbor points have been covered.

[0059] 3) Scenario 3: Although some robots have uncovered neighboring points around them, due to communication constraints, the movement direction of the robots they follow means that they can only move to a specific neighboring point in the same direction. However, the neighboring point has been covered by itself or other robots, resulting in the robot having no points to choose from and passively falling into a "deadlock" state.

[0060] For scenario one, use A. * The algorithm (commonly used in path planning) determines the appropriate backtracking point to the uncovered area. Specifically, it first sets the current position o... k Place it into the open list A, and repeat the following process: determine whether the area coverage is complete or the backtracking point has been reached. If yes, the algorithm ends; otherwise, calculate and obtain the current position o according to the formula in step 2. k All reachable neighbor points are added to the open list A. Then, the cost of reaching each target point is calculated according to the formula in step 7, and the target point with the minimum cost is taken as the current position o. k And add the target point to the closed list B. Finally, when the robot reaches position o m When its neighboring target points are not yet covered, at this time o m This is the backtracking point, and the sequence of target points in B is the movement path, thus completing the path planning.

[0061] For scenarios two and three, when some robots get stuck in a deadlock state, a follow-and-escape mechanism is used to reach the backtracking point as quickly as possible while maintaining uninterrupted communication. The specific implementation of the follow-and-escape mechanism is as follows: Figure 3As shown in the diagram, let the area formed by the points be the coverage area, and A′, B′, and C′ be three robots performing path planning tasks (C′ is the robot that is "deadlocked"). The cyan, red, and purple lines represent the historical movement trajectories of robots A, B, and C during the process of completing partial area coverage, respectively. The current coordinates of robots A, B, and C are A′(x, y, y) and A′(x, y, y). a ,y a ), B′(x b ,y b ) and C′(x c ,y c The dashed circle represents the communication range of each robot, and the cross-shaped icon outside the area (upper right corner of the square dot matrix area) represents the virtual predator's location. It is easy to see from the diagram that robot C's adjacent areas are now covered, causing C' to enter a "deadlock" state. To enable C' to quickly find a suitable backtracking point while maintaining uninterrupted communication, the following mechanism is as follows: First, based on the formula in step 7, prioritize calculating the optimal neighbor target point o for each of the "free" robots A' and B' that are not deadlocked. A′ o B′ And set them as the current positions respectively. Then, calculate the Euclidean distances between the deadlocked robots C′ and A′, and B′ respectively according to the Euclidean distance calculation formula in step 4. like At this point, robot C' should follow robot B', that is, using the previous position as a reference point, and move towards the neighboring target point in the same direction. This continues until robot C' reaches the appropriate backtracking point in the uncovered area.

[0062] Step 11: If the current iteration count reaches the preset value G or all environmental areas have been covered, then proceed to step 12; otherwise, proceed to step 2.

[0063] Step 12: Output the movement coverage path of each robot.

[0064] verify

[0065] To further explore the feasibility and performance of the proposed method, an experimental comparison was conducted, comparing it with two other adaptive coverage path planning methods based on predator theory: PP-CPP (Predator-Prey-Based Coverage Path Planning) and Dec-PPCPP (Decentralized Predator-Prey-Based Coverage Path Planning). Considering the multiple parameter settings involved in the experiment, six aspects were included in the parameter settings: weighting factor, number of robots, environment scale, environmental area variation, communication range, and initial robot position. The parameter settings are summarized in Table 1 below.

[0066] Table 1 Summary of Parameter Settings

[0067]

[0068]

[0069] Table 2 presents the experimental comparison results between the proposed method and the PP-CPP and Dec-PPCPP algorithms:

[0070] Table 2. Experimental Comparison Results

[0071]

[0072]

[0073] The above description is merely a description of preferred embodiments of this application and is not intended to limit the scope of this application in any way. Any changes or modifications made by those skilled in the art based on the above-disclosed technical content should be considered as equivalent and valid embodiments and fall within the scope of protection of the technical solution of this application.

Claims

1. A method for regional coverage path planning based on dynamic networking under local communication, characterized in that, Including the following steps: Step 1: Establish the search environment and initialize the swarm of robots; Step 2: Calculate the potential neighbor points for each robot; Step 3: Based on the communication range of the already moved robot group, obtain selectable neighbor points for robots that have not yet moved; Step 4: For each selectable neighbor point obtained in Step 3, calculate its predator avoidance reward; Step 5: For the selectable neighbor points obtained in Step 3, calculate their linear motion rewards respectively; Step 6: For the selectable neighbor points obtained in Step 3, calculate their boundary coverage rewards respectively; Step 7: Calculate the total revenue of neighboring points based on the reward value components obtained in steps 4-6 respectively; Step 8: For robots that have not yet moved in this iteration, based on the total revenue of each neighbor point calculated in Step 7, select the neighbor point corresponding to the maximum revenue value as the robot's moving target in the iteration cycle, and set the robot's moving state. Step 9: In this iteration cycle, repeat steps 3-8 until all robots have moved or there are robots stuck in a "deadlock" situation; Step 10: If all robots have moved, the iteration ends and the state of all robots is reset; if there are robots in a deadlock, the deadlock is resolved and the iteration ends. Step 11: If the current iteration count reaches the preset value or all environmental areas have been covered, proceed to step 12; otherwise, proceed to step 2. Step 12: Output the movement coverage path of each robot; In step 10, for robots that are trapped in a "deadlock", there are three possible scenarios: Scenario 1: All robots have no selectable neighbor points, and all their neighbor points have been covered; Scenario 2: Some robots have no selectable neighbor points, and all their neighbor points have been covered; Scenario 3: Although some robots have uncovered neighboring points around them, due to communication constraints, the movement direction of the robots they follow means that they can only move to a specific neighboring point in the same direction. However, this neighboring point has been covered by itself or other robots, resulting in the robot having no points to choose from and passively falling into a "deadlock" state. For scenario one, the A* algorithm is used to determine the appropriate backtracking point to the uncovered area. First, the current position is... Add to open list In the process, the following steps are repeated: determine whether the area coverage has been completed or the backtracking point has been reached; if so, the algorithm ends; otherwise, calculate and obtain the current position according to the formula in step 2. Add all reachable neighbor points to the open list Then, according to the formula in step 7, the cost of reaching each target point can be calculated separately, and the target point corresponding to the minimum cost is taken as the current position. And add the target point to the closed list. In the middle; finally, when the robot reaches the location When its neighboring target points are not yet covered, at this time That is, the backtracking point. The sequence of target points in the diagram constitutes the movement path, and path planning is complete. For scenarios two and three, when some robots get stuck in a deadlock state, a follow-and-escape mechanism is used to reach the backtracking point as quickly as possible while maintaining uninterrupted communication network.

2. The method for regional coverage path planning based on dynamic networking under local communication according to claim 1, characterized in that, Step 1 includes: Establish a search environment and set up virtual reference locations. Initialize algorithm parameters: number of robots The robot's communication distance Weighting factor ; Initialize the robot's position And ensure that they can communicate with each other through AD-HOC mode.

3. The method for regional coverage path planning based on dynamic networking under local communication according to claim 2, characterized in that, Step 2: Suppose any robot Current location is Its neighboring target points are represented as Including four connected points The region includes the top-left, bottom-left, top-right, and bottom-right neighbor regions; the candidate neighbor points for each robot are calculated using the following formula: in, For robots The current coordinates.

4. The method for regional coverage path planning based on dynamic networking under local communication according to claim 3, characterized in that, Step 3: set up The location of the already moved robot group, and the neighbor points to be selected. The shortest Euclidean distance is The calculation formula is as follows: in :when hour, ;when hour, ;when hour, , For a robot that has been moved; if and only if When the target point is within the communication range of the moved robot group, it is considered a selectable neighbor point. This represents the communication distance of the robot.

5. The method for regional coverage path planning based on dynamic networking under local communication according to claim 4, characterized in that, In step 4, the formula for calculating the predator avoidance reward is: in, The number of selectable neighbor points. This is the preset maximum reward value component. Point for neighbors To predators The Euclidean distance; similarly, For the current point The maximum Euclidean distance from the predator among all its neighboring points, and For the current point The minimum Euclidean distance between the predator and its neighbors; the formula for calculating the Euclidean distance is: 。 6. The method for regional coverage path planning based on dynamic networking under local communication according to claim 5, characterized in that, In step 5, the formula for calculating the linear motion reward is: in, The number of selectable neighbor points. This is the preset maximum reward value component. For the current point The previous historical position; Current location The One selectable neighbor point; For vectors and The oblique angle formed by the angle.

7. The method for regional coverage path planning based on dynamic networking under local communication according to claim 6, characterized in that, In step 6, the formula for calculating the boundary coverage reward is as follows: in, The number of selectable neighbor points. This is the preset maximum reward value component. express The maximum possible number of neighbors; and for The j-th neighbor point Number of uncovered neighbors.

8. The method for regional coverage path planning based on dynamic networking under local communication according to claim 7, characterized in that, Step 7 involves calculating the reward value components obtained from steps 4-6. , and Calculate neighbor points The total revenue is calculated using the following formula: in, and These are weighting factors for predator avoidance, linear movement, and boundary coverage rewards, respectively. The value of the weighting factor determines the degree of importance the forager attaches to different rewards when deciding on its next movement direction.

9. The method for regional coverage path planning based on dynamic networking under local communication according to claim 8, characterized in that, For scenarios two and three, a follow-and-escape mechanism is adopted, specifically implemented as follows: Let the area formed by the points be the coverage area. , as well as Three robots are tasked with performing path planning. The coordinates of the three robots that are stuck in a deadlock are as follows: , as well as ; First, calculate the "free" robots that are not deadlocked according to the formula in step 7. , Each's optimal neighbor target point , And set them as the current positions respectively; Next, the deadlocked robot was calculated according to the Euclidean distance calculation formula in step 4. and , European distance , ; like At this moment, the robot Follow the robot Movement, i.e., following the robot Using the previous position as a reference point, move towards the neighboring target point in the same direction; until the robot... Reach the appropriate backtracking point in the uncovered area.

Images