Multi-robot task allocation diagram segmentation method based on road network topology fusion edge points

Abstract

The invention belongs to the field of collaborative task allocation of a multi-robot system technology, and particularly relates to a multi-robot task allocation graph segmentation method based on road network topology fusion edge points, which comprises the following steps: constructing a minimum divisible block in a road network topological graph based on edge point information, and storing the minimum divisible block in a data list; generating upper and lower thresholds of the balance cost according to the number of the robots and the initial threshold factor; the minimum divisible block is transmitted according to a data list, the path cost corresponding to the transmitted block is calculated, the path cost is compared with the threshold value balance cost to judge whether task allocation at the stage is ended or the next minimum divisible block is transmitted, the next minimum divisible block is selected according to a set rule and then calculated and compared until current task allocation is ended, and the steps are repeated; and dividing the global road network topological graph into task groups with the corresponding number according to the number of the robots, reducing a set mean value cost threshold after division is completed, and obtaining a final segmentation result through iterative optimization. According to the method, blindness and redundancy are reduced while the efficiency of multi-robot road network traversal search in the complex urban environment is improved.

Description

Technical Field

[0001] The present invention belongs to the field of collaborative task allocation of multi-robot system technology, and specifically is a multi-robot task allocation graph segmentation method based on road network topology fusion edge points. Background Art

[0002] In complex urban scenarios, when multiple robots perform military reconnaissance, disaster relief, search and rescue and other tasks, the first stage of the centralized multi-robot system is to segment the task area map for the total task area, and to enable the tasks to be executed in parallel by assigning subgraphs to each robot. In essence, the road network topology input to the multi-robot system is modeled as a graph consisting of a finite non-empty set of vertices V ( G ) and a set of edges E ( G ) between vertices, and the input graph is segmented according to the number of multiple robots. In response to the task requirements of traversing the regional road network of multiple robots, the edges and points in the graph are assigned to the corresponding robots according to the rules to perform the traversal task.

[0003] According to the way of segmenting graph data, the current segmentation schemes can be divided into vertex partitioning (edge-cut partitioning) and edge partitioning (edge ​​partitioning or vertex-cut partitioning). Figure 1 As shown in , the nodes of the graph are distributed to each subgraph to maintain the integrity of the subgraphs between nodes; the edge division is as follows Figure 2 As shown in the figure, the edges of the graph are allocated to each subgraph, and each group of allocated edges constitutes a subgraph. However, with the increase in the complexity of the scene and the changes in the specific traversal requirements of multiple robots, the limitations of the traditional graph segmentation method begin to be exposed, as follows: 1. Traditional node-based graph segmentation methods often cause the edges between some nodes to be cut (edge-cut). That is, when two nodes are assigned to two subgraphs according to the rules, if the two nodes are connected in the road network topology, the edge connecting the two nodes will be cut into two segments in the actual road network topology after the assignment is completed, and assigned to different robots, causing redundancy in the actual robot traversal task, while also greatly increasing the complexity.

[0004] 2. For graph data that obeys power-law distribution, some nodes may be connected to a large number of edges. If point splitting is performed, a large number of edges will be missing and the load distribution of edges will be uneven, which is almost not feasible in actual multi-machine allocation.

[0005] 3. Traditional edge-based graph partitioning methods use edges as the basic partitioning unit, and their partitioning rules often do not consider edge connectivity. In actual multi-machine system allocation, the edges allocated to robots based on edge partitioning often become disconnected. However, the actual robot's traversal of the edges in the road network requires that the allocated road network topology is reachable, that is, connectivity must be guaranteed.

[0006] 4. The edge-based graph partitioning strategy does not consider nodes, which will cause node redundancy (vertex-cut). If the relationship between all nodes is calculated in advance to avoid redundancy, it will consume a lot of memory and converge slowly.

[0007] Therefore, when it comes to the graph segmentation task allocation problem facing the actual multi-robot road network topology, how to design a graph segmentation method that is practically feasible with low complexity and ensures the connectivity of the paths assigned to each single robot becomes an urgent problem to be solved. Summary of the invention

[0008] The purpose of the present invention is to provide a multi-robot task allocation graph segmentation method based on road network topology fusion edge points, so as to improve the efficiency of multi-robot road network traversal search in complex urban environments while reducing blindness and redundancy, and to be able to more efficiently and simply allocate path traversal tasks for multi-robot systems to meet the growing number of robots and task complexity requirements.

[0009] To achieve the above object, the present invention adopts the following technical solution: The multi-robot task allocation graph segmentation method based on road network topology fusion edge points includes the following steps: Step 1: Obtain the number of robots in the multi-robot system and the road network topology information of the task area, extract the global road network topology map based on the road network topology information of the task area, and calculate the total cost of the global road network topology map; Step 2: Process the global road network topology map to construct the smallest divisible block, number each smallest divisible block, and create a smallest divisible block data list; Step 3: Calculate the ideal equilibrium cost of each robot based on the number of robots and the total cost of the global road network topology, set the initial cost threshold factor, and calculate the upper and lower thresholds of the equilibrium cost based on the set initial cost threshold factor; Step 4: Select the smallest divisible block corresponding to the smallest number from the data list created in step 2 as input; Step 5: After inputting this smallest divisible block, calculate the sum of the costs of all paths in the smallest divisible block in the task group that surrounds the input; Step 6: Subtract the sum of the costs obtained in step 5 from the ideal equilibrium cost. If the absolute value of the difference is less than the lower threshold, the minimum divisible block of the next input is searched as input based on the rule of the closest block center distance, and the process returns to step 5 until the sum of the path costs corresponding to the minimum divisible block of the input is greater than the lower threshold and less than the upper threshold, and the task graph allocation is completed. If the absolute value of the difference is greater than the lower threshold but less than the upper threshold, the allocated minimum divisible block is removed from the data list, and the minimum divisible block with the smallest number in the data list at this moment is used as input, and the process returns to step 5 to generate the graph segmentation area for the next task until the input global road network topology graph is divided according to the number of robots; Step 7: Reduce the set mean cost threshold, repeat steps 3 to 6 to iteratively optimize the allocation results until the task balance index of each robot reaches the optimal level, and the optimal graph segmentation multi-robot task allocation result can be obtained.

[0010] Furthermore, the calculation of the block center coordinates and the search rule for the next block in step 6 are as follows: Calculate the mean of the horizontal and vertical coordinates of all points surrounding the minimum separable block to determine a new point to replace the center point of the current minimum separable block; use the Euclidean distance between the new center point and the center points of the remaining minimum separable blocks as the evaluation index, and select the minimum separable block with the smallest Euclidean distance, that is, the nearest block as the next input.

[0011] The multi-robot task allocation graph segmentation method based on road network topology fusion edge points provided by the present invention is aimed at the undirected graph segmentation task of urban road network topology modeling in the task area, and adopts the method of fusion edge point features, that is, segmentation is performed based on the smallest divisible block surrounded by edge points, and the road network topology cost balance assigned to each robot is taken as the optimization goal, and the connectivity of the road network after allocation is guaranteed according to the adjacency of the blocks, so as to generate the final graph segmentation result and realize the balanced division of the graph. One of the optimization goals in the multi-objective optimization is the balance of different allocation results each time, and the other optimization goal is the closeness of the single task group to the ideal balanced cost in each iteration.

[0012] Compared with the prior art, the present invention effectively reduces the blindness and redundancy of multi-robot searches in complex urban areas, while optimizing resource allocation, improving the balance and efficiency of multi-machine collaboration, and thus achieving reasonable task allocation. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 It is the traditional point segmentation method; Figure 2 It is the traditional edge segmentation method; Figure 3 It is a flow chart of a multi-robot task allocation graph segmentation method based on road network topology fusion edge points in an embodiment; Figure 4 Constructing the smallest separable blocks for the monotone edges for the embodiment; Figure 5 A schematic diagram of the extraction of the urban rule scenario task area network topology and the minimum divisible block division provided in the embodiment; Figure 6 A schematic diagram of the extraction of the wide-area irregular scenario task area network topology and the minimum divisible block division provided by the embodiment; Figure 7 A schematic diagram of multi-robot task allocation results based on graph segmentation of road network topology fusion edge points in the urban regular scene task area provided in the embodiment; Figure 8 A schematic diagram of multi-robot task allocation results based on graph segmentation of road network topology fused edge points in a wide-area irregular scene task area provided in an embodiment. DETAILED DESCRIPTION

[0014] In order to make the objectives, technical solutions and advantages of the present invention more clear, the present invention is described in detail below in conjunction with the implementation modes and the accompanying drawings.

[0015] like Figure 3 As shown, the multi-robot task allocation graph segmentation method based on road network topology fusion edge points provided in this embodiment includes the following steps: Step 1: Obtain the number of robots in the multi-robot system and the road network topology information of the task area, extract the global road network topology map based on the road network topology information of the task area, and calculate the total cost of the global road network topology map; the implementation process includes: A global road network topology graph is extracted from the task area road network topology information. The global road network topology graph consists of a finite non-empty set V ( G ) of vertices and a set E ( G ) of edges between vertices, expressed as: G = (V, E), wherein G represents the global road network topology graph, V is the set of vertices in the graph G, and E is the set of edges in the graph G.

[0016] In the global road network topology graph: define each vertex as P ( , (X, Y)), where is the number of the vertex, (X, Y) is the coordinate of the vertex; each edge is defined as E( ),in is the number of the edge, C is the cost corresponding to the edge, and in this embodiment, the road network length of the edge in the actual environment is used as its cost; For the input global road network topology, the total cost is calculated as follows: ; In the above formula, E n For all edges in the graph, Corresponding cost for each edge; Traverse all robots and obtain their number k; divide the global road network topology map into corresponding task groups.

[0017] Step 2: Process the global road network topology map to construct the smallest divisible blocks and number them, and number each smallest divisible block, and create a smallest divisible block data list. The implementation process includes: Define the smallest divisible block. The block is surrounded by points and edges. The smallest divisible block is defined as a block that has no further divisible blocks, that is, the degree of each vertex in the block is 2. The specific definition of each smallest divisible block is B = (N, P n , E n ), where N is the number of the smallest divisible block, P n is the set of points surrounding the block, E n is the edge set surrounding the block. If there is a point with degree 1 in the global road network topology, the edge connecting the point is a monotonic edge and cannot form a block. At this time, the smallest divisible block is constructed for the edge, that is, a repeated edge of the edge is added to construct a set of rectangular blocks with opposite edges of 0, so that the global road network topology can be completely divided into a set of smallest divisible blocks, and they are numbered and constructed and stored in the smallest divisible block data list. The smallest divisible block data list is . Figure 4 To show the process of constructing the smallest divisible block for monotone edges, Figure 4 As shown in the figure, the road network topology graph consists of five vertices: a, b, c, d, and e. Vertex a is a point with a degree of 1, that is, the edge (a, b) connecting this point is a monotonic edge and cannot form a block. On the basis of the original topology graph, a repeated edge (a', b') of this edge is added to construct a set of rectangular blocks with opposite sides being 0, that is, the cost of the edge (a, a') and the edge (b, b') is zero, forming a minimum divisible block.

[0018] Step 3: Calculate the ideal equilibrium cost of each robot based on the number of robots and the total cost of the global road network topology, and set the initial cost threshold factor as , calculate the upper and lower thresholds of the equilibrium cost according to the set initial cost threshold factor; the specific calculation formula is as follows: ; ; ; Among them, Average_Cost is the ideal equilibrium cost, k is the number of robots, Average_Up is the upper threshold of the equilibrium cost, and Average_Down is the lower threshold of the equilibrium cost.

[0019] Step 4: Based on the global road network topology map obtained in step 2, select the smallest divisible block corresponding to the smallest number from the data list created in step 2 as input; Step 5: Calculate the sum of the costs of all paths in the smallest divisible block that surrounds the input in the task group after inputting the smallest divisible block; the details are as follows: This section is defined as: The block set is ; Each block Edged Set ; Each edge There is a price ; The total cost formula for inputting the smallest divisible block is: ; Where n represents the total number of blocks in the current block set, m i Represents the number of edges of the i-th block.

[0020] Step 6: Sum the cost obtained in step 5 Subtracting from the ideal equilibrium cost, the difference between the current input block cost and the equilibrium cost is: ; If the absolute value of the difference is less than the lower threshold, the minimum divisible block of the next input is searched as input based on the rule of the closest block center distance, and return to step 5 until the sum of the path costs corresponding to the minimum divisible block of the input is greater than the lower threshold and less than the upper threshold, then the task graph allocation is completed. The calculation of the block center coordinates and the search rules for the next block are as follows: The input block set is ; For each block Edged Set ; Each point With coordinates ; The formula for the center coordinates of the input block is: ; ; List of smallest divisible blocks , t represents the smallest divisible block in the data list, and m represents the mth smallest divisible block in the data list; Each block The center coordinates are obtained from the above formula , which is simplified to: ; The distance between blocks is calculated as: ; Find the next smallest input block, that is, update the condition: ; ; ; If the absolute value of the difference is greater than the lower threshold and less than the upper threshold, the allocated minimum divisible block is removed from the data list, and the minimum divisible block with the smallest number in the data list at this moment is used as input, and the process returns to step 5 to generate the graph segmentation area for the next task until the input global road network topology map is divided according to the number of robots.

[0021] Step 7: Reduce the set mean cost threshold, iteratively optimize and repeat steps 3 to 6 until each robot task balance index reaches the optimal level, and the best multi-robot task allocation result can be obtained. Evaluate the balance of each allocation result. There are two conditions for ending the iteration. The iteration can be ended if any of the conditions is met, where: The method of reducing the set mean cost threshold is: according to each iteration, the mean cost threshold is set to 90% of the previous one, so as to reduce the mean cost threshold. , and then the cost threshold factor of the second iteration is 0.9 , and so on, until its variance no longer decreases or the path cost of each task allocation result reaches the set ideal threshold range; One of the two conditions for ending the iteration is that the cost of each allocation is within the set ideal threshold range: In the i-th iteration, if the cost of each allocation is The ideal equilibrium price The difference is within the tolerance If it is within, then this condition is met, that is: ; ; in, is the set tolerance factor; Another condition is that the variance no longer decreases: In the i-th iteration, if the variance Greater than or equal to the variance of the previous iteration , then this condition is satisfied, namely: ; ; in, represents the variance, is the total cost of each observation, that is, the corresponding block after each allocation , is the mean, here given by Instead, n is the number of observations, here it is the number of robots k, that is, the number of tasks finally assigned.

[0022] In order to verify the feasibility and advantages of the multi-robot task allocation graph segmentation method of this embodiment, it is placed in a specific scenario for illustration below: Figure 5 Demonstrated the extraction of the network topology and the minimum block division of the task area in the city rule scenario; Figure 6 This paper demonstrates the extraction of the task area network topology and the minimum block division in a wide-area irregular scenario. Figure 5 It can be seen that the extracted road network topology has a total of 15 vertices, 22 edges, and 8 smallest divisible blocks; Figure 6 It can be seen that the extracted road network topology has a total of 17 vertices, 26 edges, and 10 smallest divisible blocks. It can be seen that this embodiment is segmented based on the smallest divisible blocks surrounded by edge points, with the balance of the road network topology cost allocated to each robot as the optimization goal, and at the same time, the connectivity of the allocated road network can be guaranteed according to the adjacency of the blocks. Figure 7 A schematic diagram showing the results of multi-robot task allocation in the urban regular scene task area based on the graph segmentation of road network topology fusion edge points; Figure 8 The schematic diagram of the multi-robot task allocation results based on the graph segmentation of the road network topology fusion edge points in the wide-area irregular scene task area is shown. Figure 7 and Figure 8 It can be seen that this embodiment generates the final graph segmentation result and realizes the balanced division of the graph. The road network topology of the area surrounded by each color represents that in the two scenes, the graph segmentation results assigned to the four robots are robot 1, robot 2, robot 3 and robot 4 respectively, indicating that the method of this embodiment can more efficiently and conveniently allocate the path traversal tasks of the multi-robot system, meeting the increasing number of robots and task complexity requirements.

[0023] In summary, the multi-robot task allocation graph segmentation method based on road network topology fusion edge points in this embodiment adopts a method of fusing edge point features for the undirected graph segmentation task of urban road network topology modeling in the task area, that is, segmentation is performed based on the smallest divisible blocks surrounded by edge points, and the balance of the road network topology cost assigned to each robot is used as the optimization goal. At the same time, the connectivity of the road network after allocation is guaranteed according to the adjacency of the blocks, and the final graph segmentation result is generated in conjunction with multi-objective optimization to achieve balanced division of the graph.

[0024] It is to be understood that the present invention is described by some embodiments, and it is known to those skilled in the art that various changes or equivalent substitutions may be made to these features and embodiments without departing from the spirit and scope of the present invention. In addition, under the teachings of the present invention, these features and embodiments may be modified to adapt to specific circumstances and materials without departing from the spirit and scope of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed herein, and all embodiments falling within the scope of the claims of this application are within the scope of protection of the present invention.

Claims

1. A multi-robot task allocation graph segmentation method based on road network topology fusion edge points, characterized in that: The following steps are involved: Step 1: Obtain the number of robots in the multi-robot system and the road network topology information of the task area, extract the global road network topology map based on the road network topology information of the task area, and calculate the total cost of the global road network topology map; Step 2: Process the global road network topology map to construct the smallest divisible block, number each smallest divisible block, and create a smallest divisible block data list; Step 3: Calculate the ideal equilibrium cost of each robot based on the number of robots and the total cost of the global road network topology, set the initial cost threshold factor, and calculate the upper and lower thresholds of the equilibrium cost based on the set initial cost threshold factor; Step 4: Select the smallest divisible block corresponding to the smallest number from the data list created in step 2 as input; Step 5: After inputting this smallest divisible block, calculate the sum of the costs of all paths in the smallest divisible block in the task group that surrounds the input; Step 6: Subtract the sum of the costs obtained in step 5 from the ideal equilibrium cost. If the absolute value of the difference is less than the lower threshold, the minimum divisible block of the next input is searched as input based on the rule of the closest block center distance, and the process returns to step 5 until the sum of the path costs corresponding to the minimum divisible block of the input is greater than the lower threshold and less than the upper threshold, and the task graph allocation is completed. If the absolute value of the difference is greater than the lower threshold but less than the upper threshold, the allocated minimum divisible block is removed from the data list, and the minimum divisible block with the smallest number in the data list at this moment is used as input, and the process returns to step 5 to generate the graph segmentation area for the next task until the input global road network topology graph is divided according to the number of robots; Step 7: Reduce the set mean cost threshold, repeat steps 3 to 6 to iteratively optimize the allocation results until the task balance index of each robot reaches the optimal level, and the optimal graph segmentation multi-robot task allocation result can be obtained.

2. The multi-robot task allocation graph segmentation method based on road network topology fusion edge points according to claim 1 is characterized in that: The implementation process of step 1 includes: Extracting a global road network topology graph from the task area road network topology information, wherein the global road network topology graph is composed of a finite non-empty set V(G) of vertices and a set E(G) of edges between vertices, expressed as: G=(V, E), wherein G represents the global road network topology graph, V is the set of vertices in the graph G, and E is the set of edges in the graph G; In the global road network topology, each vertex is defined as P(Np, (X, Y)), where Np is the number of the vertex and (X, Y) is the coordinate of the vertex; each edge is defined as E(N e , C), where N e is the number of the edge, and C is the cost corresponding to the edge; For the input global road network topology, the total cost is calculated as follows: ； In the above formula, Sum_Cost is the total cost of the input global road network topology, E n For all edges in the graph, Corresponding cost for each edge; Traverse all robots and obtain their number k; divide the global road network topology map into corresponding task groups.

3. The multi-robot task allocation graph segmentation method based on road network topology fusion edge points according to claim 1 is characterized in that: The implementation process of step 2 includes: Define the smallest divisible block. The block is surrounded by points and edges. The smallest divisible block is defined as a block that has no further divisible blocks, that is, the degree of each vertex in the block is 2; For each smallest divisible block, it is specifically defined as B=(N, P n , E n ), where N is the number of the smallest divisible block, P n is the set of points surrounding the block, E n is the set of edges surrounding the block; If there is a point with degree 1 in the global road network topology, that is, the edge connecting the point is a monotone edge and cannot form a block, then add a duplicate edge of the edge to construct a set of rectangular blocks with opposite edges of 0. In this way, the global road network topology is completely divided into a set of the smallest divisible blocks. Number the smallest divisible blocks, construct them and store them in the smallest divisible block data list. The smallest divisible block data list is Where t represents the smallest divisible block in the data list, and m represents the mth smallest divisible block in the data list.

4. The multi-robot task allocation graph segmentation method based on road network topology fusion edge points according to claim 3 is characterized in that: The calculation formula used in step 3 is as follows: ; ； ； Among them, Average_Cost is the ideal equilibrium cost, k is the number of robots, Average_Up is the upper threshold of the equilibrium cost, and Average_Down is the lower threshold of the equilibrium cost.

5. The multi-robot task allocation graph segmentation method based on road network topology fusion edge points according to claim 4 is characterized in that: The implementation method of step 5 includes: Define the block set as ; Each block Edged Set ; Each edge There is a price ; The sum of the costs on all paths in the smallest divisible block that encloses the input is: ； Where n represents the total number of blocks in the current block set, m i Represents the number of edges of the i-th smallest divisible block.

6. The multi-robot task allocation graph segmentation method based on road network topology fusion edge points according to claim 5 is characterized in that: The calculation of the block center coordinates and the search rules for the next block in step 6 are as follows: Calculate the mean of the horizontal and vertical coordinates of all points surrounding the minimum separable block to determine a new point to replace the center point of the current minimum separable block; use the Euclidean distance between the new center point and the center points of the remaining minimum separable blocks as the evaluation index, and select the minimum separable block with the smallest Euclidean distance, that is, the nearest block as the next input.

7. The multi-robot task allocation graph segmentation method based on road network topology fusion edge points according to claim 1 is characterized in that: The condition for ending the iteration in step 7 should satisfy any one of the two conditions at the same time, one of the two conditions is that the cost of each allocation reaches the set ideal threshold range, and the other condition is that the variance no longer decreases.

Images