Adaptive formation obstacle avoidance method and system for intelligent drainage robot

Abstract

This application relates to the field of robot formation control, specifically to an adaptive formation obstacle avoidance method and system for intelligent flood drainage robots. The method includes: collecting point cloud data of a flooded area and constructing a grid map; determining the initial formation configuration of multiple intelligent flood drainage robots in the grid map; defining obstacle avoidance grids for the robots based on whether the neighboring grids of the robot's current position satisfy collision constraints when facing dynamic obstacles; constructing the obstacle avoidance confidence score for each obstacle avoidance grid; selecting the obstacle avoidance grid with the highest confidence score as the robot's obstacle avoidance grid, and adjusting the robot's movement direction accordingly to achieve formation obstacle avoidance. This application aims to improve the cooperative obstacle avoidance performance of multiple robots in dynamic environments and the robustness of the overall formation motion.

Description

Technical Field

[0001] This application relates to the field of robot formation control, specifically to an adaptive formation obstacle avoidance method and system for intelligent flood drainage robots. Background Technology

[0002] The intelligent drainage robot is designed with a powerful drainage system that can quickly pump out accumulated water and discharge it to designated locations. Its efficient pumping system works closely with specially optimized drainage pipelines to significantly improve drainage efficiency, enabling it to handle large amounts of accumulated water in a short time and alleviate urban flooding. Among them, the Junrong Emergency Multifunctional Drainage Robot integrates high mobility, high drainage efficiency, diverse functions, and operational safety, making it suitable for emergency drainage and rescue operations in urban and rural underground parking garages, underpasses, subway stations, and other spaces after heavy rains and typhoons.

[0003] During flood drainage, intelligent drainage robots need to pre-set a set of movement paths in the flooded area and carry out drainage tasks along the optimal path. However, the water volume in flooded areas is large, especially in intelligent drainage tasks in complex environments, such as urban water conservancy drainage and post-disaster relief scenarios. Existing methods usually combine triangulation with path planning to handle multi-robot motion planning or use Lyapunov functions to achieve multi-robot formation obstacle avoidance. However, during the movement of multiple robots, when avoiding dynamic obstacles (moving obstacles, such as rolling stones), the obstacle avoidance ability is hindered, the obstacle avoidance effect is poor, and it cannot meet the requirements for formation motion planning in complex environments. Summary of the Invention

[0004] To address the aforementioned technical problems, this application provides an adaptive formation obstacle avoidance method and system for intelligent flood drainage robots, the specific technical solution of which is as follows:

[0005] In a first aspect, one embodiment of this application provides an adaptive formation obstacle avoidance method for intelligent flood drainage robots, the method comprising the following steps:

[0006] Collect point cloud data of the flooded area and construct a raster map;

[0007] Determine the initial formation configuration of multiple intelligent drainage robots in a grid map. The initial formation configuration is determined by the target formation, formation structure, communication relationship and formation constraints.

[0008] Based on whether the robot's current position neighborhood grid satisfies collision constraints when facing dynamic obstacles, the robot's obstacle avoidance grid is defined; the first obstacle avoidance factor of each obstacle avoidance grid is determined based on the distance between the actual position of the dynamic obstacle and the robot's current position; the edge weights in the initial undirected graph with the robot as the vertex are constructed using the distance between robots and the obstacle avoidance weights; the second obstacle avoidance factor of each obstacle avoidance grid is constructed by the graph editing distance between the undirected graph constructed after the robot moves to all obstacle avoidance grids and the initial undirected graph.

[0009] The obstacle avoidance confidence of each obstacle avoidance grid of the robot is determined by combining the first and second obstacle avoidance factors; the obstacle avoidance grid with the highest obstacle avoidance confidence is selected as the grid for robot obstacle avoidance, and the robot's movement direction is adjusted accordingly to achieve formation obstacle avoidance.

[0010] Preferably, the initial formation configuration is the set of initial position coordinates of all intelligent drainage robots in the grid map, with each intelligent drainage robot occupying one grid.

[0011] Preferably, the formation constraints include collision constraints and environmental constraints, wherein the collision constraint is: the safe distance between intelligent drainage robots is greater than twice the diameter of the drainage range of the intelligent drainage robots.

[0012] Preferably, in the formation, the robot closest to the drainage location is designated as the leader robot, and the rest are designated as follower robots.

[0013] Preferably, the obstacle avoidance grid includes a first obstacle avoidance grid, a second obstacle avoidance grid, and a third obstacle avoidance grid, corresponding to obstacle avoidance without obstacles, obstacle avoidance that meets collision constraints, and obstacle avoidance in feasible areas under multi-robot cooperative obstacle avoidance conditions, respectively.

[0014] Preferably, the third obstacle avoidance grid is obtained by deleting common areas in the second obstacle avoidance grids of multiple robots, and is prioritized in combination with robot obstacle avoidance weights to achieve multi-robot cooperative obstacle avoidance.

[0015] Preferably, the obstacle avoidance weight is calculated based on the robot's working time and current battery level, using the following formula: w represents the robot's obstacle avoidance weight, t represents the robot's working time, and h represents the robot's current battery level.

[0016] Preferably, the method for calculating the second obstacle avoidance factor is as follows:

[0017] Calculate the reciprocal of the mean graph edit distance between the undirected graph constructed after the robot moves to all obstacle avoidance grids in sequence and the initial undirected graph;

[0018] For any obstacle avoidance grid of the robot, calculate the reciprocal of the mean of the graph edit distance between the undirected graph constructed after the robot moves to all obstacle avoidance grids except for the one mentioned above and the initial undirected graph;

[0019] Take the absolute value of the difference between the two reciprocals as the second obstacle avoidance factor for any obstacle avoidance grid of the robot.

[0020] Preferably, the formula for calculating the obstacle avoidance confidence level is: ; This represents the obstacle avoidance confidence of the j-th obstacle avoidance grid for the i-th robot. Let e ​​represent the first obstacle avoidance factor of the j-th obstacle avoidance grid for the i-th robot, where e is the natural constant. This represents the graph edit distance between the undirected graph constructed after the i-th robot moves the j-th obstacle avoidance grid and the initial undirected graph. This represents the second obstacle avoidance factor of the j-th obstacle avoidance grid for the i-th robot. , This indicates the preset weighting coefficient.

[0021] Secondly, another embodiment of this application also provides an adaptive formation obstacle avoidance system for intelligent drainage robots, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the aforementioned adaptive formation obstacle avoidance method for intelligent drainage robots.

[0022] This application has at least the following beneficial effects:

[0023] Considering that a single intelligent drainage robot may not be able to perform drainage work well in complex environments, this application utilizes multiple intelligent drainage robots to perform drainage tasks simultaneously and form a formation to avoid collisions between robots when there are no dynamic obstacles. In the grid map corresponding to the flooded area, based on the collision probability of each intelligent drainage robot when facing dynamic obstacles in its neighboring grid, the overlap probability with the neighboring grid of adjacent intelligent drainage robots, and the current performance of each intelligent drainage robot, the obstacle avoidance grid of each intelligent drainage robot is obtained. By using the performance data of the intelligent drainage robots and the movement characteristics of dynamic obstacles, the obstacle avoidance confidence of each obstacle avoidance grid is calculated. The movement direction of the intelligent drainage robot is adjusted by the obstacle avoidance grid corresponding to the maximum obstacle avoidance confidence, so that each robot can achieve adaptive obstacle avoidance in different obstacle encounter situations (single or multiple robots encountering obstacles) (encountering dynamic obstacles), effectively improving the cooperative obstacle avoidance performance of multiple robots in dynamic environments and the robustness of the overall formation movement. Attached Figure Description

[0024] To more clearly illustrate the technical solutions and advantages in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0025] Figure 1 This is a flowchart of an adaptive formation obstacle avoidance method for intelligent drainage robots provided in one embodiment of this application. Detailed Implementation

[0026] Example 1

[0027] This application provides an embodiment of an adaptive formation obstacle avoidance method for intelligent flood drainage robots, for details of which can be found in the following document. Figure 1 The method includes the following steps:

[0028] Step 1: Collect point cloud data of the flooded area and construct a raster map.

[0029] A drone equipped with a lidar system was used to repeatedly patrol the area over the flooded region to acquire point clouds. During the acquisition process, the lidar sensor may experience vibration interference. Therefore, a point cloud filtering algorithm (e.g., voxel filtering, bilateral filtering, etc.; this embodiment uses bilateral filtering) was used to filter the acquired point clouds and eliminate external interference. The pre-processed point cloud was used as input, and a grid size was set (2m in this embodiment). The pre-processed point cloud was then mapped onto a grid map using a grid method. In the grid map, grid cells with obstacles were marked as 1, and unobstructed grid cells were marked as 0. Both the point cloud filtering algorithm and the grid method are well-known technologies, and the specific calculation methods will not be elaborated further.

[0030] Each intelligent drainage robot is equipped with network communication devices, lidar, and other equipment, enabling communication between multiple intelligent drainage robots via wireless network; and through lidar, with a collection interval of 2 seconds, it collects information about the surrounding environment in real time, including the location and size of possible obstacles, as well as its own distance from other robots.

[0031] Step 2: Determine the initial formation configuration of multiple intelligent drainage robots in the grid map. The initial formation configuration is determined by the target formation, formation structure, communication relationship and formation constraints.

[0032] When the flooded area is a tunnel, the number of intelligent drainage robots to be used is set according to the actual physical width of the tunnel and the scale of drainage (3 robots in this embodiment); based on the number of intelligent drainage robots, a feasible target formation is predefined. In this embodiment, the 3 intelligent drainage robots are arranged in an equilateral triangle, and the initial distance between the three robots is twice the length of the robot. The intelligent drainage robot facing the apex of the triangle towards the drainage position is the leader robot, and the rest are the follower robots.

[0033] Each intelligent drainage robot is treated as a vertex, the two vertices that need to maintain communication are treated as an edge, and the Euclidean distance between the two vertices on each edge is used as the edge weight. The formation of the three intelligent drainage robots is described by an undirected graph, in which the intelligent drainage robots all have communication relationships with each other.

[0034] Define formation constraints and, for the collision problem between intelligent drainage robots, obtain the position coordinates of each intelligent drainage robot in the grid map. , Based on the dimensions of the intelligent drainage robots, determine the safe distance between them. (In this embodiment, this minimum safe distance is...) (set to be twice the diameter of the flood discharge range of the intelligent flood drainage robot), therefore the collision constraint can be defined as: , The distance between the two intelligent drainage robots in the corresponding grid area is represented by L, which represents the diameter of the drainage range of the intelligent drainage robot. In this embodiment, it is set to 4m. The environmental constraints are defined as follows: X represents a raster marker.

[0035] The formation constraints are set according to the actual situation: In this embodiment, the initial grid size is set to 2m; the normal flood discharge range is greater than 2m; therefore, there will be multiple grids between each robot, so that subsequent calculations can be performed based on the grid. Moreover, the flood discharge range setting can maximize the robot's flood discharge efficiency.

[0036] Using formation constraints, grid map, and undirected graph as input, and the target formation as the objective function, the SCIP algorithm is used to obtain a feasible initial formation configuration for three intelligent drainage robots. The initial formation configuration is the set of initial position coordinates of all intelligent drainage robots in the grid map, that is, each intelligent drainage robot occupies one grid.

[0037] Step 3: Define the robot's obstacle avoidance grid based on whether the neighborhood grid of the robot's current position satisfies the collision constraint when facing dynamic obstacles; determine the first obstacle avoidance factor for each obstacle avoidance grid of the robot based on the distance between the actual position of the dynamic obstacle and the robot's current position; construct the edge weights in the initial undirected graph with the robot as the vertex using the distance between robots and the obstacle avoidance weights; construct the second obstacle avoidance factor for each obstacle avoidance grid of the robot by using the graph editing distance between the undirected graph constructed after the robot moves to all obstacle avoidance grids and the initial undirected graph.

[0038] In a raster map, the first step is to use a path planning algorithm (e.g., A...). Algorithms, etc., are used to plan the path for the leader robot in the initial formation configuration, resulting in a set of global path points. During the actual flood drainage process, the leader robot moves according to these global path points and performs flood drainage work. The follower robots follow the leader robot according to the initial formation configuration.

[0039] Taking the i-th robot as an example, when a dynamic obstacle (e.g., a rockfall) appears on the i-th robot's path, the intelligent drainage robot can obtain the position of the dynamic obstacle by collecting point clouds of consecutive frames. Using the point cloud of consecutive frames as input, a point cloud-based motion tracking algorithm (e.g., the TrajTrack algorithm) is used. In this embodiment, the TrajTrack algorithm is used to track the motion of the point cloud of consecutive frames to obtain the current position, speed, and direction of the dynamic obstacle. An extended Kalman filter is used to predict the trajectory of the dynamic obstacle to obtain its predicted position. The actual and predicted positions of the dynamic obstacle are mapped to the same grid map using a world coordinate system, and the grid map showing the actual and predicted positions of the dynamic obstacle is marked as 2. The point cloud-based motion tracking algorithm and the extended Kalman filter are well-known technologies and will not be described in detail here.

[0040] The description of dynamic obstacles is as follows: Due to the large amount of water accumulation at the drainage location, which poses a significant risk of hazard, any moving object should be considered as a dynamic obstacle. Therefore, the dynamic obstacles refer to all objects in motion detected by the robot.

[0041] For the i-th robot, if a dynamic obstacle is detected, the current position, speed, and direction of movement of the i-th robot and the dynamic obstacle in the grid map are used as input. The obstacle avoidance distance (1m in this embodiment, i.e., one grid) and time window parameters (set as the shortest time required for the robot to pass through one grid) are set. The collision detection algorithm (e.g., the closest point algorithm, the trajectory discrete sampling algorithm) is used to determine whether the movement trajectory of the two will be less than the obstacle avoidance distance in the future, thereby determining whether a collision has occurred.

[0042] If a collision is predicted, the neighborhood of the grid containing the i-th robot is obtained within a grid map of a preset size, centered on the grid containing the i-th robot. In this embodiment, the 24-neighborhood of the grid containing the i-th robot is obtained within a 5×5 grid map.

[0043] Similarly, using a collision detection algorithm, each unobstructed grid marked as 0 in the 24 neighborhood of the i-th robot is taken as the current position. It is then determined whether the grid where the robot's current position is located will collide with the grid where the actual position of the dynamic obstacle is located. The grid marked as 2 that will not collide is recorded as the first obstacle avoidance grid of the i-th robot. The first obstacle avoidance grid is the unobstructed obstacle avoidance grid.

[0044] For the first obstacle avoidance grid, when the i-th robot moves to the first obstacle avoidance grid, it is determined whether the positions of the current 3 intelligent drainage robots meet the collision constraints. The first obstacle avoidance grid that meets the conditions is recorded as the second obstacle avoidance grid of the corresponding robot. The second obstacle avoidance grid is the obstacle avoidance grid that meets the collision constraints.

[0045] When the i-th robot predicts that a collision will occur, it interacts with other robots through communication devices to obtain the obstacle encounter status of all robots.

[0046] If only one robot encounters an obstacle, then that robot's obstacle avoidance grid becomes its second obstacle avoidance grid.

[0047] If multiple robots encounter obstacles, determine whether there is a common area in the second obstacle avoidance grid of all robots. If there is a common area, taking the i-th robot as an example, delete the common area in the second obstacle avoidance grid of the i-th robot and use it as the third obstacle avoidance grid of the i-th robot. The same applies to the other robots.

[0048] Based on the drainage data of each robot (e.g., working time, battery level), an obstacle avoidance weight is assigned to each robot. This obstacle avoidance weight is calculated from the drainage data. t represents the robot's working time, and h represents the robot's current battery level. The longer the working time and the lower the current battery level, the weaker the robot's current battery life and the higher the obstacle avoidance weight.

[0049] If two robots encounter obstacles simultaneously, the obstacle avoidance grid of the robot with the higher obstacle avoidance weight should be the third obstacle avoidance grid, while the obstacle avoidance grid of the robot with the lower obstacle avoidance weight should remain the second obstacle avoidance grid. The third obstacle avoidance grid represents the feasible area obstacle avoidance under the condition of multi-robot cooperative obstacle avoidance.

[0050] If all three robots encounter obstacles simultaneously, the obstacle avoidance grid for the two robots with higher obstacle avoidance weights should be the third obstacle avoidance grid, while the obstacle avoidance grid for the robot with lower obstacle avoidance weights should remain the second obstacle avoidance grid.

[0051] Step 4: Combine the first and second obstacle avoidance factors to determine the obstacle avoidance confidence of each obstacle avoidance grid of the robot; select the obstacle avoidance grid with the highest obstacle avoidance confidence as the obstacle avoidance grid of the robot, and adjust the robot's movement direction accordingly to achieve formation obstacle avoidance.

[0052] For the i-th robot, the distance between the dynamic obstacle and the i-th robot can be used as one of the indicators to measure the danger level of the dynamic obstacle. Therefore, distance algorithms (such as Euclidean distance, Manhattan distance, etc.) are used to calculate the distance between the actual position of the dynamic obstacle and the current position of the robot, that is, the distance between the grid corresponding to the actual position of the dynamic obstacle and the grid corresponding to the current position of the robot, which is used as the first obstacle avoidance factor for each obstacle avoidance grid of the robot.

[0053] If an undirected graph is constructed for three robots, when the i-th robot encounters an obstacle, its position needs to be moved to its corresponding obstacle avoidance grid. At this time, the undirected graph will change. If the change in the undirected graph is large, it may disrupt the connectivity of the three robots and weaken their collaborative ability. Therefore, the rationality of the obstacle avoidance grid is reflected by the degree of change in the undirected graph after the i-th robot moves.

[0054] In this embodiment, the position of each robot is taken as a vertex, the line connecting two adjacent robots is taken as an edge, and the weighted sum of the Euclidean distance between two vertices and the absolute value of the difference in obstacle avoidance weights (the weight coefficient in this embodiment is 0.5) is used as the edge weight to construct an initial undirected graph of three robots, denoted as . ;

[0055] Taking the i-th robot moving to the j-th obstacle avoidance grid as an example, after the i-th robot moves to the j-th obstacle avoidance grid, the robot's power consumption will increase, that is: , This represents the power consumption of the i-th robot moving from its current position to the j-th obstacle avoidance grid. This represents the time taken for the i-th robot to move from its current position to the j-th obstacle avoidance grid. Let represent the power consumption of the i-th robot during its movement; the obstacle avoidance weights for the i-th robot at this time should be: , This indicates the battery level when the robot is not moving. This represents the total power of the i-th robot.

[0056] After updating the obstacle avoidance weights, calculate... Similarly, calculate the undirected graph of the three robots. Using graph editing distance algorithms (e.g., GED algorithm), calculate and Similarly, the graph edit distance between them can be calculated. The graph edit distances after the robot moves to all obstacle avoidance grids are combined to form a graph edit distance sequence. The reciprocal of the mean of all elements in this sequence is then calculated. After deletion, the reciprocal of the mean of all elements in the sequence is calculated in the same way, and the absolute value of the difference between the two reciprocals is taken as the second obstacle avoidance factor of the j-th obstacle avoidance grid of the i-th robot.

[0057] By combining the first and second obstacle avoidance factors of the j-th obstacle avoidance grid of the i-th robot, the obstacle avoidance confidence of the j-th obstacle avoidance grid of the i-th robot is calculated:

[0058]

[0059] This represents the obstacle avoidance confidence of the j-th obstacle avoidance grid for the i-th robot. Let e ​​represent the first obstacle avoidance factor of the j-th obstacle avoidance grid for the i-th robot, where e is the natural constant. This represents the graph edit distance between the undirected graph constructed after the i-th robot moves the j-th obstacle avoidance grid and the initial undirected graph. This represents the second obstacle avoidance factor of the j-th obstacle avoidance grid for the i-th robot. , This represents the preset weighting coefficient. In this embodiment, .

[0060] Calculation principle: Based on the fundamental nonlinear exponential formula, and and Inversely proportional, and It is directly proportional, therefore a nonlinear exponential formula is used to represent it. .

[0061] The grid cell with the highest obstacle avoidance confidence is selected as the obstacle avoidance grid cell for the i-th robot when it encounters an obstacle.

[0062] When multiple robots encounter obstacles, the position of the robot with the lowest obstacle avoidance weight (before the update) remains unchanged. The obstacle avoidance confidence of the obstacle avoidance grid corresponding to the robot with the highest obstacle avoidance weight (before the update) is calculated, and the obstacle avoidance grid with the highest obstacle avoidance confidence is selected as the obstacle avoidance grid for the robot with the highest obstacle avoidance weight (before the update). After fixing the position, the initial undirected graph is reconstructed, and the obstacle avoidance grid with the highest obstacle avoidance confidence is selected as the obstacle avoidance grid for the robot with the lowest obstacle avoidance weight (before the update).

[0063] After obstacle avoidance is completed, the system immediately triggers the formation recovery process: each robot uploads its current position information via its communication module, taking the target formation, formation constraints, grid map, and undirected graph as input, and uses the SCIP algorithm to perform formation reconstruction. During formation reconstruction, the system continuously monitors environmental dynamics and communication quality. If a new obstacle is detected or communication is interrupted, the current process is immediately interrupted, obstacle avoidance is restarted, until all robots finally and stably reach the target formation position, completing the re-formation task.

[0064] Example 2

[0065] Another embodiment of this application provides an adaptive formation obstacle avoidance system for intelligent drainage robots, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the adaptive formation obstacle avoidance method for intelligent drainage robots described above.

[0066] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the invention herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not invented in this application.

[0067] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope.

Claims

1. An adaptive formation obstacle avoidance method for intelligent flood drainage robots, characterized in that, The method includes the following steps: Collect point cloud data of the flooded area and construct a raster map; Determine the initial formation configuration of multiple intelligent drainage robots in a grid map. The initial formation configuration is determined by the target formation, formation structure, communication relationship and formation constraints. Based on whether the robot's current position neighborhood grid satisfies collision constraints when facing dynamic obstacles, an obstacle avoidance grid for the robot is defined. A first obstacle avoidance factor for each obstacle avoidance grid is determined based on the distance between the actual position of the dynamic obstacle and the robot's current position. Edge weights in an initial undirected graph with the robot as a vertex are constructed using the distances between robots and the obstacle avoidance weights. A second obstacle avoidance factor for each obstacle avoidance grid is constructed using the graph editing distance between the undirected graph constructed after the robot has moved to all obstacle avoidance grids and the initial undirected graph. The second obstacle avoidance factor is calculated as follows: the reciprocal of the mean of the graph editing distances between the undirected graph constructed after the robot has moved to all obstacle avoidance grids and the initial undirected graph is calculated. For any obstacle avoidance grid of the robot, the reciprocal of the mean of the graph editing distances between the undirected graph constructed after the robot has moved to all obstacle avoidance grids except for the one mentioned above is calculated. The absolute value of the difference between the two reciprocals is taken as the second obstacle avoidance factor for any obstacle avoidance grid of the robot. The obstacle avoidance confidence of each obstacle avoidance grid of the robot is determined by combining the first and second obstacle avoidance factors; the obstacle avoidance grid with the highest obstacle avoidance confidence is selected as the grid for robot obstacle avoidance, and the robot's movement direction is adjusted accordingly to achieve formation obstacle avoidance. The formula for calculating the obstacle avoidance confidence level is as follows: ; This represents the obstacle avoidance confidence of the j-th obstacle avoidance grid for the i-th robot. Let e ​​represent the first obstacle avoidance factor of the j-th obstacle avoidance grid for the i-th robot, where e is the natural constant. This represents the graph edit distance between the undirected graph constructed after the i-th robot moves the j-th obstacle avoidance grid and the initial undirected graph. This represents the second obstacle avoidance factor of the j-th obstacle avoidance grid for the i-th robot. , This indicates the preset weighting coefficients.

2. The adaptive formation obstacle avoidance method for intelligent flood drainage robots as described in claim 1, characterized in that, The initial formation configuration is the set of initial position coordinates of all intelligent drainage robots in the grid map, with each intelligent drainage robot occupying one grid.

3. The adaptive formation obstacle avoidance method for intelligent flood drainage robots as described in claim 1, characterized in that, The formation constraints include collision constraints and environmental constraints. The collision constraint is that the safe distance between intelligent drainage robots is greater than twice the diameter of the drainage range of the intelligent drainage robots.

4. The adaptive formation obstacle avoidance method for intelligent flood drainage robots as described in claim 1, characterized in that, In the formation, the robot closest to the drainage location is designated as the leader robot, and the rest are designated as follower robots.

5. The adaptive formation obstacle avoidance method for intelligent flood drainage robots as described in claim 1, characterized in that, The obstacle avoidance grid includes a first obstacle avoidance grid, a second obstacle avoidance grid, and a third obstacle avoidance grid, which correspond to obstacle avoidance without obstacles, obstacle avoidance that meets collision constraints, and obstacle avoidance in feasible areas under the condition of multi-robot cooperative obstacle avoidance, respectively.

6. The adaptive formation obstacle avoidance method for intelligent flood drainage robots as described in claim 5, characterized in that, The third obstacle avoidance grid is obtained by deleting common areas in the second obstacle avoidance grids of multiple robots, and is then prioritized based on the robot obstacle avoidance weights to achieve multi-robot collaborative obstacle avoidance.

7. The adaptive formation obstacle avoidance method for intelligent flood drainage robots as described in claim 6, characterized in that, The obstacle avoidance weight is calculated based on the robot's working time and current battery level, using the following formula: w represents the robot's obstacle avoidance weight, t represents the robot's working time, and h represents the robot's current battery level.

8. An adaptive formation obstacle avoidance system for intelligent flood drainage robots, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the adaptive formation obstacle avoidance method for intelligent drainage robots as described in any one of claims 1-7.

Images