Multi-ship-brushing-robot collaborative operation path planning method

Abstract

The invention discloses a path planning method for collaborative operation of multiple ship washing robots, which belongs to the technical field of robot path planning, and comprises the following steps: initializing parameters, and collecting real-time data acquired by sensors carried on ship washing robots; making a robot group priority strategy; rasterizing the surface of the ship body by adopting a grid contour method; an improved # imgabs0 # algorithm is adopted to generate an initial path of the ship washing robot group; adding a constraint tree of a binary tree structure, and performing path planning; and when there is no obstacle on a front and back node connecting line of a certain node, deleting redundant nodes of the path, only retaining starting and ending points and inflection points, then deleting redundant inflection points, extracting key nodes as intermediate target points, and outputting an optimal path scheme. According to the technical scheme, the cooperative work efficiency of the multiple ship brushing robots can be improved, and the technical problems of dynamic obstacle avoidance response lag, low multi-machine cooperation efficiency and the like existing in traditional path planning in a complex curved surface environment are solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of robot path planning, and particularly relates to a multi-brush ship robot collaborative operation path planning method. Background Art

[0002] In the context of the increasingly frequent global shipping activities, the maintenance of ship navigation equipment has become a key issue in ensuring the efficiency of maritime transportation. When various ships are in the seawater immersion environment for a long time, complex marine biological communities will gradually grow in the hull immersion area, forming a destructive biological fouling layer. Attachment organisms represented by barnacles and serpulids secrete biological adhesives to form a high-strength bonding interface with the hull protective layer, making it difficult to effectively remove by conventional cleaning means.

[0003] The continuous accumulation of this stubborn biological fouling will cause multiple technical problems: First, the change in the hydrodynamic characteristics of the hull surface significantly increases the navigation resistance, resulting in a sharp increase in fuel consumption; Second, the weight load of the biological community not only changes the ship's load distribution, but may also cause safety hazards such as center of gravity shift, seriously weakening the navigation stability; Finally, the electrochemical interaction between the attachment and the metal hull will also accelerate the corrosion process. Therefore, building an efficient hull biological cleaning system has become an important direction for the development of modern ship maintenance technology, and its technical breakthrough has double practical significance for improving shipping economy and navigation safety.

[0004] Brush ship robots are a type of special robots that have emerged in the shipping industry and the marine engineering field in recent years. Their R & D background is closely related to ship maintenance needs, the upgrading of environmental protection regulations, and the development of automation technology. Ships sail in the marine environment for a long time, and it is very easy for marine organisms such as algae and shellfish to attach to the hull surface (i.e., biological fouling). This will not only increase the navigation resistance, resulting in an increase in fuel consumption and carbon emissions, but may also accelerate the hull corrosion and threaten navigation safety. Traditional manual cleaning methods are inefficient, costly, and rely on divers to work underwater, posing safety hazards. With the improvement of the International Maritime Organization (IMO)'s requirements for ship energy efficiency and environmental protection (such as the "Anti-fouling Systems Convention"), the industry urgently needs more efficient and sustainable hull cleaning solutions, which has become the direct driving force for the birth of brush ship robots.

[0005] In the field of underwater maintenance of brush ship robots, the traditional single-machine operation mode is limited by problems such as the equipment movement space constraint, the operation efficiency bottleneck, and the trajectory conflict, and there is an urgent need to innovate the operation mode through multi-robot collaborative path planning technology.

[0006] Therefore, aiming at the key technical challenges such as cleaning blind spots, rising operation and maintenance costs, and insufficient adaptive ability for complex curved surfaces commonly existing in the hull outer shell biofouling removal project, carrying out the research on the cooperative operation path planning of multi-brush ship robots has important engineering value. Through constructing a dynamic task allocation model and a distributed conflict resolution mechanism, this technology can realize the spatio-temporal cooperative optimization of the multi-robot operation trajectories, effectively eliminate the coverage blind spots in the traditional serial operation mode, and at the same time significantly reduce the risk of manipulator interference through kinematic coupling planning. Compared with the discretized operation scheme, the cooperative path planning can systematically shorten the dry-docking operation cycle and optimize the energy consumption distribution of the high-pressure water jet system, providing key technical support for the intelligent transformation of the ship green maintenance system. Summary of the Invention

[0007] In view of the above problems existing in the prior art, the present invention proposes a cooperative operation path planning method for multi-brush ship robots, which is reasonably designed, solves the deficiencies of the prior art, and has good effects.

[0008] A cooperative operation path planning method for multi-brush ship robots includes the following steps: Step 1, initialize the parameters, where the parameters include: hull surface map information, terrain obstacles and constraint information, and the start and end point information of the ship cleaning robots; Step 2, collect the real-time data collected by the sensors carried on the ship cleaning robots. The sensors include encoders, IMUs, and temperature and humidity sensors; the encoders collect the real-time speed of the ship cleaning robots during operation, the IMUs collect the position coordinates and attitude information of the ship cleaning robots, and the temperature and humidity sensors monitor the temperature and humidity changes of the seawater; Step 3, determine the number of ship cleaning robots, allocate the number of robots of the same type according to the task scale, and formulate a priority strategy for the robot groups to ensure that there are no conflicts in the paths during multi-robot cooperation; Step 4, rasterize the hull surface using the grid contour method; Step 5, use the improved algorithm to generate the initial path of the ship cleaning robot group; Step 6, add a constraint tree with a binary tree structure to detect whether there are conflicts in the solutions of each node in the initial path of the ship cleaning robot population; when a conflict is detected in the current path, generate two child nodes for each conflict and add constraints respectively, and re-plan the affected paths for the robots with lower priority; the algorithm terminates until all paths have no conflicts and the total cost is optimal; Step 7, perform smoothing optimization processing on the path, traverse all the nodes on the robot path, when there are no obstacles on the line connecting the front and rear nodes of a certain node, delete the redundant nodes of the path, only retain the start and end points and the inflection points, and then delete the redundant inflection points and extract the key nodes as the intermediate target points to output the optimal path plan.

[0009] Furthermore, step 5 includes the following sub-steps: Step 5.1: Initialize the generation sets Open List and Close List. Open List stores all nodes to be expanded, and Close List stores all expanded nodes. Insert the start node into Open List. Step 5.2: When Open List is not empty, select the node with the lowest total cost from Open List. The total cost of the nth node's total cost The calculation formula is: ; where is the actual path cost from the starting point to the current point, is the estimated path cost from the current point to the target point; Optimize using the Q-Learning algorithm, using the Euclidean distance formula, The expression is: ; where is the coordinate of the current node; is the coordinate of the start node; Delete the node with the lowest total cost from Open List, add it to Close List, and use it as the current node; Step 5.3: If the current node is the target node, backtrack the parent nodes to generate a path and end the algorithm. If the current node is not the target node, proceed to the next step; Step 5.4: Find all adjacent nodes of the current node. For each adjacent node, first check if it is in Close List. If it is in Close List, skip this node. If not, calculate the total cost of the adjacent node ; Then check if it is in Open List. If not, add it to Open List and update the total cost and parent node of the adjacent node. If it is in Open List, compare the current total cost and the previously calculated total cost . If the current total cost is smaller, update the parent node and update the total cost of the adjacent node to ; Step 5.5: Return to Step 5.2 to continue the loop until the target node is found.

[0010] Furthermore, in step 5.2, optimizing using the Q-Learning algorithm includes the following sub-steps: Step 5.2.1: Initialize the parameters of the Q-Learning algorithm, the obstacle information in the environment, the number of robots, and the coordinates of the starting point and the target point; Construct the basic environment model for robot path planning; Discretize the operation area into uniform grids, and label each grid as free space, obstacle, or task target point; The robot state space is defined as the two-dimensional coordinates (x, y) and the obstacle distribution within the surrounding 3×3 neighborhood; Define the action space of the robot , For the corresponding robot, move one step from the current grid cell to the adjacent th direction, corresponding to the eight-neighborhood movement direction set. Feasibility verification is required before each action: the target grid needs to be within the map boundary and not occupied by static obstacles; Step 5.2.2: Use the R-Learning algorithm to obtain the optimal strategy by maximizing the reward value and introducing a collision avoidance reward and punishment mechanism. Specifically: The total reward obtained by the brushing robot in the current state s after performing the action a through the selected strategy consists of the collision avoidance reward and the step length reward, and the expression is: ; Among them, is the sparse reward function, and the expression is: ; is the distance reward function, and the expression is: ; In the formula: is the reward coefficient, is the Euclidean distance between the target starting point and the end point; During the training process, the Q value is updated following the Bellman optimal equation, and the expression is: ; In the formula, , are the states at the t-th iteration and the (t + 1)-th iteration; , are the actions at the t-th iteration and the (t + 1)-th iteration respectively, and are the learning rate and the discount coefficient respectively, and the value range is [0, 1], is the immediate reward obtained by the particle when performing the action in the state , is the state of the particle Take action The expected Q-value is the maximum expected future reward for all possible actions in the new state; Step 5.2.3: Obtain the Q-values of each state-action pair through Q-Learning training. Traverse all grid nodes and record the maximum Q-value of the th grid node. Store the maximum Q-value result in a matrix, where the index of the matrix is the node coordinates and the value is the maximum Q-value corresponding to the node; Embed the trained Q-values into the algorithm to reconstruct the heuristic function , and its formula is: ; In the formula: is the maximum Q-value that can be obtained by choosing the optimal action starting from the node. The Q-value represents the long-term benefit in historical experience; is the Euclidean distance from the current node to the target node, which is used to dynamically adjust the weight according to the distance from the node to the target; The parameter is the attenuation coefficient, and its value range is [0.3, 1].

[0011] Furthermore, the said Step 6 includes the following sub-steps: Step 6.1: Establish a constraint tree in the form of a binary tree based on the conflict search mechanism. Each constraint tree node N includes three pieces of information: a constraint set, a solution to the problem, and the cost of solving the problem by this node. The constraint set contains constraints on all scrubbing robots in the problem; Step 6.2: In the algorithm initialization stage, generate an unconstrained optimal path for each scrubbing robot through the improved algorithm to form an initial solution set; Step 6.3: Start the conflict detection mechanism and describe the method in the form of a 4-tuple: A conflict is a 4-tuple , indicating that and occupy the node simultaneously at ; Step 6.4: When a conflict is detected, activate the branch generation mechanism of the constraint tree: For the conflict tuple , two child nodes Nc1 and Nc2 will be generated by expansion. The two child nodes inherit all the constraints and solutions of the parent node; Step 6.5: Add a new constraint to the node Nc1, and add a new constraint to Nc2; Step 6.6: Perform local path planning for the scrubbing robots with lower priority according to the constraints, while keeping the original paths of other scrubbing robots unchanged.

[0012] Beneficial technical effects brought by the present invention: The present invention combines algorithm with Q-Learning algorithm. In view of the situation that the algorithm is prone to falling into local optimal solutions during the solving process, the Q-Learning algorithm is adopted to drive the iterative update of the Q-table through multiple reward functions. During the algorithm integration stage, the Q value is mapped to the dynamic heuristic function H(n), which significantly improves the search efficiency and obstacle avoidance ability of the algorithm in complex environments. In terms of path collaborative planning, by creating a robot priority mechanism, this strategy divides multiple scrubbing robots into different priorities and adds a constraint tree with a binary tree structure. When it is detected that the paths of different robots conflict, two child nodes are generated for each conflict and constraints are added respectively, and the paths of the robots with lower priorities are replanned to avoid mutual influence among multiple robots during collaborative operations. The technical solution of the present invention can improve the collaborative working efficiency of multiple scrubbing robots and solve technical problems such as the lag in dynamic obstacle avoidance response and the low efficiency of multi-robot cooperation existing in traditional path planning in complex curved surface environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 It is a flowchart of basic information preprocessing in the present invention.

[0014] Figure 2 It is a flowchart of generating the initial path of the scrubbing robot group by using the improved algorithm in the present invention.

[0015] Figure 3 It is a flowchart of outputting the optimal path plan in the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0016] The following further describes the specific embodiments of the present invention in conjunction with specific embodiments: A path planning method for multi-scrubbing robot collaborative operation includes the following steps: Step 1, initialize the parameters, including: hull surface map information, terrain obstacles and constraint information, and scrubbing robot start and end point information; Step 2, collect the real-time data collected by the sensors carried on the scrubbing robot. The sensors include an encoder, an IMU, and a temperature and humidity sensor; the encoder collects the real-time speed of the scrubbing robot during operation, the IMU collects the position coordinates and attitude information of the scrubbing robot, and the temperature and humidity sensor monitors the temperature and humidity changes of seawater; Among them, the encoder monitors the rotational speed and displacement of the robot's drive wheels, feeds back the motion state through real-time speed data (such as linear speed and angular speed), and corrects the speed deviation in real time by combining the preset path planning to reduce the path tracking error; the IMU provides three-dimensional attitude angles (pitch angle, roll angle, yaw angle) and acceleration information, constructs a real-time six-degree-of-freedom model of the robot's pose, and dynamically adjusts the contact pressure of the cleaning robotic arm in combination with the attitude data during the robot's complex curved surface operation to avoid cleaning omissions or equipment wear caused by uneven surfaces; the temperature and humidity sensor monitors the seawater temperature, air humidity, and surface condensation state, reflecting the real-time changes in the robot's working environment. Through multi-sensor data fusion, a redundant perception system is constructed to reduce the failure risk caused by a single sensor failure.

[0017] Step 3: Determine the number of ship-washing robots, allocate the number of robots of the same type according to the task scale, and number them as , , etc. to represent different individuals, formulate a priority strategy for the robot group, and allocate priorities in the order of the numbers (such as > > ) to ensure that there are no path conflicts during multi-robot cooperation; in the priority planning framework, after the path planning of the high-priority robot is successful, its path is locked, and the path information that has been planned is updated before the path planning of the next robot to avoid collisions.

[0018] Step 4: Use the grid contour method to rasterize the hull surface. The specific content is as follows: Obtain the geometric data of the hull surface, select one of the geometric reference points as the origin, determine the grid size according to the cleaning range of the robot, and generate a grid that covers and surrounds the entire hull surface area to be cleaned, as shown in Figure 1 ; Specifically, first obtain the two-dimensional boundary coordinates of the hull surface to be cleaned. The contour of the hull is defined by a series of points, forming a closed polygon; then determine the resolution of the grid, as well as the number of rows and columns required to cover the entire hull; according to the set parameters, create a grid that covers the entire hull area to be cleaned. By discretizing the continuous surface into grid cells, simplifying complex geometric problems into discrete topological problems, and dynamically allocating tasks in combination with the priority strategy, the dimension of path search can be reduced, and the calculation time can be significantly shortened.

[0019] Step 5: Use the improved algorithm to generate the initial path of the ship-washing robot group; The implementation of the algorithm relies on two lists, namely the open list and the closed list. The open list stores a list of nodes to be inspected, through which the path may pass. The closed list stores the nodes that have been visited. First, the nodes adjacent to the starting point and that can be accessed are saved in the open list, and they are sorted in ascending order of the evaluation value in the list. Since the node with the smallest evaluation value is the first node, only the first node in the list is the node to be visited. Then, if an adjacent node is the best node, this node is saved to the closed list, and at the same time, the adjacent optional nodes with this node as the parent node are added to the open list, and continue to select until the target point of the search is reached.

[0020] The working area of the robot is divided by the grid method, and the working space is simplified into a two-dimensional grid. Each grid is numbered through a grid array to save its environmental information, which can represent the starting point, the target point, obstacles, and free areas. Through The algorithm starts from the starting point A and checks its adjacent nodes in turn, and then continuously expands outwards until the target point B is searched. The grids passed from point A to point B are the path between AB.

[0021] The correct selection of the evaluation function will be directly related to the success of the algorithm. The determination of the function is closely related to the actual situation. Therefore, the selection of the heuristic function is the key. An inappropriate heuristic function will lead to a decline in the quality of the algorithm path planning. The closer the estimated value is to the actual value, the more appropriate the heuristic function is selected. To avoid the algorithm converging prematurely to a suboptimal solution during the search process and thus falling into a local optimum, the Q-Learning algorithm is used to optimize the heuristic function H(n). The Q-Learning algorithm is based on the Bellman equation. Through the immediate reward after action selection and the optimal Q value of the next state, h(n)=max is iteratively updated a Q(s,a), making it gradually approach the minimum cumulative cost from the current state to the target. At the same time, the ε-greedy strategy is used to retain the exploration of unknown states, so as to reduce the expansion of invalid nodes in algorithms such as, and improve the search efficiency and path quality.

[0022] Such as Figure 2 shown, step 5 includes the following sub-steps: Step 5.1: Initialize and generate the sets Open List and Close List. Open List stores all nodes to be expanded, and Close list stores all expanded nodes. Insert the start node into Open List; Step 5.2: When the Open List is not empty, select the node with the lowest total cost from the Open List. The total cost of the th node's total cost The calculation formula is: ; Among them, is the actual path cost from the starting point to the current point, is the estimated path cost from the current point to the target point; Adopt the Q-Learning algorithm to optimize , Common Euclidean distance formula, Manhattan distance formula and Chebyshev distance formula are used. In the present invention, the Euclidean distance formula is adopted, The expression is: ; Among them, is the coordinate of the current node; is the coordinate of the starting node; Delete the node with the lowest total cost from the Open List, add it to the Close List, and use it as the current node; Step 5.3: If the current node is the target node, backtrack the parent node to generate a path and end the algorithm. If the current node is not the target node, then proceed to the next step; Step 5.4: Find all adjacent nodes of the current node. For each adjacent node, first determine whether it is in the Close List. If it is in the Close List, skip this node. If not, calculate the total cost of the adjacent node ; Secondly, calculate the total cost of the adjacent node; Then determine whether it is in the Open List. If not, add it to the Open List and update the total cost and parent node of the adjacent node. If it is in, compare the current total cost and the previously calculated total cost . If the current total cost is smaller, update the parent node and update the total cost of the adjacent node to ; Step 5.5: Return to Step 5.2 and continue to loop until the target node is found.

[0023] In Step 5.2, it is necessary to calculate the total cost of the node , where the heuristic function When the admissibility criterion (i.e., always not higher than the true path cost) is satisfied and the consistency condition is met, the algorithm can effectively guarantee the generation of a globally optimal path. However, in practical engineering applications, if the parameter settings of the heuristic function are improper or there are biases in the heuristic guidance direction, it may lead to premature convergence of the search process to a suboptimal solution, manifested as the algorithm performing redundant expansions in a specific area and being unable to break through the constraints of local extrema, thus falling into a local optimum, which in turn affects the quality and efficiency of the overall path planning. Using the Q-Learning algorithm to optimize the heuristic function H(n), the Q-Learning algorithm is based on the Bellman equation. Through the immediate reward after action selection and the optimal Q value of the next state, it iteratively updates h(n)=max a Q(s,a), making it gradually approach the minimum cumulative cost from the current state to the goal. At the same time, the ε-greedy strategy is used to retain the exploration of unknown states, thereby reducing the expansion of invalid nodes in and other algorithms, improving the search efficiency and path quality. The specific process includes the following steps: Step 5.2.1: Initialize the parameters of the Q-Learning algorithm, the obstacle information in the environment, the number of robots, and the coordinates of the starting point and the target point; Construct the basic environment model for robot path planning; Discretize the working area into uniform grids, and each grid is labeled as free space, obstacle, or task target point; The robot state space is defined as the two-dimensional coordinates (x,y) and the obstacle distribution within its surrounding 3×3 neighborhood; Define the action space of the robot , For the corresponding robot to move one step from the current grid cell to the adjacent th direction, corresponding to the eight-neighborhood movement direction set (orthogonal directions: N, S, E, W; diagonal directions: NE, NW, SE, SW). Before each action is executed, a feasibility verification is required: the target grid needs to be within the map boundary and not occupied by static obstacles; Step 5.2.2: Use the R-Learning algorithm to obtain the optimal strategy by maximizing the reward value and introducing a collision avoidance reward and punishment mechanism, specifically: The total reward obtained by the scrubber robot in the current state s after performing action a after selecting the strategy consists of a collision avoidance reward and a step length reward, and the expression is: ; Among them, For the sparse reward function, the target reward is used to directly encourage the convergence of the path end point, and the collision penalty is used to avoid path collisions. Usually, Q-Learning selects a fixed value as the reward function. The reward value for an action reaching the end point is 100, the reward value for an action without collision is -1, and the reward value for an action with collision is -50. The expression is: ; For the distance reward function, in the long-distance path planning scenario of the ship brushing robot, in the initial stage, due to the lack of an effective guidance mechanism, the exploration process shows significant randomness, resulting in a large number of invalid iterations and high computational resource consumption. To optimize the algorithm performance, a distance reward and penalty mechanism is introduced to improve the reward function of the Q-Learning algorithm. This mechanism establishes the following evaluation function through mathematical modeling: ; In the formula: is the reward coefficient, is the Euclidean distance between the target start point and the end point. The distance reward can enable the Q-Learning algorithm to reduce the calculation time, eliminate redundant paths, and optimize the path efficiency. In the initial exploration of the ship brushing robot, it is more inclined to select actions close to the target point; During the training process, the Q value is updated following the Bellman optimal equation, and the expression is: ; In the formula, , are the states at the t-th iteration and the (t + 1)-th iteration; , are the actions at the t-th iteration and the (t + 1)-th iteration respectively, and are the learning rate and the discount coefficient respectively, and their value ranges are [0, 1], is the immediate reward obtained by the particle when executing the action in the state , is the expected Q value of the particle when taking the action in the state , is the maximum expected future reward for all possible actions in the new state; Step 5.2.3: Obtain the Q value of each state-action pair through Q-Learning training. Traverse all grid nodes and record the maximum Q value of the th grid node. Store the maximum Q value result in a matrix. The index of the matrix is the node coordinates, and the value is the maximum Q value corresponding to the node. Embed the trained Q value into algorithm to reconstruct the heuristic function H(n), and its formula is: ; where: is the maximum Q value that can be obtained by selecting the optimal action starting from the node, and the Q value represents the long-term benefit in historical experience; is the Euclidean distance from the current node to the target node, which is used to dynamically adjust the weight according to the distance from the node to the target; the parameter is the attenuation coefficient, and its value range is [0.3, 1].

[0024] Step 6: Use a constraint tree with a binary tree structure to detect whether there are conflicts in the solutions of each node in the initial path of the ship brushing robot population; when a conflict in the current path is detected, generate two child nodes for each conflict and add constraints respectively, and re-plan the affected paths for the robots with lower priorities; the algorithm terminates until all paths have no conflicts and the total cost is optimal.

[0025] This mechanism realizes the conflict resolution and global optimization of the group path through a three-stage progressive process: First, establish a spatio-temporal dimension conflict detection model for the initial path set, and use a quadtree spatial index and a time window prediction algorithm to identify resource competition problems in the node solution space; secondly, construct a two-way constraint branch for the detected conflict events - each conflict point generates left and right child nodes with a logical inheritance relationship, the left branch inherits the physical constraints of the parent level and superimposes dynamic priority rules, and the right branch inherits the timing constraints and expands the safety buffer interval; finally, implement a priority-oriented path correction strategy to re-plan the paths for the robots with lower priorities in the conflict until all paths have no conflicts and the total cost is optimal, and the algorithm terminates.

[0026] As Figure 3 shown, Step 6 includes the following sub-steps: Step 6.1: Establish a constraint tree in the form of a binary tree based on the conflict search mechanism. Each constraint tree node N includes three pieces of information: a constraint set, a solution to the problem, and the cost of this node to solve the problem. The constraint set contains the constraints on all ship brushing robots in the problem; Step 6.2: In the initialization stage of the algorithm, use the improved algorithm to generate an unconstrained optimal path for each ship brushing robot to form an initial solution set; Step 6.3: Start the conflict detection mechanism and use a 4-tuple form description method: A conflict is a 4-tuple , indicating and simultaneously occupy the node at the time; Step 6.4: When a conflict is detected, activate the branch generation mechanism of the constraint tree: For the conflict tuple , the expansion generates two child nodes Nc1 and Nc2, and the two child nodes inherit all the constraints and solutions of the parent node; Step 6.5: Add a new constraint to node Nc1 , and add a new constraint to Nc2 ; Step 6.6: Perform local path planning for the low-priority ship-brushing robots according to the constraints, while keeping the original paths of other ship-brushing robots unchanged.

[0027] Step 7: Perform smoothing optimization on the path. Traverse all the nodes on the robot path. When there are no obstacles on the line connecting the front and rear nodes of a certain node, delete the redundant nodes on the path, only keep the starting point, the ending point and the inflection points, and then delete the redundant inflection points and extract the key nodes as the intermediate target points, and output the optimal path plan.

[0028] Perform smoothing optimization on the generated path. Path optimization is an important post-processing link in the motion planning of mobile robots. Traditional The initial paths generated by algorithms usually have problems of redundant turning points and multi-segment line connections, resulting in the trajectory deviating from the optimal kinematic characteristics. To improve the continuity and executability of the path, this study proposes a path smoothing strategy based on node optimization. Specifically, during implementation, first perform a traversal analysis on the path node sequence. If the line connecting adjacent nodes does not intersect with the obstacle area, remove the intermediate node. By iteratively executing this optimization process, the unnecessary inflection points in the path can be effectively eliminated, and finally a smooth trajectory that satisfies the kinematic constraints is generated.

[0029] Of course, the above description is not a limitation of the present invention, and the present invention is not limited to the above examples. Changes, modifications, additions or substitutions made by those skilled in the art within the essence of the present invention should also fall within the protection scope of the present invention.

Claims

1. A path planning method for collaborative operation of multi-brush ship robots, characterized in that, It includes the following steps: Step 1: Initialize the parameters, including the hull surface map information, terrain obstacles, constraint information, and the starting and ending points of the ship washing robot; Step 2: Collect the real-time data collected by the sensors carried on the ship washing robot. The sensors include an encoder, an IMU, and a temperature and humidity sensor. The encoder collects the real-time speed of the ship washing robot during operation, the IMU collects the position coordinates and attitude information of the ship washing robot, and the temperature and humidity sensor monitors the temperature and humidity changes of the seawater; Step 3: Determine the number of ship washing robots, allocate the number of robots of the same type according to the task scale, and formulate a priority strategy for the robot group to ensure that there is no conflict in the paths during multi-robot cooperation; Step 4: Use the grid contour method to rasterize the hull surface; Step 5: Use the improved algorithm to generate the initial path of the ship washing robot group; Step 6: Add a constraint tree with a binary tree structure to detect whether there is a conflict in the solutions of each node in the initial path of the ship brushing robot population. When a conflict in the current path is detected, two child nodes are generated for each conflict and constraints are added respectively, and the paths affected by the robot with a lower priority are re-planned. The algorithm terminates until all paths have no conflicts and the total cost is optimal; Step 7: Perform a smoothing optimization process on the path. Traverse all the nodes on the robot path. When there are no obstacles on the line connecting the previous and next nodes of a certain node, delete the redundant nodes of the path, only retain the starting and ending points and the inflection points, and then delete the redundant inflection points to extract the key nodes as intermediate target points, and output the optimal path plan.

2. The method for collaborative operation path planning of a multi-brush ship robot according to claim 1, characterized in that The said Step 5 includes the following sub-steps: Step 5.1: Initialize and generate the Open List and Close List. The Open List stores all the nodes to be expanded, and the Close List stores all the expanded nodes. Insert the starting node into the Open List; Step 5.2: When the Open List is not empty, select the node with the lowest total cost from the Open List. The total cost of the total cost of the calculation formula is: ； Among them, is the actual path cost from the starting point to the current point, is the estimated path cost from the current point to the target point; The Q-Learning algorithm is used for optimization , The Euclidean distance formula is used, The expression is: ； Among them, is the coordinate of the current node; is the coordinate of the starting node; Delete the node with the lowest total cost from the Open List, add it to the Close List, and use it as the current node; Step 5.3: If the current node is the target node, backtrack the parent node to generate a path and end the algorithm. If the current node is not the target node, then proceed to the next step; Step 5.

4. Find all adjacent nodes of the current node. For each adjacent node, first check whether it is in the CloseList. If it is in the Close List, skip this node. If not, calculate the total cost of the adjacent node ; Then check whether it is in the Open List. If not, add it to the Open List and update the total cost and parent node of the adjacent node. If it is in, compare the current total cost with the previously calculated total cost . If the current total cost is smaller, update the parent node and update the total cost of the adjacent node to ; Step 5.5: Return to Step 5.2 and continue to loop until the target node is found.

3. A method for collaborative operation path planning of a multi-brush ship robot according to claim 2, characterized in that, In step 5.2, the Q-Learning algorithm is used for optimization , including the following sub-steps: Step 5.2.1: Initialize the parameters of the Q-Learning algorithm, the obstacle information in the environment, the number of robot information, and the coordinates of the starting point and the target point; Construct the basic environment model for robot path planning; Discretize the working area into uniform grids, and each grid is labeled as free space, obstacle, or task target point; The robot state space is defined as the two-dimensional coordinates (x, y) and the obstacle distribution within its surrounding 3×3 neighborhood; Define the action space of the robot , corresponds to the robot moving one step from the current grid cell to the th adjacent direction, corresponding to the eight-neighborhood movement direction set. Feasibility verification is required before each action: the target grid needs to be within the map boundary and not occupied by static obstacles; Step 5.2.2: Use the R-Learning algorithm to obtain the optimal strategy by maximizing the reward value and introducing a collision avoidance reward and punishment mechanism. Specifically: The total reward obtained by the ship brushing robot in the current state s after performing the action a through the selection strategy consists of a collision avoidance reward and a step reward, and the expression is: ； Among them, is a sparse reward function, and its expression is: ； is the distance reward function, and its expression is: ； In the formula: is the reward coefficient, is the Euclidean distance between the target starting point and the ending point; During the training process, the Q value update follows the Bellman optimal equation, and the expression is: ； Wherein, and are the states at the t-th iteration and the (t + 1)-th iteration; and are the actions at the t-th iteration and the (t + 1)-th iteration respectively, and are the learning rate and the discount factor respectively, and their value ranges are [0, 1], is the immediate reward obtained when the particle executes the action in the state , is the expected Q value when the particle takes the action in the state , is the maximum expected future reward for all possible actions in the new state; Step 5.2.3: Obtain the Q-values of each state-action pair through Q-Learning. Traverse all grid nodes and record the maximum Q-value of the th grid node . Store the maximum Q-value result in a matrix, where the index of the matrix is the node coordinates and the value is the maximum Q-value corresponding to the node. Embed the trained Q-values into the algorithm to reconstruct the heuristic function , and its formula is: ； Wherein: is the maximum Q value that can be obtained by selecting the optimal action starting from the node, and the Q value represents the long-term benefit in historical experience; is the Euclidean distance from the current node to the target node, which is used to dynamically adjust the weight according to the distance from the node to the target; parameter is the attenuation coefficient, and its value range is [0.3, 1].

4. A method for collaborative operation path planning of a multi-brush ship robot according to claim 3, characterized in that, The said Step 6 includes the following sub-steps: Step 6.1: Establish a constraint tree in the form of a binary tree based on the conflict search mechanism. Each constraint tree node N includes three pieces of information: a constraint set, a solution to the problem, and the cost of the node to solve the problem. The constraint set contains constraints on all the ship-cleaning robots in the problem. Step 6.2: In the algorithm initialization stage, an improved algorithm generates an unconstrained optimal path for each brush boat robot to form an initial solution set; Step 6.3: Start the conflict detection mechanism and describe the method in the form of a quadruple: A conflict is a quadruple , indicating and both occupy the node at the same time at the moment ; Step 6.4: When a conflict is detected, activate the branch generation mechanism of the constraint tree: for the conflicting tuple , two child nodes Nc1 and Nc2 will be generated by extension, and the two child nodes inherit all the constraints and solutions of the parent node; Step 6.5: Add a new constraint to node Nc1 , and add a new constraint to Nc2 ; Step 6.6: Perform local path planning for the low-priority ship-cleaning robots according to the constraints, while keeping the original paths of other ship-cleaning robots unchanged.

Images