Method for multi-robot cooperative path planning for cleaning of ocean-going container ships

Abstract

This invention relates to the field of path planning technology, specifically to a multi-robot collaborative path planning method for cleaning ocean-going container ships. The invention selects the robot to be reset at the current moment. Based on the difference in the direction of external environmental forces between the robot's current working area and its adjacent working areas, and the traction trend of the external environmental forces in the adjacent working areas on the predetermined track, it selects alternative detour areas for the robot at the current moment. Based on the distance the robot needs to enter the alternative detour areas, and the degree of conformity of the environmental forces in the alternative detour areas to the entry heading, it selects the optimal detour area from the alternative detour areas. It then plans the return path of the robot to its predetermined track via the optimal detour area and controls the robot to move along the return path. This invention utilizes external environmental forces to plan the optimal detour path, converting water flow resistance into thrust, achieving low-consumption and high-efficiency off-track reset.

Description

Technical Field

[0001] This invention relates to the field of path planning technology, specifically to a multi-robot collaborative path planning method for cleaning ocean-going container ships. Background Technology

[0002] During long voyages, marine organisms easily accumulate on the hull of ocean-going vessels, increasing drag, fuel consumption, and causing corrosion. Therefore, timely cleaning of these organisms is crucial for the efficiency and safety of navigation.

[0003] Currently, multiple underwater robots often work together to efficiently clean ship surfaces along pre-planned paths. However, during cleaning operations, underwater cleaning robots are highly susceptible to deviation from their intended cleaning trajectory due to complex hydrodynamic factors such as ocean currents, wave impacts, and the ship's own navigation. To address this deviation, existing technologies typically use the geometric shortest path—the line connecting the deviance point and the current position—to control the robot's linear repositioning and facilitate rapid resumption of work. However, this conventional linear repositioning strategy ignores the root cause of the robot's deviation: the powerful impact of external flow fields such as waves and ocean currents. In this situation, forcing repositioning along the geometric shortest path often means the robot must directly confront enormous flow resistance, leading to a significant increase in propulsion energy consumption and extremely slow repositioning due to excessive resistance. Summary of the Invention

[0004] To address the technical problems of excessive energy consumption and slow recovery caused by conventional shortest line regression strategies when underwater cleaning robots deviate from their tracks due to water flow resistance, this invention aims to provide a multi-robot collaborative path planning method for cleaning ocean-going container ships. The specific technical solution adopted is as follows: One embodiment of the present invention provides a multi-robot collaborative path planning method for cleaning ocean-going container ships, the method comprising: On a two-dimensional grid map of the surface of an ocean-going container ship, work areas and predetermined tracks are assigned to multiple robots used for cleaning the ship; the actual position of each robot during the cleaning process is obtained in real time. Based on the degree of deviation of the actual position of each robot from its predetermined trajectory and the degree of heading deviation within the preset analysis period at the current moment, the robot to be reset at the current moment is selected. Based on the difference in the direction of external environmental forces between the working area where the robot to be reset is located at the current moment and its adjacent working areas, as well as the traction trend of the external environmental forces in the adjacent working areas on the predetermined track, the alternative detour area of ​​the robot to be reset at the current moment is selected. Based on the distance cost of the robot to be reset entering each alternative detour area, and the degree of conformity of the environmental external forces in the corresponding alternative detour area to the cutting-in heading, the optimal detour area is selected from the alternative detour areas of the robot to be reset at the current moment. Plan the return path of the robot to be reset through the optimal detour area to its predetermined track, and control the robot to be reset to move along the return path.

[0005] Furthermore, the selection of the robot to be reset at the current moment includes: The shortest distance from the actual position of each robot to its predetermined trajectory at each moment is recorded as the off-track distance. For each robot, if the deviation distance of the robot at each time within the preset analysis period at the current time is greater than the preset allowable deviation threshold, and the deviation distance at the time before the start time of the preset analysis period is less than or equal to the preset allowable deviation threshold, then the first-order difference sequence of the deviation distance of the robot at each time within the preset analysis period is obtained. The sum of the positive elements and the sum of the absolute values ​​of the negative elements in the first-order difference sequence are respectively counted and recorded as the off-track cumulative amount and the return-track cumulative amount; the ratio of the off-track cumulative amount to the return-track cumulative amount is taken as the net deviation trend ratio of the robot at the current moment. Based on the difference between the robot's actual displacement direction and the ideal displacement direction during the preset analysis period at the current moment, the heading coincidence coefficient of the robot at the current moment is obtained; Based on the net deviation trend ratio and the heading coincidence coefficient, the robot's comprehensive deviation index at the current moment is obtained; the robot whose comprehensive deviation index is greater than the preset deviation threshold is selected as the robot to be reset at the current moment.

[0006] Furthermore, obtaining the robot's heading alignment coefficient at the current moment includes: The product of the robot's ideal speed during the preset analysis period and the duration of the preset analysis period is calculated as the expected displacement. The tangent direction of the point on the robot's predetermined track that is closest to its actual position at the start of the preset analysis period is taken as the ideal direction of travel. Starting from the actual position at the start of the period, the expected displacement is extended along the ideal direction of travel to obtain the robot's expected position at the current moment. The actual displacement vector and the ideal displacement vector are respectively denoted as the actual displacement vector and the ideal displacement vector, respectively, pointing from the robot's actual position at the start of the preset analysis period at the current moment to the actual position and the expected position at the current moment. The cosine similarity between the actual displacement vector and the ideal displacement vector is normalized to obtain the heading coincidence coefficient of the robot at the current moment.

[0007] Furthermore, the selection of the alternative detour area for the robot to be reset at the current moment includes: Calculate the difference vector between the actual displacement vector and the ideal displacement vector of the robot to be reset in the current time within a preset analysis period. Multiply the difference vector by the reciprocal of the number of time points in the preset analysis period to obtain the fluid disturbance vector of the working area where the robot to be reset is located at the current time. For each adjacent working area of ​​the working area where the robot to be reset is located at the current moment, the vector from the center point of the adjacent working area to the point on the predetermined track of the robot to be reset that is closest to the center point is used as the track traction vector of the adjacent working area. If the angle between the fluid disturbance vector of the working area where the robot to be reset is located at the current moment and the fluid disturbance vector of each of its adjacent working areas is an obtuse angle, and the dot product of the fluid disturbance vector of the adjacent working area and the track traction vector is greater than zero, then the adjacent working area is taken as the alternative detour area for the robot to be reset at the current moment.

[0008] Furthermore, the step of selecting the optimal detour area from the detour candidate area of ​​the robot to be reset at the current moment includes: For each alternative detour region of the robot to be reset at the current moment, the vector pointing from the actual position of the robot to be reset at the current moment to the point on the boundary of the alternative detour region that is closest to the actual position is used as the shortest boundary crossing vector of the alternative detour region. The ratio of the off-track distance of the robot to be reset at the current moment to the magnitude of the shortest out-of-bounds vector of each alternative detour region is used as the detour cost-benefit ratio of that alternative detour region. Calculate the fluid disturbance vector of the working area where the robot to be reset is located at the current moment, and the angles between the fluid disturbance vector of each alternative detour area and the shortest cross-boundary vector at that moment, respectively, and record them as the flow field deflection angle and the cross-boundary heading deflection angle. The flow field compliance of the robot to be reset in each alternative detour area at the current moment is obtained by comparing the negative of the cosine of the flow field deflection angle with the difference between the constant 1 and the cosine of the out-of-bounds heading angle. Based on the overall deviation index of the robot to be reset at the current moment, as well as the detour cost-benefit ratio and flow field compliance of each detour candidate area at that moment, the overall recommendation degree of the corresponding detour candidate area is obtained; From all the alternative routes available to the robot to be reset at the current moment, select the alternative route with the highest overall recommendation score that is greater than the preset recommendation threshold as the optimal alternative route for that moment.

[0009] Furthermore, the planned return path for the robot to be reset, via the optimal detour area to its predetermined orbit, includes: The point on the predetermined track of the robot to be reset that is closest to the actual position of the start time of the preset analysis period at the current moment is recorded as the reset point. Starting from the actual position of the robot to be reset at the current moment, and ending at the reset point; the point on the boundary of the optimal detour area of ​​the robot to be reset at the current moment that is closest to the starting point is recorded as the region entry node; the ant colony algorithm is used to plan the path from the starting point through the region entry node to the end point, which is used as the return path; Specifically, the heuristic factor in the ant colony algorithm is set to the reciprocal of the oncoming resistance cost; when performing path search between adjacent grid nodes using the ant colony algorithm, the vector pointing from the currently traversed grid node to the adjacent grid node to be transferred is calculated, and the angle between this vector and the fluid interference vector of the optimal bypass area is taken as the transfer flow field deflection angle; the cosine value of the transfer flow field deflection angle is negatively correlated and mapped, and the product of the mapping result and the distance between the corresponding adjacent grid nodes is set as the oncoming resistance cost between adjacent grid nodes.

[0010] Furthermore, if the overall recommendation degree of all alternative detour areas of the robot to be reset at the current moment is less than or equal to the preset recommendation threshold, then the straight line from the actual position of the robot to be reset at the current moment to the operation reset point is taken as the homing path.

[0011] Furthermore, when the shortest outbound vector is a zero vector, or when the cosine of the outbound heading angle is equal to 1, the comprehensive recommendation degree is assigned as the limit recommendation degree; wherein, the limit recommendation degree is greater than the preset recommendation threshold.

[0012] Furthermore, the net deviation trend ratio is positively correlated with the comprehensive deviation index, and the heading coincidence coefficient is negatively correlated with the comprehensive deviation index.

[0013] Furthermore, the current time is at the end of its preset analysis period.

[0014] The present invention has the following beneficial effects: In this embodiment of the invention, by comprehensively analyzing the degree of deviation of the robot's actual position from the predetermined track and the degree of heading deviation within a preset analysis period, the robot to be reset due to substantial deviation caused by the external force of water flow can be accurately located. When facing the need for resetting due to deviation, by comparing the difference in the direction of external environmental forces between the current working area and adjacent working areas, as well as the traction trend of external environmental forces in adjacent working areas on the predetermined track, alternative detour areas that can provide reverse water flow and help the robot move closer to the track are identified. Based on this, the spatial distance cost of entering each alternative detour area and the degree of conformity of external environmental forces to the entry heading are further comprehensively analyzed to accurately locate the optimal detour area with the least pre-existing physical consumption and the lowest cross-regional upstream resistance, and the final return path is planned accordingly. This solution cleverly transforms the environmental forces that cause the robot to deviate from its course into thrust, effectively solving the technical problem that the conventional shortest straight-line return strategy for underwater cleaning robots, which is forced to resist flow field resistance, results in excessive propulsion energy consumption and slow reset after deviating from the course due to water flow forces. This significantly improves the robot's reset efficiency and overall endurance in underwater operations. Attached Figure Description

[0015] Figure 1 A flowchart illustrating the steps of a multi-robot collaborative path planning method for cleaning ocean-going container ships, as provided in an embodiment of the present invention. Figure 2 A flowchart illustrating a method for acquiring a robot to be reset, provided in one embodiment of the present invention; Figure 3 This is a flowchart illustrating a method for obtaining an optimal detour area according to an embodiment of the present invention. Detailed Implementation

[0016] The following description, in conjunction with the accompanying drawings, details a specific scheme for a multi-robot collaborative path planning method for cleaning ocean-going container ships provided by the present invention.

[0017] Example 1: This invention proposes a multi-robot collaborative path planning method for cleaning ocean-going container ships. Please refer to [link / reference]. Figure 1 The diagram illustrates a flowchart of a multi-robot collaborative path planning method for cleaning ocean-going container ships, provided by an embodiment of the present invention. The method includes: Step S1: On a two-dimensional grid map on the surface of the ocean-going container ship, assign working areas and predetermined tracks to multiple robots used for cleaning the ocean-going container ship; obtain the actual position of each robot in real time during the cleaning process.

[0018] A 3D laser scanner was used to acquire 3D data of the bottom and sides of an ocean-going container ship, constructing a 3D point cloud model of the ship's surface. This 3D point cloud model was then unfolded into a 2D plane along the ship's longitudinal reference plane. Subsequently, a grid method was used to discretize this 2D plane, uniformly dividing it into multiple square grids to construct a 2D grid map of the ocean-going container ship's surface. The grid side length can be set according to the robot's effective cleaning coverage width, for example, 0.5 meters. To distinguish cleaning status, insurmountable physical obstacles were assigned a grid attribute value of 1, grids that had been cleaned and traversed were assigned a value of 2, and other feasible area grids were assigned a value of 0.

[0019] Based on a two-dimensional grid map, and according to the number of cleaning robots deployed, the entire two-dimensional grid map to be cleaned is divided into multiple non-overlapping sub-regions using an equal-division method, which serve as the working areas for each robot. Then, within each working area, a depth-first search algorithm is used to plan a travel route traversing all the grid cells to be cleaned within that area, serving as the predetermined track for the corresponding robot. During the actual cleaning operation, the inertial measurement units onboard the robots collect their physical coordinates in real time and map them onto the two-dimensional grid map to obtain the actual position of each robot.

[0020] Step S2: Select the robot to be reset at the current moment based on the degree of distance deviation and heading deviation of each robot's actual position relative to its predetermined trajectory within the preset analysis period at the current moment.

[0021] The deviation of a robot from its intended path underwater due to fluid disturbances is a gradual and cumulative process. Complex ocean currents not only cause lateral spatial displacement errors in the robot but also induce distortions and deflections in its orientation. By comprehensively considering the degree of distance deviation and heading deflection of the robot's actual position within a preset analysis period, we can analyze the robot's deviation from its intended path from both spatial distance and orientation dimensions. This allows for precise quantification of the robot's true state of loss of control and deterioration, identifying robots that are truly in a state of continuous deterioration and complete derailment and require resetting.

[0022] In this embodiment, the current time is the end time of its preset analysis period, which contains 10 times. The implementer can adjust the time according to the sea state level of the sea area where the ocean container ship is located. That is, the worse the sea state, the shorter the preset analysis period should be.

[0023] Step S3: Based on the difference in the direction of external environmental forces between the working area where the robot to be reset is located at the current moment and its adjacent working areas, and the traction trend of the external environmental forces in the adjacent working areas on the predetermined track, select the alternative detour area for the robot to be reset at the current moment.

[0024] The root cause of the robot's deviation from its intended path lies in the continuous entrainment by the harsh currents in its working area. If a conventional shortest-line return method were used, it would face immense resistance from the current. Analyzing the difference in the direction of external environmental forces between the robot's current working area and its adjacent areas allows us to determine whether the adjacent areas possess reverse currents that could help it escape the current unfavorable flow field. Analyzing the traction trend of external environmental forces in the adjacent working areas on the predetermined trajectory clarifies whether these reverse currents can provide a favorable thrust for the robot to return to its original path. Combining these two aspects allows us to accurately identify alternative detour areas that offer both escape and leverage for return, fundamentally breaking away from traditional escape strategies that rely on resisting external forces and laying a fluid dynamics foundation for low-energy repositioning.

[0025] In this embodiment, the boundary grid of the current working area is extracted, and the working areas of the remaining grids within the eight neighboring areas of the boundary grid are taken as the adjacent working areas of the current working area.

[0026] Step S4: Based on the distance cost of the robot to be reset entering each alternative detour area, and the degree of conformity of the environmental external forces in the corresponding alternative detour area to the cutting-in heading, select the optimal detour area from the alternative detour areas of the robot to be reset at the current moment.

[0027] While detouring through adjacent work areas can utilize water flow, cross-area detours incur a certain initial travel cost. To ensure the detour strategy is economical, it is necessary to comprehensively weigh the spatial distance cost of the robot's cross-area entry into each alternative area, as well as the degree of environmental assistance to the heading during the entry process. The optimal detour area with the highest global energy efficiency must be precisely identified, ensuring that the robot only needs to expend a very short initial journey and minimal cross-area resistance to avoid a long, high-resistance return journey, thereby maximizing the benefits of leveraging water flow.

[0028] Step S5: Plan the return path of the robot to be reset through the optimal detour area to its predetermined track, and control the robot to be reset to move along the return path.

[0029] After precisely identifying the optimal springboard, a return path is planned, bypassing the optimal detour area to return to the predetermined track, and the robot's movement along the path is controlled. This guides the robot to completely escape the unfavorable flow field and smoothly reconnect to the original cleaning trajectory with an extremely smooth, downstream attitude. This not only achieves a perfect closed loop for reset control but also effectively solves the core technical pain points of conventional linear return systems, such as excessive energy consumption and slow reset due to struggling against flow field resistance, at the physical engineering level. This significantly improves the overall cleaning efficiency and endurance of multi-robot systems.

[0030] Preferably, in some possible implementations of the embodiments of the present invention, the method for obtaining the robot to be reset is described in [reference needed]. Figure 2The diagram illustrates a flowchart of a method for acquiring a robot to be reset according to an embodiment of the present invention. The method includes: Step S210: Record the shortest distance from the actual position of each robot to its predetermined track at each moment as the off-track distance; for each robot, if the off-track distance of the robot at each moment in the preset analysis period at the current moment is greater than the preset allowable deviation threshold, and the off-track distance at the moment before the start of the preset analysis period is less than or equal to the preset allowable deviation threshold, then obtain the first-order difference sequence of the off-track distance of the robot at each moment in the preset analysis period.

[0031] It should be noted that the deviation distance of the robot from the track at the moment preceding the start time of the preset analysis period is less than or equal to the preset allowable deviation threshold, which is used to accurately pinpoint the moment the robot just deviates from the track; the deviation distance at each moment within the preset analysis period is greater than the preset allowable deviation threshold, which is used to confirm that the robot's deviation state has real continuity. The combination of the above conditions can accurately define the real process of the robot just experiencing a definite deviation and continuing to move outward.

[0032] In this embodiment, the preset allowable deviation threshold can be set empirically based on the grid accuracy of the two-dimensional grid map on the surface of the ocean container ship, for example, it can be set to 0.8 times the side length of a single grid.

[0033] Step S220: Calculate the sum of positive elements and the sum of the absolute values ​​of negative elements in the first-order difference sequence, and record them as the off-track cumulative amount and the return-track cumulative amount respectively; use the ratio of the off-track cumulative amount to the return-track cumulative amount as the robot's net deviation trend ratio at the current moment.

[0034] It should be noted that the first-order difference sequence characterizes whether the robot continuously moves away from the track or exhibits a physical tendency to move closer to the track between adjacent time points. Positive values ​​represent the robot moving away from the track, while negative values ​​represent the robot moving closer to the track. The cumulative amount of derailment and the cumulative amount of return to track represent the cumulative absolute displacement of the robot moving away and moving closer to the track within the preset analysis period, respectively. The ratio of these two values, i.e., the net deviation trend ratio, effectively isolates and offsets the interference of reciprocating oscillations caused by the water flow, accurately reflecting the robot's true net trend of outward deviation. The larger the net deviation trend ratio, the more severe the deviation trend, indicating that the robot is mainly being pulled away from the track by force, with little effect from struggling to pull back.

[0035] Step S230: Based on the difference between the robot's actual displacement direction and the ideal displacement direction during the preset analysis period at the current moment, obtain the robot's heading consistency coefficient at the current moment.

[0036] In this embodiment, the method for obtaining the heading coincidence coefficient includes: calculating the product of the robot's ideal speed and the duration of the preset analysis period as the expected displacement; taking the tangent direction of the point on the robot's predetermined track closest to the actual position at the start of the preset analysis period as the ideal direction of travel; extending the expected displacement along the ideal direction of travel from the actual position at the start of the preset analysis period to obtain the robot's expected position at the current moment; recording the vectors pointing from the robot's actual position at the start of the preset analysis period at the current moment to the actual position and the expected position at the current moment as the actual displacement vector and the ideal displacement vector, respectively; and normalizing the cosine similarity between the actual displacement vector and the ideal displacement vector to obtain the heading coincidence coefficient of the robot at the current moment.

[0037] It should be noted that the expected displacement represents the theoretical distance the robot will travel within a preset analysis period under conditions of no external environmental forces, i.e., the ideal displacement without interference. Considering that the robot may already have a slight positional deviation at the start of the preset analysis period, to accurately reflect its intended trajectory, the tangent direction of the point on the predetermined track closest to its actual position at that start time is selected as the ideal direction of travel. To determine the robot's expected position without external forces, the expected displacement is obtained by extending the ideal direction of travel from the actual position at the start time. The actual displacement vector represents the actual spatial displacement path of the robot under real fluid disturbances, while the ideal displacement vector represents the theoretical spatial displacement path of the robot without external force interference. The smaller the heading coincidence coefficient determined based on the cosine similarity of the two, the more the robot's actual displacement direction deviates from the expected trajectory, indicating a more severe heading deflection caused by external environmental forces.

[0038] In this embodiment of the invention, the cosine similarity is normalized by taking half of the sum of the cosine similarity between the actual displacement vector and the ideal displacement vector and the constant 1.

[0039] In this embodiment, the robot's ideal speed can be obtained by querying a pre-built mapping table. The mapping table is constructed as follows: In an ideal still water environment, propulsion state data (such as thruster output power or motor speed) characterizing the robot's power output intensity is divided into multiple levels and adjusted sequentially. After the robot stabilizes at a uniform linear motion in each level, the current moving speed is synchronously measured using a speed measuring device. The propulsion state data and moving speed at each level are mapped one-to-one to obtain the mapping table. In actual cleaning operations, the robot's propulsion state data for all moments within a preset analysis period are acquired in real time, and the mapping table is queried. The average value of the corresponding ideal speed is calculated as the ideal speed for the preset analysis period. If the propulsion state data falls between two calibrated levels, it is calculated using a linear interpolation algorithm.

[0040] Step S240: Based on the net deviation trend ratio and the heading coincidence coefficient, obtain the robot's comprehensive deviation index at the current moment; select the robot whose comprehensive deviation index is greater than the preset deviation threshold as the robot to be reset at the current moment.

[0041] It should be noted that the net deviation trend ratio and heading consistency coefficient analyze the robot's deviation from its intended path from two perspectives: spatial distance and direction of motion, respectively. A larger net deviation trend ratio indicates a more severe lateral deviation; a smaller heading consistency coefficient reflects a greater deviation of the robot's actual displacement direction from the expected trajectory, indicating a more severe heading deflection caused by external forces. Therefore, the net deviation trend ratio is positively correlated with the comprehensive deviation index, while the heading consistency coefficient is negatively correlated with the comprehensive deviation index. A larger comprehensive deviation index indicates a more severe deviation from the predetermined path due to external forces, and a greater likelihood that the robot has completely left the effective cleaning area.

[0042] In this embodiment of the invention, the product of the robot's net deviation trend ratio at the current moment and the reciprocal of the heading consistency coefficient is used as the comprehensive deviation index. It should be noted that when the cumulative return to track or the heading consistency coefficient is zero, the comprehensive deviation index is set to a preset upper limit cutoff value. This preset upper limit cutoff value is set as the product of a preset deviation threshold and a safety amplification factor, such as 10.

[0043] In this embodiment, the preset deviation threshold can be obtained by calibrating historical test data of the robot under extreme permissible deviation conditions. The method is as follows: extract the robot's operating data when the environmental external force interference in the test environment reaches the tolerance limit, calculate the net deviation trend ratio within a preset analysis period under this extreme condition, and simultaneously obtain the heading coincidence coefficient between the actual displacement vector and the ideal displacement vector under this condition; fuse the extreme net deviation trend ratio and the extreme heading coincidence coefficient, and the result can be used as the preset deviation threshold.

[0044] Preferably, in some possible implementations of the embodiments of the present invention, the method for obtaining the detour candidate area includes: calculating the difference vector between the actual displacement vector and the ideal displacement vector of the robot to be reset at the current moment during a preset analysis period; multiplying the difference vector by the reciprocal of the number of moments in the preset analysis period to obtain the fluid disturbance vector of the working area where the robot to be reset is located at the current moment; for each adjacent working area of ​​the working area where the robot to be reset is located at the current moment, the vector pointing from the center point of the adjacent working area to the point closest to the center point on the predetermined track of the robot to be reset is taken as the track traction vector of the adjacent working area; if the angle between the working area where the robot to be reset is located at the current moment and the fluid disturbance vector of each of its adjacent working areas is an obtuse angle, and the dot product of the fluid disturbance vector and the track traction vector of the adjacent working area is greater than zero, then the adjacent working area is taken as the detour candidate area of ​​the robot to be reset at the current moment.

[0045] It should be noted that the difference between the actual displacement vector and the ideal displacement vector reflects the positional offset of the robot caused by external water flow impact within the preset analysis period; the fluid disturbance vector characterizes the average physical direction and intensity of the external force of the water flow within the working area. Specifically, the fluid disturbance vector of adjacent working areas is obtained by acquiring real-time operating data of other robots currently operating within that adjacent working area and using the same calculation method as the robot to be reset; if there are no other robots operating in that adjacent working area at the current moment, the historical fluid disturbance vector of that area from the previous moment is used; if there is no historical record, the average fluid disturbance vector of the remaining robots around the robot to be reset at the current moment is used instead.

[0046] The orbital traction vector represents the ideal orientation for the robot to return to its predetermined orbit from the adjacent work area using the shortest spatial geometric path. Considering that the robot's current deviation is mainly due to strong and unfavorable water flow interference, the planned reset path needs to find areas with backflow or opposite-direction water flow to escape the current unfavorable flow field and reduce reset energy consumption by utilizing the thrust of the water flow itself. Therefore, if the angle between the fluid interference vector of the current work area where the robot is located and each of its adjacent work areas is obtuse, it indicates that there is an opposite-direction water flow in the adjacent work area, which is beneficial for the robot to escape. If the dot product of the fluid interference vector and the orbital traction vector of the adjacent work area is greater than zero, that is, the angle between the two vectors is acute, it means that the direction of the water flow in the adjacent work area is roughly consistent with the ideal orientation of the return orbit, which can help the robot approach the predetermined orbit. Adjacent work areas that meet both of the above conditions are selected as alternative detour areas for the robot to be reset at the current moment.

[0047] Preferably, in some possible implementations of the embodiments of the present invention, the method for obtaining the optimal detour area is described in [reference needed]. Figure 3 The diagram illustrates a flowchart of a method for obtaining an optimal detour area according to an embodiment of the present invention, the method comprising: Step S410: For each alternative detour area of ​​the robot to be reset at the current moment, the vector from the actual position of the robot to be reset at the current moment to the point on the boundary of the alternative detour area that is closest to the actual position is used as the shortest boundary crossing vector of the alternative detour area.

[0048] It should be noted that, considering that prolonged exposure of the robot in a harsh flow field will increase energy consumption and the risk of secondary deviation, in order to minimize the upfront costs before entering the downstream water area, it is necessary to plan the shortest route to escape the current unfavorable flow field and obtain the shortest boundary crossing vector, which represents the optimal physical orientation and path length of the robot to cross the current boundary and enter the downstream bypass alternative area with the minimum distance and energy consumption.

[0049] Step S420: The ratio of the off-track distance of the robot to be reset at the current moment to the magnitude of the shortest out-of-bounds vector of each alternative detour area is used as the detour cost-benefit ratio of that alternative detour area.

[0050] It should be noted that the off-track distance represents the total physical cost that the robot needs to overcome to return to its straight-line trajectory, and the magnitude of the shortest out-of-bounds vector represents the initial travel distance required to enter the downstream region; the ratio of the two, i.e., the detour cost-benefit ratio, quantifies the geometric economy of the detour operation. The larger the detour cost-benefit ratio, the more the robot can avoid the high-resistance forced return-to-track distance by only incurring a very short detour distance, and the higher the geometric benefit of the detour decision.

[0051] Step S430: Calculate the fluid disturbance vector of the working area where the robot to be reset is located at the current moment, and the angles between it and the fluid disturbance vector of each alternative bypass area and the shortest cross-boundary vector at the current moment, respectively, and record them as the flow field deflection angle and the cross-boundary heading deflection angle; perform a ratio operation on the negative of the cosine value of the flow field deflection angle and the difference between the constant 1 and the cosine value of the cross-boundary heading deflection angle to obtain the flow field compliance of the robot to be reset in each alternative bypass area at the current moment.

[0052] It should be noted that the flow field deflection angle reflects the difference in water flow direction between the current working area of ​​the robot to be reset and the alternative detour area; the cross-boundary heading angle reflects the angle between the robot's direction of travel when entering the alternative detour area and the current water flow direction. The larger the difference in water flow direction between two adjacent areas, i.e., the larger the flow field deflection angle, the more significant the downstream thrust the robot gains after entering the alternative detour area. At the same time, when the robot crosses the boundary to enter the alternative detour area, the more its direction of travel is in line with the current water flow, i.e., the cross-boundary heading angle approaches 0 degrees, the lower the water flow resistance and energy consumption when crossing the area. To comprehensively measure the overall effect of the above water flow leverage, a ratio calculation is constructed: in the numerator, the negative value of the cosine of the flow field deflection angle is taken, converting the negative number representing the opposite direction into a positive value representing the leverage gain; in the denominator, when the robot enters against the current, the difference between the constant 1 and the cosine of the cross-boundary heading angle becomes larger, thereby reducing the ratio result and thus reducing the high energy consumption route against the current. The greater the flow field compliance, the greater the water flow thrust provided by the alternative detour area, and the smaller the resistance during the robot's entry into the area, resulting in a better overall leverage effect.

[0053] Step S440: Based on the comprehensive deviation index of the robot to be reset at the current moment, as well as the detour cost-benefit ratio and flow field compliance of each detour candidate area at that moment, obtain the comprehensive recommendation degree of the corresponding detour candidate area; from all the detour candidate areas of the robot to be reset at the current moment, select the detour candidate area with the largest comprehensive recommendation degree that is greater than the preset recommendation threshold as the optimal detour area at that moment.

[0054] In one specific implementation of this invention, the comprehensive recommendation score is expressed by the formula:

[0055] In the formula, Y is the comprehensive recommendation degree of each alternative detour area of ​​the robot to be reset at the current time; P is the comprehensive deviation index of the robot to be reset at the current time; and d is the deviation distance of the robot to be reset at the current time. Let be the shortest out-of-bounds vector for each alternative detour region of the robot to be reset at the current moment; The flow field deflection angle of the robot to be reset in each alternative detour region at the current moment; Let be the out-of-bounds heading angle of the robot to be reset in each alternative detour area at the current moment; cos is the cosine function. It is the modulo function; The detour cost-benefit ratio of the robot to be reset to each alternative detour area at the current moment; The flow field compliance of the robot to be reset in each alternative detour region at the current moment.

[0056] It should be noted that the comprehensive deviation index reflects the urgency of getting out of trouble: the larger the value, the more severe the robot's deviation, the extremely high energy cost of straight-line resetting, and the stronger the urgency of taking a detour; the detour cost-benefit ratio reflects the economy of detour: the larger the value, the shorter the pre-detour distance to enter the detour candidate area, and the higher the geometric benefit of avoiding long-distance backflow; the flow field compliance characterizes the water flow leverage effect: the larger the value, the stronger the downstream thrust provided by the detour candidate area and the minimal resistance to crossing boundaries. Therefore, the larger the above three values, the more advantageous it is in terms of urgent need, distance cost, and energy consumption reduction when detouring through the detour candidate area, and thus all are positively correlated with the comprehensive recommendation degree. The higher the comprehensive recommendation degree, the more energy-efficient and effective it is to return via the detour candidate area than to return in a straight line. To ensure that the detour plan brings positive benefits and accurately pinpoint the best leverage route, the area with the largest comprehensive recommendation degree greater than the preset recommendation threshold is selected from all detour candidate areas as the optimal detour area.

[0057] In this embodiment, homing sample data of the robot in historical testing or fluid dynamics simulation environments are extracted. The homing sample data is a set of data on the actual propulsion energy consumption or homing time consumed by the robot to successfully reset after deviating from its course using a straight-line homing strategy and a cross-area detour homing strategy, respectively. From the homing sample data, critical operating condition samples in which the straight-line homing cost and the detour homing cost are equal under the same comprehensive deviation index are selected. The comprehensive recommendation degree calculated based on the comprehensive deviation index, detour cost-benefit ratio, and flow field compliance degree corresponding to the critical operating condition sample is used as a preset recommendation threshold.

[0058] It is important to note that when the shortest outbound vector of the alternative detour area is zero, or the outbound heading angle is 0 degrees, it indicates that the robot is already at the area boundary or the entry heading is completely downstream. In this case, the prior physical distance cost and upstream resistance when entering the alternative detour area are practically zero, representing an excellent leverage-based repositioning route. To avoid the denominator being zero in the formula calculation, which would render the fraction meaningless, the overall recommendation degree of this alternative detour area is directly set as the limit recommendation degree. The limit recommendation degree is set as the product of a preset recommendation threshold and a preset multiplier factor, such as 5, ensuring that the limit recommendation degree is greater than the preset recommendation threshold and the overall recommendation degree obtained from all conventional calculations.

[0059] Preferably, in some possible implementations of the embodiments of the present invention, the method for obtaining the homing path includes: recording the point on the predetermined track of the robot to be reset that is closest to the actual position at the start time of the preset analysis period at the current time as the operation reset point; taking the actual position of the robot to be reset at the current time as the starting point and the operation reset point as the ending point; recording the point on the boundary of the optimal bypass area of ​​the robot to be reset at the current time that is closest to the starting point as the area entry node; using an ant colony algorithm to plan a path from the starting point through the area entry node to the ending point as the homing path; wherein, the heuristic factor in the ant colony algorithm is set as the reciprocal of the upstream resistance cost; when performing path search between adjacent grid nodes using the ant colony algorithm, calculating the vector from the currently traversed grid node to the adjacent grid node to be transferred, and using the angle between this vector and the fluid interference vector of the optimal bypass area as the transfer flow field deflection angle; performing a negative correlation mapping on the cosine value of the transfer flow field deflection angle, and setting the product of the mapping result and the distance between the corresponding adjacent grid nodes as the upstream resistance cost between adjacent grid nodes.

[0060] It should be noted that, to ensure a seamless transition to the original cleaning task after the robot resets, the point on the robot's predetermined track closest to its actual position at the start of the current preset analysis period is recorded as the reset point, representing the reasonable working position the robot should have been in before a substantial deviation from its course. Considering that conventional shortest geometric paths often face significant fluid resistance underwater, resulting in high energy consumption and susceptibility to secondary deviation, fluid dynamics factors must be incorporated into pathfinding. Therefore, an ant colony algorithm is used to plan the homing path from the region entry node to the reset point; and the heuristic factor of the ant colony algorithm is set as the reciprocal of the upstream resistance cost. To accurately quantify the resistance or assistance encountered by the robot in a single step, when the ant colony algorithm searches adjacent grids, the angle between the vector from the current node to the node to be transferred and the fluid disturbance vector of the optimal bypass area is used as the transfer flow field deflection angle, reflecting the robot's compliance with the water flow field during local movement. The shift angle of the transferred flow field tends to 180 degrees, indicating that the robot's travel direction is in a high-energy-consuming headwind and upstream state. To effectively penalize such upstream routes during path search, the cosine value of this shift angle is negatively correlated with the upstream resistance cost, which characterizes the magnitude of the water flow resistance to be overcome. In this embodiment, the upstream resistance cost is expressed by the formula:

[0061] In the formula, C is the cost of the frontal resistance between each pair of adjacent grid nodes; L is the Euclidean distance between each pair of adjacent grid nodes. θ is the deflection angle of the transferred flow field between each pair of adjacent grid nodes; cos is the cosine function; is the preset environmental adjustment coefficient; max is the maximum value function. The preset environmental adjustment coefficient represents the degree of water flow interference to the robot, set as the ratio of the magnitude of the fluid interference vector in the optimal bypass area to the robot's maximum interference-free movement speed. A higher resistance cost indicates that the robot is in a headwind / upstream state when traversing the adjacent grid path, requiring it to overcome a higher water flow thrust. Consequently, its heuristic factor is smaller, allowing the ant colony algorithm to reject such high-resistance paths and globally optimize a return path with the highest energy efficiency and maximum leverage with the current.

[0062] If the overall recommendation score of all alternative detour areas for the robot to be reset at the current moment is less than or equal to the preset recommendation threshold, then the straight line from the actual position of the robot to be reset at the current moment to the reset point will be taken as the homing path.

[0063] This invention is now complete.

Claims

1. A method for multi-robot cooperative path planning for ocean-going container ship cleaning, characterized in that, The method includes: On a two-dimensional grid map of the surface of an ocean-going container ship, work areas and predetermined tracks are assigned to multiple robots used for cleaning the ship; the actual position of each robot during the cleaning process is obtained in real time. Based on the degree of deviation of the actual position of each robot from its predetermined trajectory and the degree of heading deviation within the preset analysis period at the current moment, the robot to be reset at the current moment is selected. Based on the difference in the direction of external environmental forces between the working area where the robot to be reset is located at the current moment and its adjacent working areas, as well as the traction trend of the external environmental forces in the adjacent working areas on the predetermined track, the alternative detour area of ​​the robot to be reset at the current moment is selected. Based on the distance cost of the robot to be reset entering each alternative detour area, and the degree of conformity of the environmental external forces in the corresponding alternative detour area to the cutting-in heading, the optimal detour area is selected from the alternative detour areas of the robot to be reset at the current moment. Plan the return path of the robot to be reset through the optimal detour area to its predetermined track, and control the robot to be reset to move along the return path.

2. The multi-robot collaborative path planning method for cleaning ocean-going container ships according to claim 1, characterized in that, The selection of the robot to be reset at the current moment includes: The shortest distance from the actual position of each robot to its predetermined trajectory at each moment is recorded as the off-track distance. For each robot, if the deviation distance of the robot at each time within the preset analysis period at the current time is greater than the preset allowable deviation threshold, and the deviation distance at the time before the start time of the preset analysis period is less than or equal to the preset allowable deviation threshold, then the first-order difference sequence of the deviation distance of the robot at each time within the preset analysis period is obtained. The sum of the positive elements and the sum of the absolute values ​​of the negative elements in the first-order difference sequence are respectively counted and recorded as the off-track cumulative amount and the return-track cumulative amount; the ratio of the off-track cumulative amount to the return-track cumulative amount is taken as the net deviation trend ratio of the robot at the current moment. Based on the difference between the robot's actual displacement direction and the ideal displacement direction during the preset analysis period at the current moment, the heading coincidence coefficient of the robot at the current moment is obtained; Based on the net deviation trend ratio and the heading coincidence coefficient, the robot's comprehensive deviation index at the current moment is obtained; the robot whose comprehensive deviation index is greater than the preset deviation threshold is selected as the robot to be reset at the current moment.

3. The multi-robot collaborative path planning method for cleaning ocean-going container ships according to claim 2, characterized in that, The acquisition of the robot's heading alignment coefficient at the current moment includes: The product of the robot's ideal speed during the preset analysis period and the duration of the preset analysis period is calculated as the expected displacement. The tangent direction of the point on the robot's predetermined track that is closest to its actual position at the start of the preset analysis period is taken as the ideal direction of travel. Starting from the actual position at the start of the period, the expected displacement is extended along the ideal direction of travel to obtain the robot's expected position at the current moment. The actual displacement vector and the ideal displacement vector are respectively denoted as the actual displacement vector and the ideal displacement vector, respectively, pointing from the robot's actual position at the start of the preset analysis period at the current moment to the actual position and the expected position at the current moment. The cosine similarity between the actual displacement vector and the ideal displacement vector is normalized to obtain the heading coincidence coefficient of the robot at the current moment.

4. The multi-robot collaborative path planning method for cleaning ocean-going container ships according to claim 3, characterized in that, The selection of alternative detour areas for the robot to be reset at the current moment includes: Calculate the difference vector between the actual displacement vector and the ideal displacement vector of the robot to be reset in the current time within a preset analysis period. Multiply the difference vector by the reciprocal of the number of time points in the preset analysis period to obtain the fluid disturbance vector of the working area where the robot to be reset is located at the current time. For each adjacent working area of ​​the working area where the robot to be reset is located at the current moment, the vector from the center point of the adjacent working area to the point on the predetermined track of the robot to be reset that is closest to the center point is used as the track traction vector of the adjacent working area. If the angle between the fluid disturbance vector of the working area where the robot to be reset is located at the current moment and the fluid disturbance vector of each of its adjacent working areas is an obtuse angle, and the dot product of the fluid disturbance vector of the adjacent working area and the track traction vector is greater than zero, then the adjacent working area is taken as the alternative detour area for the robot to be reset at the current moment.

5. The multi-robot collaborative path planning method for cleaning ocean-going container ships according to claim 4, characterized in that, The process of selecting the optimal detour area from the current detour candidate area for the robot to be reset includes: For each alternative detour region of the robot to be reset at the current moment, the vector pointing from the actual position of the robot to be reset at the current moment to the point on the boundary of the alternative detour region that is closest to the actual position is used as the shortest boundary crossing vector of the alternative detour region. The ratio of the off-track distance of the robot to be reset at the current moment to the magnitude of the shortest out-of-bounds vector of each alternative detour region is used as the detour cost-benefit ratio of that alternative detour region. Calculate the fluid disturbance vector of the working area where the robot to be reset is located at the current moment, and the angles between the fluid disturbance vector of each alternative detour area and the shortest cross-boundary vector at that moment, respectively, and record them as the flow field deflection angle and the cross-boundary heading deflection angle. The flow field compliance of the robot to be reset in each alternative detour area at the current moment is obtained by comparing the negative of the cosine of the flow field deflection angle with the difference between the constant 1 and the cosine of the out-of-bounds heading angle. Based on the overall deviation index of the robot to be reset at the current moment, as well as the detour cost-benefit ratio and flow field compliance of each detour candidate area at that moment, the overall recommendation degree of the corresponding detour candidate area is obtained; From all the alternative routes available to the robot to be reset at the current moment, select the alternative route with the highest overall recommendation score that is greater than the preset recommendation threshold as the optimal alternative route for that moment.

6. The multi-robot collaborative path planning method for cleaning ocean-going container ships according to claim 5, characterized in that, The planned return path for the robot to be reset, via the optimal detour area to its predetermined orbit, includes: The point on the predetermined track of the robot to be reset that is closest to the actual position of the start time of the preset analysis period at the current moment is recorded as the reset point. Starting from the actual position of the robot to be reset at the current moment, and ending at the reset point; the point on the boundary of the optimal detour area of ​​the robot to be reset at the current moment that is closest to the starting point is recorded as the region entry node; the ant colony algorithm is used to plan the path from the starting point through the region entry node to the end point, which is used as the return path; Specifically, the heuristic factor in the ant colony algorithm is set to the reciprocal of the oncoming resistance cost; when performing path search between adjacent grid nodes using the ant colony algorithm, the vector pointing from the currently traversed grid node to the adjacent grid node to be transferred is calculated, and the angle between this vector and the fluid interference vector of the optimal bypass area is taken as the transfer flow field deflection angle; the cosine value of the transfer flow field deflection angle is negatively correlated and mapped, and the product of the mapping result and the distance between the corresponding adjacent grid nodes is set as the oncoming resistance cost between adjacent grid nodes.

7. The multi-robot collaborative path planning method for cleaning ocean-going container ships according to claim 6, characterized in that, If the overall recommendation score of all alternative detour areas for the robot to be reset at the current moment is less than or equal to the preset recommendation threshold, then the straight line from the actual position of the robot to be reset at the current moment to the operation reset point will be taken as the homing path.

8. The multi-robot collaborative path planning method for cleaning ocean-going container ships according to claim 5, characterized in that, When the shortest outbound vector is zero, or when the cosine of the outbound heading angle is equal to 1, the comprehensive recommendation degree is assigned as the limit recommendation degree; wherein, the limit recommendation degree is greater than the preset recommendation threshold.

9. The multi-robot collaborative path planning method for cleaning ocean-going container ships according to claim 2, characterized in that, The net deviation trend ratio is positively correlated with the comprehensive deviation index, and the heading coincidence coefficient is negatively correlated with the comprehensive deviation index.

10. The multi-robot collaborative path planning method for cleaning ocean-going container ships according to claim 1, characterized in that, The current time is at the end of its preset analysis period.

Images