Dynamic water surface environment path planning method, device and system and storage medium

Abstract

The invention discloses a dynamic water surface environment path planning method, device and system, and a storage medium. The method comprises the following steps: S1, constructing a fusion energy consumption estimation model based on the cubic law of fluid mechanics and direction penalty; and S2, according to the fused energy consumption estimation model, obtaining a dynamic target re-evaluation and switching strategy in combination with a hysteresis threshold and a minimum committed distance so as to realize energy consumption sensing adaptive path planning of the water surface cleaning robot. By the adoption of the technical scheme, the technical problems that when the water surface cleaning robot executes garbage searching and collecting tasks in the complex dynamic marine environment, energy efficiency is low, path planning adaptability is insufficient, and response to the dynamic environment is slow are solved.

Description

Technical Field

[0001] This invention belongs to the field of water surface cleaning robot technology, specifically relating to a dynamic water surface environment path planning method, device, system, and storage medium. Background Technology

[0002] In the field of path planning for marine surface cleaning robots, various algorithms have been proposed and applied to optimize the robot's trajectory and improve waste collection efficiency. Traditional path planning methods, when applied to dynamic marine scenarios, often simplify the problem to path search in static or quasi-static environments, failing to fully consider the nonlinear effects of time-varying ocean currents on robot propulsion power, real-time obstacles and waste targets, and energy constraints under limited endurance. This leads to situations where paths planned by existing methods may experience a surge in energy consumption due to counter-current navigation, ineffective navigation due to target drift or the appearance of new targets, or getting trapped in locally suboptimal paths due to sudden environmental changes. Ultimately, this results in high overall task energy consumption, low waste collection efficiency, and insufficient system robustness. Summary of the Invention

[0003] To address the problems existing in the prior art, this invention provides a dynamic water surface environment path planning method, device, system, and storage medium, which solves the technical problems of low energy efficiency, insufficient path planning adaptability, and slow response to dynamic environments when water surface cleaning robots perform garbage search and collection tasks in complex and dynamic marine environments.

[0004] To achieve the above objectives, the present invention provides the following solution: A dynamic water surface environment path planning method, comprising: Step S1: Construct a fusion energy consumption estimation model based on the cubic law of fluid dynamics and directional penalty; Step S2: Based on the fusion energy consumption estimation model, a dynamic target re-evaluation and switching strategy combining hysteresis threshold and minimum committed distance is obtained to realize energy consumption perception adaptive path planning for the water surface cleaning robot.

[0005] Preferably, in step S1, after each move, a loss function is constructed based on the error between the measured energy consumption per unit distance and the energy consumption per unit distance predicted by the current weights and the fused energy consumption estimation model, and the weights are updated using the stochastic gradient descent method.

[0006] Preferably, in step S2... Hysteresis switching condition: only when a new candidate target The estimated total cost meets Target switching is only triggered at that time; Minimum commitment distance constraint: in the direction of a selected target The distance traveled did not exceed the preset minimum commitment distance. Previously, switching was prohibited even if a better target became available.

[0007] Preferably, in step S2, at each decision step, the robot selects the grid that minimizes the cost per step within its limited neighborhood based solely on a unified heuristic cost function as the next moving position.

[0008] The present invention also provides a dynamic water surface environment path planning device, comprising: The first processing module is used to construct a fusion energy consumption estimation model based on the cubic law of fluid dynamics and direction penalty; The second processing module is used to obtain a dynamic target re-evaluation and switching strategy that combines hysteresis threshold and minimum committed distance based on the fusion energy consumption estimation model, so as to realize the energy consumption perception adaptive path planning of the water surface cleaning robot.

[0009] Preferably, the first processing module is used to construct a loss function based on the error between the measured energy consumption per unit distance and the energy consumption per unit distance predicted by the current weights and the fused energy consumption estimation model after each movement, and to update the weights using the stochastic gradient descent method.

[0010] As a preferred option, the second process is performed as follows: Hysteresis switching condition: only when a new candidate target The estimated total cost meets Target switching is only triggered at that time; Minimum commitment distance constraint: in the direction of a selected target The distance traveled did not exceed the preset minimum commitment distance. Previously, switching was prohibited even if a better target became available.

[0011] Preferably, the second processing module is used so that, at each decision step, the robot selects the grid that minimizes the single-step cost within its limited neighborhood as the next move position, based solely on a unified heuristic cost function.

[0012] The present invention also provides a dynamic water surface environment path planning system, comprising: a memory and a processor, wherein the memory stores a computer program executed by the processor, and the computer program executes a dynamic water surface environment path planning method when run by the processor.

[0013] The present invention also provides a storage medium storing a computer program, which executes a dynamic water surface environment path planning method during runtime.

[0014] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention combines a refined energy consumption model based on the cubic law of fluid dynamics with an online learning mechanism of stochastic gradient descent. This allows the algorithm to continuously learn the relationship between environmental characteristics and energy consumption during task execution, and dynamically adjust the weights of distance and energy consumption factors in path cost, thereby achieving adaptive optimization in dynamic ocean current environments. Furthermore, this invention abandons the traditional "target-locked" static planning approach, innovatively introducing a dynamic target re-evaluation mechanism based on opportunity cost analysis and a step-by-step walking strategy. This enables the robot to consider not only the immediate movement cost in each movement decision, but also the optimality of all candidate targets based on updated environmental information, and to switch targets when necessary, effectively avoiding path failure or energy waste caused by environmental changes. This framework, while ensuring the real-time performance of the algorithm, significantly improves the energy efficiency and task completion performance of surface cleaning robots in complex, non-stationary marine environments, providing a practical technical solution for long-endurance, highly autonomous marine robot operations. Attached Figure Description

[0015] To more clearly illustrate the technical solution of the present invention, the drawings used in the embodiments are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a flowchart of the dynamic water surface environment path planning method according to an embodiment of the present invention; Figure 2 The trend of total energy consumption of various path planning algorithms under different robot moving speeds; Figure 3 A comparison of the impact of robot movement speed on the amount of garbage collected; Figure 4 The total energy consumption performance of each path planning algorithm under different obstacle coverage rates; Figure 5 Analysis of the impact of obstacle density on waste collection efficiency; Figure 6 The impact of ocean current velocity variations on the energy consumption performance of path planning; Figure 7 A comparison of the amount of garbage collected under different ocean current velocities. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0019] Example 1 like Figure 1 As shown, this invention provides a dynamic water surface environment path planning method, named EcoNav. This method constructs a unified decision-making framework integrating environmental perception, energy consumption modeling, and online learning, enabling robots to achieve energy-efficient path planning in complex and dynamic marine environments; specifically, it includes: Step S1: Construct a fusion energy consumption estimation model based on the cubic law of fluid dynamics and directional penalty; Step S2: Based on the fusion energy consumption estimation model, a dynamic target re-evaluation and switching strategy combining hysteresis threshold and minimum committed distance is obtained to realize energy consumption perception adaptive path planning for the water surface cleaning robot.

[0020] As one embodiment of the present invention, in step S1, a heuristic cost function is designed and dynamically optimized to quantitatively evaluate the cost from the current position. Heading to candidate waste targets The total cost. This function summarizes the decision factors into three parts: historical cost, estimated distance cost, and estimated energy cost, and its mathematical expression is as follows: ; in, This indicates that the robot has moved from the starting point to its current position. The accumulated actual path cost. To estimate the future travel distance, Euclidean distance is used for calculation: ; This is the estimated energy cost, calculated by comprehensively considering the robot's base propulsion power and the nonlinear effects of ocean currents. For accurate modeling, two key sub-energy rates are first defined: the base energy rate... and ocean current corrected energy consumption rate The base energy rate is determined by the cruise power. With set speed Decide: ; The ocean current-corrected energy rate further considers changes in effective speed, among which... This represents the component of the current ocean current in the direction of movement (positive for countercurrent). To prevent small constants from being divided by zero, The fluid drag index (usually taken as 3): ; Combining the above two points, and introducing a directional consistency penalty to prevent the robot from aimlessly wandering in non-critical areas, the final energy consumption cost estimate is expressed as: ; In the formula The current movement direction vector The angle between the vector and the global task direction vector (such as the geometric center pointing to all uncollected trash). This represents the directional penalty coefficient.

[0021] In the cost function and These are two key dynamic weight parameters that control the importance of distance cost and energy cost in the overall decision-making process, respectively, and satisfy normalization constraints. To enable the algorithm to adapt to non-stationary marine environments, this invention employs an online learning mechanism to adjust these two weights in real time. This mechanism is based on a stochastic gradient descent framework: in the... After the first movement, the energy consumption per unit distance was measured. Compared with the predicted value The loss function is constructed from the error between them. And update the weights in the opposite direction of the gradient: ; in The learning rate is used. Normalization is performed immediately after the weights are updated to ensure that the constraints always hold.

[0022] Furthermore, the fusion energy consumption estimation model based on the cubic law of fluid dynamics and direction penalty does not simply use a fixed power consumption value or a linear model. Instead, it embeds the cubic law of fluid dynamics (drag is proportional to the cube of velocity) into the energy consumption prediction, specifically through the formula... To calculate the ocean current corrected energy rate, where The typical value is 3. Additionally, the model incorporates a penalty term related to the global task direction. ,in It is the angle between the current direction of movement and the overall task direction (such as the direction of the geometric center of uncollected trash). This design aims to prevent the robot from making ineffective explorations in directions away from the main task area, thereby balancing local propulsion efficiency and global task orientation in terms of energy consumption estimation.

[0023] Furthermore, during task execution, the distance cost weight in the cost function is dynamically adjusted in real time and continuously. Weighted by energy consumption cost The specific implementation method is as follows: after each movement, based on the measured energy consumption per unit distance... Compared with the energy consumption per unit distance predicted by the current weights and the model The error between them is used to construct a loss function, and the weights are updated using stochastic gradient descent, with the update rule being: This mechanism enables the algorithm to autonomously learn and adapt to the dynamic characteristics of the current marine environment (such as changes in ocean current intensity) without relying on preset fixed weights or offline environmental calibration, thereby significantly improving the robustness and long-term energy efficiency of the planning system in non-stationary environments.

[0024] As one embodiment of the present invention, in step S2, based on the above cost model, EcoNav adopts an execution strategy combining "stepwise decision-making and dynamic re-evaluation". The robot does not plan the entire path all at once during the task, but rather re-evaluates all candidate targets at each discrete time step based on updated environmental information and weights. For the current position... The algorithm will consider its eight neighboring grids. And calculate the one-step heuristic cost of moving to each neighborhood: ; in, This represents the actual distance traveled in a single step. The robot chooses to... The smallest neighborhood is used as the next moving target, thus achieving gradual advancement towards local optima. Simultaneously, the algorithm continuously monitors environmental changes (such as drastic ocean current changes or the appearance of new debris) or its own state (such as battery level falling below a threshold). Once a re-evaluation condition is met, such as the existence of another candidate target, the algorithm will proceed. satisfy (in (The preset switching threshold, such as 0.9), and the current target has already been switched to. The distance traveled exceeded the minimum committed distance. If so, the current path will be immediately interrupted. The system updates to the new current goal and restarts the stepwise decision-making loop based on the new goal. This mechanism ensures that the robot maintains flexibility and near-optimal decision-making at the global level.

[0025] When the robot is within the capture radius of the target debris, collection is considered successful. The target is removed from the pending list, and a new round of global target selection is immediately triggered. The entire planning process is executed cyclically until all debris is collected or the robot's battery level drops to a safe return threshold. At this point, the algorithm will force the robot to switch to the nearest charging station and enter an energy-saving return mode. Through the synergistic effect of the above technical solutions, EcoNav can significantly improve the energy efficiency and mission reliability of surface cleaning robots in dynamic marine environments while ensuring real-time computing efficiency.

[0026] Furthermore, in step S2, this strategy abandons the traditional models of "one-time planning, fixed execution" or "greedy instant switching," and adopts a stable dynamic re-evaluation mechanism. Its core rules include two points: a) Hysteresis handover condition: only when the new candidate target The estimated total cost meets (in Target switching is only triggered when the threshold is less than 1 (e.g., 0.85 or 0.9). This avoids frequent and volatile target switching caused by minor fluctuations in cost estimation.

[0027] b) Minimum commitment distance constraint: When moving towards a selected target The distance traveled did not exceed the preset minimum commitment distance. Previously, switching was prohibited even if a better objective emerged. This constraint-enforcing algorithm maintains a certain "inertia" for the current decision, ensuring the consistency of actions and robustness to instantaneous environmental noise.

[0028] The path generation method of this invention does not calculate a complete trajectory all at once, but rather employs a hierarchical, cyclical architecture of "global target selection -> local single-step decision -> continuous feedback and reassessment". At each decision step, the robot selects a path within its limited neighborhood (e.g., an 8-neighborhood) based solely on a unified heuristic cost function, maximizing the single-step cost. The smallest grid cell is used as the next move position. Single-step cost. Inheriting the core elements of the global cost function, this architecture ensures that local movements always serve the global optimization objective. It decomposes the complex global path optimization problem into a series of locally optimal decisions guided by global information, significantly reducing computational complexity while guaranteeing real-time response to dynamic environments, thus achieving a balance between algorithmic efficiency and planning quality.

[0029] The core objective of this invention is not simply to find the shortest path or achieve area coverage, but rather to establish an integrated planning framework in a dynamic and uncertain marine environment that can sense environmental changes online, predict energy consumption impacts in real time, and dynamically adjust decision weights. This invention aims to enable surface cleaning robots to possess the following capabilities: 1. Energy consumption perception and dynamic modeling: The nonlinear influence of ocean currents on propulsion power (based on the cubic law of fluid dynamics) is directly embedded into the decision cost function, so that the path selection can fundamentally avoid high energy consumption areas.

[0030] 2. Online learning and adaptation: Through the stochastic gradient descent mechanism, the algorithm can automatically adjust the weight of the distance cost and energy cost based on the actual energy consumption feedback during task execution, thereby adapting to the needs of different flow rates and different task stages.

[0031] 3. Dynamic replanning and target switching: Abandoning the static planning mode of "locking in the target", a real-time target re-evaluation mechanism based on opportunity cost analysis is introduced, which enables the robot to switch to a better global target in a timely manner when the environment (ocean current, new garbage) or its own state (battery power) changes, thus avoiding ineffective or high-cost navigation.

[0032] 4. Incremental Optimization and Real-Time Guarantee: Employing a "step-by-step" strategy, the complex global optimization problem is decomposed into a series of locally optimal decisions. While ensuring that each move is immediately feasible, a heuristic cost function guides global convergence. This method has approximately linear computational complexity, meeting the real-time planning requirements of large-scale dynamic environments.

[0033] To make the simulation experiment more realistic, the embodiments of the present invention use total power loss. The calculation formula is shown below: ; in, For Totaltimesteps; Distance traveled at step (km); Oceancurrentvelocityatstep (km / h); for Baseenergyforsensors / computation(constant).

[0034] Equation (2) defines the average amount of marine debris collected. ,in Indicates the first The amount of marine debris collected in this experiment. Additionally, if the robot is less than 1 meter away from the marine debris, the collection is considered successful; otherwise, the collection is considered unsuccessful. This represents the total number of runs, i.e.

[0035] ;

[0036] The impact of robot movement speed This experiment analyzes the impact of robot movement speed on path planning efficiency. Intuitively, a higher movement speed may shorten the total task time, but it may also lead to a sharp increase in propulsion energy consumption due to the cubic law, resulting in increased energy consumption per unit distance. To quantify this trade-off, in this embodiment of the invention, with other parameters fixed, the robot speed was gradually increased from 2 m / s to 4 m / s, with the movement speed of the robot in each experimental group increasing by 0.2 m / s. The average ocean current velocity was fixed at 0.1 m / s, and the obstacle coverage rate was set at 20%. 35 experiments were conducted under each condition.

[0037] At each speed setting, this embodiment of the invention records the total energy consumption, final path length, and amount of garbage successfully collected for EcoNav, Greedy, RRT, SMURF, and GA. By comparison, it can be observed whether EcoNav can maintain good energy efficiency under high-speed conditions, and whether its online learning mechanism can automatically adjust weights to adapt to different speed ranges.

[0038] from Figure 2 The results show that as the robot's moving speed increases from 2 m / s to 4 m / s, the total energy consumption of all algorithms increases significantly, but EcoNav maintains a significantly lower energy consumption curve across all speed ranges. In contrast, the energy consumption of Greedy, RRT, SMURF, and GA increases sharply with speed, exhibiting a typical cubic law effect; while EcoNav, through its energy-aware heuristic and online learning weight adjustment mechanism, can actively choose the downstream direction or a lower-cost trajectory even at high speeds, resulting in a much smaller increase in overall energy consumption compared to other methods. This demonstrates that EcoNav is not only more adaptable to speed changes but also maintains stable and efficient energy consumption performance across different speed scenarios.

[0039] Figure 3Further examining the garbage collection volume under different speed conditions, EcoNav achieved the highest collection volume across all settings, with a slight increase as speed increased, maintaining overall stability and outperforming the baseline methods. In contrast, the collection volume of the other four algorithms did not significantly improve with increased speed, indicating that simply increasing movement speed is insufficient to improve task completion. EcoNav, by integrating environmental information, energy consumption models, and online adjustment mechanisms, can maintain high target collection efficiency under varying speeds. Overall, EcoNav demonstrates a good balance between energy consumption control and task performance, maintaining excellent path planning quality even in high-energy-consuming environments at high speeds, highlighting its robustness and overall advantages.

[0040] The effect of obstacle size In real marine environments, there may be numerous obstacles such as aquaculture facilities or reefs, the size and density of which directly affect feasible paths and obstacle avoidance costs. This invention, while keeping the amount of debris and the robot's speed constant (3 m / s), varies the obstacle coverage to simulate environments of different complexities. Specifically, the obstacle coverage increases from 5% to 50%, with each experimental group increasing the obstacle coverage by 5%, and 35 trials are performed under each condition.

[0041] Under each obstacle density condition, the embodiments of the present invention compare the differences between EcoNav and four other algorithms in terms of total energy consumption and garbage collection efficiency. Two points are of particular concern: (1) in high-density obstacle environments, whether EcoNav's step-by-step walking strategy and energy consumption-aware cost can effectively avoid lengthy detours; (2) in low-density environments, whether EcoNav can still maintain or outperform the energy consumption and efficiency performance of traditional algorithms.

[0042] from Figure 4The results show that as obstacle coverage increases from 5% to 50%, the passable area of ​​the overall environment decreases, causing the total energy consumption of all path planning algorithms to increase. However, EcoNav maintains significantly lower energy consumption across all obstacle densities. At low obstacle densities (5%–15%), EcoNav's energy consumption curve remains almost flat with only a slight increase, while traditional algorithms show significant detour cost accumulation at this stage. As obstacle density further increases to the medium-high range (30%–50%), the energy consumption of RRT, Greedy, and GA spikes dramatically, mainly because these methods tend to perform large-scale detours or generate less regular paths in complex terrain. In contrast, EcoNav's stepwise walking strategy and energy-aware cost calculation allow it to more proactively assess local obstacle density and avoid lengthy detours, resulting in a much lower energy consumption increase than the other four methods. This demonstrates that EcoNav possesses stronger path selection resilience in complex environments and can effectively reduce the additional energy costs associated with high-density obstacle scenarios.

[0043] Figure 5 The data shows the number of trash collected under different obstacle coverage rates. The overall trend is a slight decrease with increasing obstacle density, but EcoNav consistently achieves the highest collection volume under all obstacle density conditions. In low obstacle density areas, the differences between methods are not significant; however, when the obstacle coverage exceeds 25%, the collection volume of traditional methods begins to decline rapidly, especially RRT and Greedy, which miss a large number of targets in high-density environments. In contrast, EcoNav, through its joint cost function of energy consumption and distance, tends to find feasible and energy-efficient paths, enabling it to maintain relatively stable collection efficiency in high-density scenarios. Overall, the results of this invention verify the robustness of EcoNav under different environmental complexities: it not only avoids significant energy consumption degradation caused by excessive obstacles but also consistently maintains superior trash collection performance compared to other algorithms, providing good support for its deployment in real marine environments.

[0044] The influence of ocean current velocity Ocean currents are one of the key factors affecting the energy consumption and path selection of surface robots. This invention explores the energy consumption advantages and adaptability of EcoNav under different ocean current velocities. In the simulation, the robot speed was kept at 3.0 m / s and the obstacle coverage was fixed at 20%. The average ocean current velocity was gradually increased from 0.2 m / s to 2.0 m / s in step increments of 0.2 m / s. 35 experiments were performed under each set of conditions.

[0045] Under varying ocean current intensities, EcoNav dynamically adjusts its path based on an energy consumption model and online learning, favoring downstream areas and avoiding prolonged upstream navigation. This invention's embodiments statistically analyze total energy consumption and the amount of debris successfully collected under various speed conditions, comparing them with Greedy, RRT, SMURF, and GA to verify whether EcoNav can maintain good energy efficiency and collection performance in high-velocity environments.

[0046] from Figure 6 The results show that as the average ocean current velocity increases from 0.2 m / s to 2.0 m / s, the total energy consumption of all path planning algorithms generally increases. This is mainly because the energy cost of countercurrent navigation is significantly amplified under high current conditions. However, EcoNav maintains the flattest and lowest energy consumption curve across all ocean current intensities. In the low to medium current velocity range (0.2–1.0 m / s), EcoNav's energy consumption remains relatively stable, indicating that its online learning mechanism can effectively identify downstream areas and actively avoid upstream sections. In contrast, Greedy, RRT, and GA show a multiple increase in energy consumption under medium to high current velocities, especially above 1.2 m / s. Due to the lack of energy consumption modeling for current direction, these methods often result in passive countercurrent navigation or large-angle path corrections, causing a significant spike in overall energy consumption. EcoNav's advantage is particularly evident under high current conditions, demonstrating that its energy-aware strategy can effectively offset the additional propulsion costs caused by ocean current interference.

[0047] Figure 7 The study demonstrates the amount of trash collected under different ocean current intensities. The overall trend shows a slight decrease with increasing current velocity, but EcoNav maintains the highest trash collection volume across all velocity ranges. At low current velocities, the differences between algorithms are minimal; however, as the current velocity exceeds 1.0 m / s, the collection volume of traditional algorithms significantly decreases. This is because high current velocities affect their path stability, causing them to deviate from the target area or fail to effectively cover the trash distribution area. In contrast, EcoNav, through its energy consumption model and ocean current sensing mechanism, can more accurately adjust feasible paths, maintaining stable task coverage and a high target collection volume even at medium to high current velocities. Overall, this invention verifies EcoNav's robustness under different ocean current intensities—maintaining not only a significant energy consumption advantage but also efficient target search and task execution capabilities, demonstrating its good adaptability and practical value in dynamic marine environments.

[0048] Example 2 The present invention also provides a dynamic water surface environment path planning device, comprising: The first processing module is used to construct a fusion energy consumption estimation model based on the cubic law of fluid dynamics and direction penalty; The second processing module is used to obtain a dynamic target re-evaluation and switching strategy that combines hysteresis threshold and minimum committed distance based on the fusion energy consumption estimation model, so as to realize the energy consumption perception adaptive path planning of the water surface cleaning robot.

[0049] As one embodiment of the present invention, the first processing module is used to construct a loss function based on the error between the measured energy consumption per unit distance and the energy consumption per unit distance predicted by the current weight and the fusion energy consumption estimation model after each movement, and to update the weights using the stochastic gradient descent method.

[0050] As one embodiment of the present invention, the second process is performed as follows: Hysteresis switching condition: only when a new candidate target The estimated total cost meets Target switching is only triggered at that time; Minimum commitment distance constraint: in the direction of a selected target The distance traveled did not exceed the preset minimum commitment distance. Previously, switching was prohibited even if a better target became available.

[0051] As one embodiment of the present invention, the second processing module is used to, at each decision step, allow the robot to select the grid that minimizes the single-step cost within its limited neighborhood as the next movement position based solely on a unified heuristic cost function.

[0052] Example 3 The present invention also provides a dynamic water surface environment path planning system, comprising: a memory and a processor, wherein the memory stores a computer program executed by the processor, and the computer program executes a dynamic water surface environment path planning method when run by the processor.

[0053] Example 4 The present invention also provides a storage medium storing a computer program, which executes a dynamic water surface environment path planning method during runtime.

[0054] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A dynamic water surface environment path planning method, characterized in that, include: Step S1: Construct a fusion energy consumption estimation model based on the cubic law of fluid dynamics and directional penalty; Step S2: Based on the fusion energy consumption estimation model, a dynamic target re-evaluation and switching strategy combining hysteresis threshold and minimum committed distance is obtained to realize energy consumption perception adaptive path planning for the water surface cleaning robot.

2. The dynamic water surface environment path planning method as described in claim 1, characterized in that, In step S1, after each move, a loss function is constructed based on the error between the measured energy consumption per unit distance and the energy consumption per unit distance predicted by the current weights and the fusion energy consumption estimation model, and the weights are updated using the stochastic gradient descent method.

3. The dynamic water surface environment path planning method as described in claim 2, characterized in that, In step S2, Hysteresis switching condition: only when a new candidate target is selected The estimated total cost meets Target switching is only triggered at that time; Minimum commitment distance constraint: in the direction of a selected target The distance traveled did not exceed the preset minimum commitment distance. Previously, switching was prohibited even if a better target became available.

4. The dynamic water surface environment path planning method as described in claim 3, characterized in that, In step S2, at each decision step, the robot selects the grid that minimizes the cost per step within its limited neighborhood based solely on a unified heuristic cost function as the next move position.

5. A dynamic water surface environment path planning device, characterized in that, include: The first processing module is used to construct a fusion energy consumption estimation model based on the cubic law of fluid dynamics and direction penalty; The second processing module is used to obtain a dynamic target re-evaluation and switching strategy that combines hysteresis threshold and minimum committed distance based on the fusion energy consumption estimation model, so as to realize the energy consumption perception adaptive path planning of the water surface cleaning robot.

6. The dynamic water surface environment path planning device as described in claim 5, characterized in that, The first processing module is used to construct a loss function based on the error between the measured energy consumption per unit distance and the energy consumption per unit distance predicted by the current weights and the fusion energy consumption estimation model after each movement, and to update the weights using the stochastic gradient descent method.

7. The dynamic water surface environment path planning device as described in claim 6, characterized in that, The second process is executed as follows: Hysteresis switching condition: only when a new candidate target is selected The estimated total cost meets Target switching is only triggered at that time; Minimum commitment distance constraint: in the direction of a selected target The distance traveled did not exceed the preset minimum commitment distance. Previously, switching was prohibited even if a better target became available.

8. The dynamic water surface environment path planning device as described in claim 7, characterized in that, The second processing module is used to select the grid that minimizes the single-step cost as the next move position in the robot's limited neighborhood at each decision step, based solely on a unified heuristic cost function.

9. A dynamic water surface environment path planning system, characterized in that, include: A memory and a processor, wherein the memory stores a computer program executed by the processor, the computer program performing the dynamic water surface environment path planning method as described in any one of claims 1-4 when executed by the processor.

10. A storage medium, characterized in that, The storage medium stores a computer program that, when executed, performs the dynamic water surface environment path planning method as described in any one of claims 1-4.

Images