Autonomous obstacle avoidance system of underwater robot based on fuzzy control

Abstract

The invention discloses a fuzzy-control-based autonomous obstacle avoidance system for an underwater robot, and relates to the technical field of fuzzy control, which comprises the steps of realizing accurate perception and boundary correction of obstacles by analyzing sonar data, optical images and inertial navigation information, extracting a target position in combination with echo signal features, enhancing the accuracy of environmental data, and obtaining an autonomous obstacle avoidance result. A fuzzy K-nearest neighbor algorithm is adopted, target clustering, moving target classification and environmental impact factor calculation are carried out based on dynamic information of obstacles, so that obstacle avoidance path planning can adapt to different topographic conditions and obstacle distribution, the applicability of path calculation is improved, a Dikstra algorithm is adopted to demarcate a traffic area, and the obstacle avoidance path planning efficiency is improved. And combining path interference calculation and risk assessment, optimizing a path weight matrix, and ensuring that an obstacle avoidance path realizes an optimal decision at a balance point of safety and trafficability. The path optimization link is based on energy consumption calculation and propulsion power optimization, the navigation stability is improved, and the burden of a propulsion system is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of underwater robots, and in particular to an autonomous obstacle avoidance system of an underwater robot based on fuzzy control. Background Art

[0002] Existing underwater robots have a lag in perception information when environmental parameters change significantly or obstacle distribution is complex, resulting in the inability of obstacle avoidance decisions to adapt to emergencies in real time. During obstacle detection, the sonar data and inertial information are fused in a single way, which is easily affected by echo interference and water flow changes, resulting in target recognition errors and affecting obstacle avoidance accuracy. In addition, path planning relies on fixed rules and lacks a comprehensive judgment of terrain features and the state of moving targets, resulting in obstacle avoidance strategies being limited to local optimality and difficult to adapt to the complexity of dynamic environments. As a result, underwater robots are unstable in obstacle avoidance in complex mission environments, and even missions are interrupted due to obstacle avoidance failures, affecting operation continuity and safety.

[0003] The field of fuzzy control technology aims to solve uncertainty, nonlinearity and complex system control problems that are difficult to deal with by traditional control methods. Through fuzzy set theory, fuzzy logic reasoning and fuzzy rule base, experience and expert knowledge are converted into computer-executable control strategies, enabling the system to maintain stable operation in complex environments and large parameter changes.

[0004] Therefore, there is an urgent need for an underwater robot autonomous obstacle avoidance system based on fuzzy control to realize the autonomous obstacle avoidance of underwater robots, ensure stable operation, enhance the intelligence level of underwater robots in tasks, and improve the stability and safety of task execution. Summary of the Invention

[0005] In order to solve the above technical problems, the present invention proposes an autonomous obstacle avoidance system for an underwater robot based on fuzzy control, which can realize autonomous obstacle avoidance in complex underwater environments. It uses fuzzy reasoning and expert experience rules to make flexible decisions in uncertain environments, enabling the robot to autonomously perceive obstacles, determine obstacle avoidance paths and adjust its heading or speed to avoid collisions, ensure stable operation, enhance the robot's intelligence level in tasks, and improve the stability and safety of task execution.

[0006] In order to achieve the above object, the technical solution of the present invention is as follows:

[0007] An autonomous obstacle avoidance system for an underwater robot based on fuzzy control, characterized by comprising:

[0008] Environmental perception module: acquires underwater sonar data, optical images, and inertial navigation information, analyzes echo signals, extracts target positions and corrects boundaries, performs image analysis, segments obstacle contours, calculates inertial data, corrects obstacle relative positions, and obtains dynamic obstacle information;

[0009] State modeling module: Based on the dynamic information of the obstacles, the fuzzy K-nearest neighbor algorithm is used to perform terrain modeling, segment the terrain types to cluster targets and classify moving targets, and perform fuzzy weight assignment, calculate environmental impact factors, analyze trajectories, predict target states, and obtain environmental state characteristics;

[0010] Obstacle avoidance strategy calculation module: Based on the environmental state characteristics, the Dijkstra algorithm is used to delineate the passage area, determine the feasible area, perform path interference calculation, analyze path conflicts, perform risk classification, evaluate the obstacle avoidance cost, and obtain the obstacle avoidance path weight matrix;

[0011] Path optimization module: Based on the obstacle avoidance path weight matrix, it performs path evaluation, calculates comprehensive scores, calculates energy consumption, optimizes propulsion power, selects paths, adjusts path smoothness, and obtains an optimized navigation path;

[0012] Real-time control module: Based on the optimized navigation path, it calculates control variables, adjusts propulsion torque, makes course corrections, compensates for navigation deviations, adjusts the propulsion system, matches power requirements, and obtains navigation execution instructions;

[0013] Navigation adjustment module: Based on the navigation execution instructions, it performs deviation correction, adjusts the navigation path, suppresses errors, optimizes navigation stability and fine-tunes waypoints, adjusts target positions, and obtains navigation correction parameters.

[0014] According to the autonomous obstacle avoidance system based on fuzzy control for an underwater robot according to claim 1, the environment perception module comprises:

[0015] Target positioning submodule: This module collects underwater sonar data, optical images, and inertial navigation information through sensors, collects and filters echo signals, separates background noise, extracts target signals, calculates signal propagation delay, performs beamforming processing, analyzes echo energy distribution, determines target reflection areas, analyzes echo signal characteristics, matches target geometric parameters, corrects signal interference errors, and calculates target spatial coordinates.

[0016] Obstacle parsing submodule: Based on the target spatial coordinates, it performs denoising and contrast enhancement on the optical image data, separates the background area, extracts the target contour features, performs edge detection and morphological analysis, segments the obstacle area, extracts the boundary point set, matches the structural parameters of the obstacle, calculates the connectivity of the target morphology, corrects the recognition error, and obtains the obstacle morphological contour;

[0017] Dynamic calculation submodule: Based on the obstacle shape profile, it interpolates the inertial navigation information, compensates for sensor drift errors, analyzes the angular velocity and acceleration information in the inertial data, calculates the obstacle's motion trajectory, corrects the obstacle's relative position, associates the obstacle's historical motion state, calculates the trajectory change trend, establishes an obstacle motion parameter model, and obtains the obstacle's dynamic information.

[0018] Preferably, the state modeling module includes:

[0019] Terrain modeling submodule: Based on the dynamic information of the obstacles, collect and analyze terrain data, use ranging information to calculate regional depth, remove noise points and perform gradient analysis, identify terrain boundaries, calculate the terrain height change rate and screen steep areas, use the fuzzy K-nearest neighbor algorithm to classify terrain data, calculate the membership of sample points to different terrain types, perform terrain connectivity analysis, identify transition areas, calculate the terrain surface curvature and separate different terrain types, establish regional topological relationships, and obtain terrain structural characteristics;

[0020] Target classification submodule: Based on the terrain structure characteristics, it screens target data, separates dynamic targets from static background, calculates the target's speed change rate and direction angle offset, extracts target motion trajectories and performs trajectory similarity analysis, clusters targets with different motion patterns, calculates target behavioral stability and performs category matching, assigns fuzzy weights, calculates environmental impact factors, and obtains target motion categories;

[0021] State prediction submodule: Based on the target motion category, it traces back historical trajectory data, analyzes the change trend of the target position over time, calculates the target path deviation rate, identifies trajectory intersections and performs motion conflict analysis, establishes a motion state transition matrix, calculates the target's future motion trend and associates it with environmental influencing factors, predicts the target's subsequent state, and obtains environmental state characteristics.

[0022] Preferably, the fuzzy K-nearest neighbor algorithm has the following formula:

[0023]

[0024] Where: u ij is the membership degree of terrain category j in the underwater robot environment, d ij is the Euclidean distance between the current position of the underwater robot and the centroid of terrain category j, d ik is the Euclidean distance between the current position of the underwater robot and the centroid of terrain category k, c is the total number of terrain categories in the environment, m is the fuzzy index, v ij is the terrain curvature change rate of the underwater robot at terrain category j, g ij is the gradient value of the underwater robot at terrain category j, h ijis the terrain height change of the underwater robot at terrain category j, w1, w2, w3 are weight coefficients, w n is the normalized weight, E n is information entropy, p n is the normalized probability distribution of parameter n.

[0025] Preferably, the obstacle avoidance strategy calculation module includes:

[0026] The passage area delineation submodule: Based on the environmental state characteristics, the Dijkstra algorithm is used to analyze the underwater terrain data, extract the boundary point set of fixed obstacles, calculate the obstacle height change, screen the steep areas and mark the inaccessible areas, combine the dynamic obstacle trajectory data, calculate the target activity range, perform regional connectivity detection, eliminate isolated and invalid areas, build a passage network model, delineate the passage range, and obtain the feasible navigation area;

[0027] Path interference calculation submodule: Based on the feasible navigation area, it screens feasible paths, calculates the intersection points between path segments, analyzes the relative movement speed of obstacles, calculates the occupancy of navigation paths at different times, analyzes the time window of potential conflict points, calculates the interference intensity of navigation paths, screens high-risk conflict areas, adjusts path priorities, marks the degree of interference on paths, and obtains path conflict distribution;

[0028] Obstacle avoidance cost evaluation submodule: Based on the path conflict distribution, calculate the energy consumption change required for path adjustment, analyze the impact of obstacle avoidance detour on navigation distance, analyze the time consumption of path adjustment, calculate the navigation stability of the obstacle avoidance path, extract the stability parameters of different obstacle avoidance schemes, construct a weight scoring system for obstacle avoidance adjustment schemes, screen path adjustment schemes, and obtain the obstacle avoidance path weight matrix.

[0029] Preferably, the Dijkstra algorithm has the following formula:

[0030] C(n)=min(C(m)+p1·P(m,n)+p2·H(n)+p3·G(n)+p4·F(n))

[0031] Where: C(n) is the shortest path cost of the underwater robot from the starting position to the current node n, C(m) is the shortest path cost of the underwater robot from the starting position to the previous node n, P(m,n) is the basic path weight of the underwater robot from node m to node n, H(n) is the obstacle height change at the current node n, G(n) is the steep area gradient at the current node n, F(n) is the dynamic obstacle trajectory influence factor at the current node n, p1 is the adjustment coefficient of the path basic weight, p2 is the weight coefficient of the obstacle height change, p3 is the weight coefficient of the steep area gradient, p4 is the weight coefficient of the dynamic obstacle trajectory influence factor, Eq is the information entropy of underwater environment characteristic data, r q is the normalized probability distribution of feature data, p q is the calculated feature weight value.

[0032] Preferably, the path optimization module includes:

[0033] Path scoring submodule: Based on the obstacle avoidance path weight matrix, it screens available path data, analyzes path traffic conditions, extracts obstacle distribution within the path, calculates path risk level, measures the density of path interference areas, analyzes the impact of path adjustment on navigation time, calculates connectivity between navigation paths, evaluates the stability of path changes, screens path traffic priority, and obtains path scoring results;

[0034] Propulsion optimization submodule: Based on the path scoring results, calculate the hydrodynamic parameters of the path segment, analyze the impact of fluid resistance on propulsion efficiency, extract the depth changes along the path, calculate the navigation propulsion power requirement, analyze the adaptability of the propulsion system to different paths, adjust the propulsion torque distribution, optimize the navigation propulsion power, match the path energy consumption requirements, and obtain optimized propulsion parameters;

[0035] Smoothing adjustment submodule: Based on the optimized propulsion parameters, analyze the path curvature changes, screen the low curvature path segments, calculate the heading adjustment angle, adjust the position of the heading change point, optimize the smoothness of the connection between path segments, eliminate the influence of mutation points, smooth the path curve, optimize the continuity of the navigation path, adjust the path node distribution, and obtain the optimized navigation path.

[0036] Preferably, the real-time control module includes:

[0037] Control variable calculation submodule: Based on the optimized navigation path, analyze the path heading change points, calculate the relative position of the navigation target point, analyze the path deviation trend, extract the navigation state parameters, calculate the navigation control variables, screen the key parameters for navigation adjustment, establish the navigation adjustment plan, and obtain the navigation control parameters;

[0038] Heading correction submodule: Based on the navigation control parameters, it analyzes the deviation between the current heading and the target heading, calculates the heading angle adjustment amount, analyzes the impact of the heading correction on the propulsion force, selects the key points for heading adjustment, calculates the adjustment sequence of the heading change points, optimizes the navigation correction process, and obtains the heading correction instruction;

[0039] Propulsion adjustment submodule: Based on the heading correction instruction, analyze the propulsion torque demand, calculate the propulsion force adjustment range, match the power distribution of different propulsion modes, analyze the impact of propulsion adjustment on navigation stability, screen the propulsion control scheme, adjust the propulsion system power, and obtain the navigation execution instruction.

[0040] Preferably, the navigation adjustment module includes:

[0041] Deviation correction submodule: Based on the navigation execution instruction, it extracts real-time navigation data, analyzes the deviation between the heading angle and the target heading, calculates the impact of navigation speed on the deviation, analyzes the lateral drift caused by uneven thrust during navigation, determines the longitudinal position change trend, selects the heading deviation mutation point, calculates the correction force required for heading adjustment, adjusts the propulsion torque distribution, corrects the heading control error, and obtains the deviation correction parameter;

[0042] Path adjustment submodule: Based on the deviation correction parameters, extract the adjusted navigation path data, analyze the connection relationship between path segments, calculate the distance change of the path adjustment, detect the curvature change of the path change point, calculate the impact of the adjusted path on the speed, optimize the connection sequence of path segments, detect path continuity and eliminate unstable routes, calculate the correction amplitude of the navigation path, adjust the navigation trajectory, and obtain the adjusted navigation path;

[0043] Waypoint fine-tuning submodule: Based on the adjusted navigation path, analyze the waypoint coordinate offset, calculate the impact of waypoint adjustment on path smoothness, measure the spacing change between waypoints, screen the waypoint area that needs to be adjusted, calculate the waypoint fine-tuning range, adjust the waypoint position to match the navigation trajectory, correct the target positioning error, optimize the navigation direction, and obtain the navigation correction parameters.

[0044] Based on the above technical solution, the beneficial effects of the present invention are:

[0045] 1. The present invention uses the fuzzy K-nearest neighbor algorithm to perform target clustering, moving target classification, and environmental impact factor calculation based on the dynamic information of obstacles, so that the obstacle avoidance path planning can adapt to different terrain conditions and obstacle distributions, thereby improving the applicability of path calculation.

[0046] 2. This invention uses the Dijkstra algorithm to define traversable areas. It then combines path interference calculation with risk assessment to optimize the path weight matrix, ensuring that the optimal obstacle avoidance path strikes a balance between safety and passability. Path optimization, based on energy consumption calculation and propulsion power optimization, enables smooth path adjustments, improves navigation stability, and reduces the burden on the propulsion system.

[0047] 3. The present invention ensures the precise execution of navigation instructions through propulsion torque adjustment, heading correction and power matching. At the same time, it combines deviation suppression and error correction to enhance the stability of the navigation path. The combination of path optimization and real-time adjustment enables the robot to dynamically adjust its heading in complex underwater environments, reduce navigation deviations caused by environmental changes, and enhance its intelligent adaptability when performing tasks. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] The specific embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

[0049] Figure 1 is a flow chart of an autonomous obstacle avoidance system of an underwater robot based on fuzzy control in one embodiment;

[0050] Figure 2 The present invention is a schematic block diagram of an autonomous obstacle avoidance system of an underwater robot based on fuzzy control in an embodiment. DETAILED DESCRIPTION

[0051] In order to explain the present invention more clearly, the present invention is further described below in conjunction with preferred embodiments. Those skilled in the art should understand that the following specific description is illustrative rather than restrictive and should not be used to limit the scope of protection of the present invention.

[0052] See also Figure 1 This embodiment provides an autonomous obstacle avoidance system for an underwater robot based on fuzzy control, comprising:

[0053] Environmental perception module: acquires underwater sonar data, optical images, and inertial navigation information, analyzes echo signals, extracts target positions and corrects boundaries, performs image analysis, segments obstacle contours, calculates inertial data, corrects obstacle relative positions, and obtains dynamic obstacle information;

[0054] State modeling module: Based on the dynamic information of obstacles, the fuzzy K-nearest neighbor algorithm is used to perform terrain modeling, segment terrain types, cluster targets, classify moving targets, and perform fuzzy weight assignment, calculate environmental impact factors, analyze trajectories, predict target states, and obtain environmental state characteristics;

[0055] Obstacle avoidance strategy calculation module: Based on environmental state characteristics, the Dijkstra algorithm is used to delineate the passage area, determine the feasible area, perform path interference calculations, analyze path conflicts, perform risk classification, evaluate obstacle avoidance costs, and obtain the obstacle avoidance path weight matrix;

[0056] Path optimization module: Based on the obstacle avoidance path weight matrix, it performs path evaluation, calculates comprehensive scores, calculates energy consumption, optimizes propulsion power, screens paths, adjusts path smoothness, and obtains the optimized navigation path;

[0057] Real-time control module: Based on the optimized navigation path, it calculates control variables, adjusts propulsion torque, makes course corrections, compensates for navigation deviations, adjusts the propulsion system, matches power requirements, and obtains navigation execution instructions;

[0058] Navigation adjustment module: Based on the navigation execution instructions, it performs deviation correction, adjusts the navigation path, suppresses errors, optimizes navigation stability and fine-tunes waypoints, adjusts target positions, and obtains navigation correction parameters.

[0059] See also Figure 2 In one embodiment, an autonomous obstacle avoidance system for an underwater robot based on fuzzy control also provides specific settings of an environment perception module, including the following:

[0060] Target positioning submodule: This module collects underwater sonar data, optical images, and inertial navigation information through sensors, collects and filters echo signals, separates background noise, extracts target signals, calculates signal propagation delay, performs beamforming processing, analyzes echo energy distribution, determines target reflection areas, analyzes echo signal characteristics, matches target geometric parameters, corrects signal interference errors, and calculates target spatial coordinates.

[0061] Obstacle parsing submodule: Based on the target spatial coordinates, it performs denoising and contrast enhancement on the optical image data, separates the background area, extracts the target contour features, performs edge detection and morphological analysis, segments the obstacle area, extracts the boundary point set, matches the structural parameters of the obstacle, calculates the connectivity of the target morphology, corrects the recognition error, and obtains the obstacle morphological contour;

[0062] Dynamic calculation submodule: Based on the obstacle shape contour, it interpolates the inertial navigation information, compensates for sensor drift error, analyzes the angular velocity and acceleration information in the inertial data, calculates the obstacle's motion trajectory, corrects the obstacle's relative position, associates the obstacle's historical motion state, calculates the trajectory change trend, establishes an obstacle motion parameter model, and obtains the obstacle's dynamic information.

[0063] Specifically, the target positioning submodule: based on the underwater sonar data, optical images and inertial navigation information collected by the sensor, the bandpass filter parameters are set to a center frequency of 100kHz and a bandwidth of 20kHz to filter the echo signal, and the short-time Fourier transform window size is 1024, the overlap rate is 50% to decompose the frequency components, separate the background noise, extract the target signal and calculate the signal propagation delay, use the coherent beamforming algorithm to use the Kaiser window function in a weighted manner, parameter β = 5 to perform coherent superposition on the multipath echo, and perform matched filtering template signal based on the transmitted pulse, matching window length 64 points, analyze the echo energy distribution, and use the two-dimensional peak detection threshold set to 60% of the maximum energy value to determine the target reflection area. The improved Hough transform voting threshold is set to 100, and the minimum detection straight line length is 50 pixels to analyze the echo signal characteristics, and the target geometric parameter matching is based on the ICP iterative nearest point algorithm with a maximum number of iterations of 50 and a convergence threshold of 0.001. The signal interference error is corrected and the target spatial coordinates are calculated.

[0064] Obstacle parsing submodule: Based on the target spatial coordinates, the optical image data is subjected to median filtering with a window size of 5×5 for denoising and histogram equalization with a grayscale level of 256 for contrast enhancement. The background area is separated, and the Canny edge detection algorithm is used with a high threshold of 0.3, a low threshold of 0.1, and a filter kernel size of 3×3 to extract the target contour features. Morphological gradient operation with a kernel size of 5×5 is performed to calculate the local gradient amplitude. The watershed segmentation algorithm is used based on the Euclidean distance transform, marking the foreground area as the connected area after morphological expansion to segment the obstacle area. The Douglas-Peucker algorithm is used with a maximum tolerance of 2 pixels to extract the boundary point set. The RANSAC random sampling consensus algorithm is used with 1000 iterations and an inlier threshold of 1.5 pixels to match the structural parameters of the obstacle. The connectivity of the target morphology is calculated, and a morphological closing operation with a kernel size of 7×7 is performed to correct the recognition error and obtain the obstacle morphological contour.

[0065] Dynamic calculation submodule: Based on the obstacle morphological contour, cubic spline interpolation is performed on the inertial navigation information with the number of interpolation nodes set to 50. Natural boundaries are used as boundary conditions to compensate for sensor drift errors. The angular velocity in the inertial data is calculated based on the fourth-order Runge-Kutta method with a step size of 0.01s and a low-pass filter threshold of 5Hz for acceleration information. The Kalman filter state transfer matrix is set to the unit matrix and the measurement noise variance is set to 0.01 to calculate the motion trajectory of the obstacle. The trajectory is then fitted based on the least squares method with a fitting order of 3 and a data window size of 10 frames to correct the relative position of the obstacle. The dynamic time warping algorithm DTW is used with a window width set to 10% of the trajectory sequence length to associate the historical motion state of the obstacle and calculate the trajectory change trend. A particle filter with 1000 particles and a resampling threshold of 0.5 is used to establish the obstacle motion parameter model and obtain the obstacle dynamic information.

[0066] See also Figure 2 In one embodiment, an autonomous obstacle avoidance system for an underwater robot based on fuzzy control also provides specific settings of a state modeling module, including the following:

[0067] Terrain modeling submodule: Based on the dynamic information of obstacles, it collects and analyzes terrain data, uses ranging information to calculate regional depth, removes noise points and performs gradient analysis, identifies terrain boundaries, calculates the rate of change of terrain height and screens steep areas, uses the fuzzy K-nearest neighbor algorithm to classify terrain data, calculates the membership of sample points to different terrain types, performs terrain connectivity analysis, identifies transition areas, calculates terrain surface curvature and separates different terrain types, establishes regional topological relationships, and obtains terrain structural characteristics;

[0068] Target classification submodule: Based on terrain structure features, it filters target data, separates dynamic targets from static backgrounds, calculates the target's velocity change rate and directional angle offset, extracts target motion trajectories and performs trajectory similarity analysis, clusters targets with different motion patterns, calculates target behavioral stability and performs category matching, assigns fuzzy weights, calculates environmental impact factors, and obtains target motion categories;

[0069] State prediction submodule: Based on the target motion category, it traces back historical trajectory data, analyzes the target position change trend over time, calculates the target path deviation rate, identifies trajectory intersections and performs motion conflict analysis, establishes a motion state transition matrix, calculates the target's future motion trend and associates it with environmental influencing factors, predicts the target's subsequent state, and obtains environmental state characteristics;

[0070] Specifically, the terrain modeling submodule collects and analyzes terrain data based on the dynamic information of obstacles, calculates the regional depth using the laser ranging method, performs mean filtering on the ranging data with a window size of 5×5 to remove noise points, and performs Sobel gradient calculation with a kernel size of 3×3 to calculate the gradient value. Canny edge detection is used with a high threshold of 0.3 and a low threshold of 0.1 to identify the terrain boundary. The second-order derivative change rate is calculated with a step size of 0.5m to obtain the terrain height change rate, and threshold segmentation is performed with a threshold of 20° to screen out steep areas. The terrain data is classified using the fuzzy K nearest neighbor algorithm with the number of neighbors set to 5 and the fuzzy coefficient m=2. The membership of sample points to different terrain types is calculated based on Euclidean distance calculation. The Dijkstra algorithm is used with the edge weight set to the path slope to perform terrain connectivity analysis. K-Means clustering is used with the number of clusters set to 3 to identify transition areas. Gaussian curvature is calculated with a curvature threshold of 0.05 to separate different terrain types. A Voronoi diagram is established to calculate the regional topological relationship based on the centroid to obtain the terrain structure characteristics.

[0071] Target classification submodule: Based on the terrain structure features, target data is screened. The background mean is calculated based on the mean model using the background subtraction method and a threshold is set to separate dynamic targets from static backgrounds. The target's speed change rate is calculated based on the optical flow method. The target speed is calculated with a Lucas-Kanade window size of 15×15, and the direction angle offset is calculated based on Hough transform line detection to obtain the target direction angle. The target motion trajectory is extracted and dynamic time warping (DTW) is performed. The window width is set to 10% of the trajectory length to calculate the trajectory similarity. DBSCAN clustering is used with a minimum sample number of 5 and a neighborhood radius of 2m to cluster targets with different motion modes. The target's behavioral stability is calculated based on the variance analysis window size set to 10 frames. K nearest neighbor matching is performed with the K value set to 3 for category matching. The fuzzy weight assignment method is used to assign fuzzy weights with a fuzzy factor set to 1.5. The environmental impact factor is calculated based on the standardized Euclidean distance to obtain the target motion category.

[0072] State prediction submodule: Based on the target motion category, historical trajectory data is traced back, the target position change trend over time is analyzed, and prediction is performed based on second-order curve fitting. The target path deviation rate is calculated based on the deviation vector. The average deviation of the historical trajectory is calculated based on the deviation vector. The trajectory intersection is identified and the intersection is calculated based on the Bezier curve intersection judgment. The conflict analysis is performed based on the Markov chain probability transfer matrix to calculate the conflict probability. The motion state transition matrix is established with the matrix dimension set to 3×3 to calculate the state transfer probability. The future motion trend of the target is calculated using the long short-term memory network with the hidden layer set to 64 to predict the future state of the target. The environmental influencing factors are associated and the influence of environmental factors is calculated based on the weighted Euclidean distance. The subsequent state of the target is predicted and the environmental state characteristics are obtained.

[0073] Specifically, the above fuzzy K-nearest neighbor algorithm is based on the formula:

[0074]

[0075] Where: u ij is the membership degree of terrain category j in the underwater robot environment, d ij is the Euclidean distance between the current position of the underwater robot and the centroid of terrain category j, d ik is the Euclidean distance between the current position of the underwater robot and the centroid of terrain category k, c is the total number of terrain categories in the environment, m is the fuzzy index, v ij is the terrain curvature change rate of the underwater robot at terrain category j, g ij is the gradient value of the underwater robot at terrain category j, h ij is the terrain height change of the underwater robot at terrain category j, w1, w2, w3 are weight coefficients, w n is the normalized weight, E n is information entropy, p n is the normalized probability distribution of parameter n.

[0076] Execution process: First, obtain the terrain data of the underwater environment, including the Euclidean distance d from the robot's current position to the center of mass of different terrain categories ij and d ik , measures the spatial proximity of the robot to different terrain categories, and then extracts the terrain curvature change rate v ij and v ik The calculation method is to perform a second-order derivative operation on the terrain surface, extract the local curvature characteristics and perform normalization processing to measure the degree of terrain undulation, and then calculate the terrain gradient value g ij and g ik , process underwater terrain data, calculate slope information, identify steep areas, and further calculate the terrain height change h ij and hik , local height changes are calculated through elevation data, and normalized to reduce the impact of scale differences. In order to optimize the influence weights of terrain feature parameters, the system uses entropy weight method to calculate weights w1, w2 and w3, normalizes all terrain feature data, calculates the relative weight distribution of each parameter, and then calculates the information entropy E of each parameter. n , by normalizing the probability distribution p n Calculate the amount of information. The larger the information entropy value, the more information the parameter provides. Then calculate the parameter weight w based on the information entropy. n , use the normalization formula to determine the weight distribution so that the weight value satisfies ∑w n =1, finally determine the specific values of w1, w2 and w3, and combine d i j, v ij 、g ij and h i j calculates the membership degree u of the underwater robot to different terrain categories ij , generate a terrain classification membership matrix, perform terrain connectivity analysis based on the matrix, identify the crossing areas between different terrain categories, calculate the local terrain surface curvature, separate different terrain types, establish regional topological relationships, obtain the structural characteristics of underwater terrain, and improve the robot's obstacle avoidance ability and navigation accuracy in complex underwater environments.

[0077] See also Figure 2 In one embodiment, an autonomous obstacle avoidance system for an underwater robot based on fuzzy control also provides specific settings of an obstacle avoidance strategy calculation module, including the following:

[0078] The traffic area delineation submodule: Based on environmental state characteristics, the Dijkstra algorithm is used to analyze underwater terrain data, extract the boundary point set of fixed obstacles, calculate the height change of obstacles, screen steep areas and mark inaccessible areas. Combined with dynamic obstacle trajectory data, the target activity range is calculated, regional connectivity is detected, isolated and invalid areas are eliminated, and a traffic network model is constructed to delineate the traffic range and obtain the feasible navigation area.

[0079] Path interference calculation submodule: Based on the feasible navigation area, it screens feasible paths, calculates the intersection points between path segments, analyzes the relative movement speed of obstacles, calculates the occupancy of the navigation path at different times, analyzes the time window of potential conflict points, calculates the interference intensity of the navigation path, screens high-risk conflict areas, adjusts path priorities, marks the degree of interference on the path, and obtains the path conflict distribution;

[0080] Obstacle avoidance cost evaluation submodule: Based on the path conflict distribution, calculate the energy consumption change required for path adjustment, analyze the impact of obstacle avoidance detour on navigation distance, analyze the time consumption of path adjustment, calculate the navigation stability of the obstacle avoidance path, extract the stability parameters of different obstacle avoidance schemes, construct a weight scoring system for obstacle avoidance adjustment schemes, screen path adjustment schemes, and obtain the obstacle avoidance path weight matrix.

[0081] Specifically, the traffic area demarcation submodule: Based on the environmental state characteristics, the Dijkstra algorithm is used to initialize the priority queue, the starting node cost is set to 0, and the costs of all nodes are set to infinity. The underwater terrain data is parsed, and the boundary point set of the fixed obstacle is extracted using the MarchingSquares algorithm with a step size of 1m. The obstacle height change is calculated based on bilinear interpolation to calculate the elevation difference of adjacent points, and the steep area is screened based on the gradient threshold set to 15° to calibrate the inaccessible area. Combined with the dynamic obstacle trajectory data, the target activity range is calculated. The convex hull algorithm CH is used to set the minimum boundary area, and regional connectivity detection is performed. The connected area is searched based on the FloodFill method, and isolated invalid areas are eliminated based on the area threshold set to 3m. 2 ,The traffic network model is constructed by using the adjacency matrix to store the traffic nodes, and the traffic range is ,defined by searching the passable paths based on BFS breadth-first ,search to obtain the feasible navigation area.

[0082] Path interference calculation submodule: Based on the feasible navigation area, the feasible paths are screened using the A* algorithm with the heuristic function set to Euclidean distance to calculate the path cost. The intersections between path segments are calculated based on the Bezier curve intersection detection algorithm to calculate the path intersection. The relative motion speed of obstacles is analyzed using the Kalman filter state transfer matrix set to the unit matrix and the observation noise variance set to 0.01 to calculate the obstacle speed estimation. The occupancy of the navigation path at different times is calculated based on the time discretization method with the time step set to 0.5s to calculate the path occupancy time slice. The time window for analyzing potential conflict points is based on the sliding time window method with the window size set to 5s to calculate the conflict probability. The interference intensity of the navigation path is calculated based on the Poisson distribution model λ=0.2 to calculate the path conflict frequency. The high-risk conflict areas are screened based on the DBSCAN density clustering with the neighborhood radius set to 2m and the minimum number of samples set to 3. The path priority is adjusted using the fuzzy hierarchical analysis method with the weight set to 0.6. The degree of the interfered path is marked and the disturbance level is calculated based on the path interference score to obtain the path conflict distribution.

[0083] Obstacle avoidance cost evaluation submodule: Based on the path conflict distribution, the energy consumption change required for path adjustment is calculated using the energy consumption model. The speed v is set to 1.5m / s and the fluid density ρ = 1025kg / m 3 , drag coefficient Cd = 0.9, windward area A = 0.5m 2, analyze the impact of obstacle avoidance on navigation distance, use Dijkstra algorithm to calculate the shortest path length and compare it with the original path difference, analyze the time consumption of path adjustment, use time integration method based on speed time series integration to calculate navigation time, calculate the navigation stability of obstacle avoidance path, use cubic B-spline curve fitting to calculate path curvature, extract the stability parameters of different obstacle avoidance schemes, calculate the heading deviation based on the root mean square error, construct a weight scoring system for obstacle avoidance adjustment schemes, use hierarchical analysis method to set the weight matrix to 3×3, screen the path adjustment schemes, and calculate the advantages and disadvantages of the schemes based on the multi-objective optimization method Pareto sorting to obtain the obstacle avoidance path weight matrix.

[0084] Specifically, the above Dijkstra algorithm is based on the formula:

[0085] C(n)=min(C(m)+p1·P(m,n)+p2·H(n)+p3·G(n)+p4·F(n))

[0086] Where: C(n) is the shortest path cost of the underwater robot from the starting position to the current node n, C(m) is the shortest path cost of the underwater robot from the starting position to the previous node m, P(m, n) is the basic path weight of the underwater robot from node m to node n, H(n) is the obstacle height change at the current node n, G(n) is the steep area gradient at the current node n, F(n) is the dynamic obstacle trajectory influence factor at the current node n, p1 is the adjustment coefficient of the path basic weight, p2 is the weight coefficient of the obstacle height change, p3 is the weight coefficient of the steep area gradient, p4 is the weight coefficient of the dynamic obstacle trajectory influence factor, E q is the information entropy of underwater environment characteristic data, r q is the normalized probability distribution of feature data, p q is the calculated feature weight value;

[0087] Execution process: First, calculate the basic path weight P(m, n) of each node in the underwater environment, and evaluate the passage cost by measuring the Euclidean distance between nodes m and n, combining water flow resistance and energy consumption cost. Then calculate the obstacle height change H(n), use elevation data to obtain the height difference of underwater terrain, extract the mutation factor domain, filter out the inaccessible obstacle area, and calculate the steep area gradient G(n) at the same time, extract local slope information, judge the possibility of the robot passing in different slope areas, increase the path cost for the steeper area, and further calculate the dynamic Obstacle trajectory influencing factor F(n) is used to track the trajectory of dynamic obstacles using underwater sonar or visual sensors. The Markov prediction model is used to estimate the possible future movement range, assigning higher travel costs to high-frequency activity areas to improve obstacle avoidance stability. Subsequently, the weight coefficients p1, p2, p3, and p4 are determined, and the entropy weight method is used for weight calculation. First, the path weight P(m, n), obstacle height change H(n), steep area gradient G(n), and dynamic obstacle trajectory influencing factor F(n) are normalized to calculate the information entropy E of each feature. q , measure the amount of information provided by different environmental variables, and then calculate the weight value p of each parameter q The weight value is calculated by normalizing all information entropies to ensure that the sum of the weights of each parameter is 1. The improved Dijkstra algorithm is substituted to calculate the shortest path weights from each point in the underwater environment to the target point, and a traffic accessibility matrix is generated. Combined with regional connectivity detection, isolated and invalid areas are eliminated, a traffic network model is constructed, the traffic range is delineated, and a feasible navigation area is obtained to ensure the robot's safe passage and efficient obstacle avoidance in complex underwater environments.

[0088] Analyze the impact of obstacle avoidance and detour on the navigation distance. The impact of obstacle avoidance and detour on the navigation distance can be calculated by using the shortest path algorithm to calculate the path growth, and the additional yaw caused by the detour can be evaluated in combination with the path curvature. If the obstacle avoidance involves complex terrain, the influence of fluid dynamics is considered, including strong water currents in the detour area, so that the effective navigation distance of the detour path increases more than the result of geometric calculation. The total distance change caused by the detour is calculated by the path integral method, and combined with the speed adjustment strategy of the propulsion system, the additional time and energy consumption are evaluated. The planning of the detour path takes into account the acceleration and deceleration limits of navigation to avoid additional energy consumption and decreased stability caused by frequent turns.

[0089] See also Figure 2 In one embodiment, a fuzzy-controlled autonomous obstacle avoidance system for an underwater robot further provides specific settings of a path optimization module, including the following:

[0090] Path scoring submodule: Based on the obstacle avoidance path weight matrix, it screens available path data, analyzes path traffic conditions, extracts obstacle distribution within the path, calculates path risk level, measures the density of path interference areas, analyzes the impact of path adjustment on navigation time, calculates connectivity between navigation paths, evaluates the stability of path changes, screens path traffic priority, and obtains path scoring results.

[0091] Propulsion optimization submodule: Based on the path scoring results, it calculates the hydrodynamic parameters of the path segment, analyzes the impact of fluid resistance on propulsion efficiency, extracts depth changes along the path, calculates the propulsion power requirement for navigation, analyzes the adaptability of the propulsion system to different paths, adjusts the propulsion torque distribution, optimizes the propulsion power for navigation, matches the path energy consumption requirements, and obtains the optimized propulsion parameters;

[0092] Smooth adjustment submodule: Based on the optimized propulsion parameters, it analyzes the path curvature changes, screens the low curvature path segments, calculates the heading adjustment angle, adjusts the position of the heading change point, optimizes the smoothness of the connection between path segments, eliminates the influence of mutation points, smoothes the path curve, optimizes the continuity of the navigation path, adjusts the distribution of path nodes, and obtains the optimized navigation path.

[0093] Specifically, the path scoring submodule: based on the obstacle avoidance path weight matrix, the available path data is screened using the Kruskal minimum spanning tree algorithm with the weight set as the path cost to screen the passable path, the path pass conditions are analyzed, the path risk level is calculated using the hierarchical analysis method (AHP) with the hazard factor weight matrix set to calculate the path risk, the density of the path interference area is measured using the Poisson distribution model with λ set to 0.3 to calculate the path conflict probability, the impact of path adjustment on navigation time is analyzed using the path integral method with the time step set to 0.1s to calculate the time required for path change, the connectivity between navigation paths is calculated using the Floyd-Warshall algorithm to calculate the path connectivity matrix, the stability of path changes is evaluated using the mean square error calculation method with the time window set to 10s to calculate the change stability, the path priority is screened using the fuzzy comprehensive evaluation method with the weight set to 0.6 to calculate the path priority, and the path scoring results are obtained.

[0094] Propulsion optimization submodule: Based on the path scoring results, the hydrodynamic parameters of the path segment are calculated to analyze the effect of fluid resistance on propulsion efficiency. ρ is set to 1025 kg / m 3 , v is set to 1.5m / s, Cd is set to 0.9, and A is set to 0.5m 2The propulsion resistance is calculated, and the depth variation along the path is extracted. The bilinear interpolation method is used to calculate the depth distribution with the interpolation points set to twice the number of path nodes. The propulsion power demand is calculated using the propulsion power calculation formula P = F × v to calculate the propulsion energy consumption. The adaptability of the propulsion system to different paths is analyzed and the dynamic response of the propulsion system is calculated using the modal analysis method. The propulsion torque distribution is adjusted using the optimization allocation algorithm Lagrange multiplier method to calculate the optimal propulsion torque. The sailing propulsion power is optimized using the gradient descent algorithm with a step size of 0.01 to calculate the optimal power output. The path energy consumption demand is matched using the Pareto multi-objective optimization to calculate the matching degree between energy consumption and propulsion power, and the optimized propulsion parameters are obtained.

[0095] Smoothing adjustment submodule: Based on the optimized propulsion parameters, the path curvature change is analyzed and the curvature distribution is calculated using the Bezier curve fitting algorithm with the number of control points set to 5. The low curvature path segments are screened using the second-order curvature differential calculation method with the curvature threshold set to 0.1 to screen the stable path segments. The heading adjustment angle is calculated using the fourth-order Runge-Kutta method with a step size set to 0.02 to calculate the heading change rate. The position of the heading change point is adjusted using the dynamic programming algorithm to relocate the heading point based on the minimum path cost. The smoothness of the connection between path segments is optimized using the cubic spline interpolation node number set to 1.5 times the number of path nodes to calculate the path smoothness. The influence of mutation points is eliminated using the Gaussian filtering method with a window size set to 3×3 to remove trajectory mutations. The smooth path curve is calculated using the Catmull-Rom spline curve interpolation node number set to twice the total number of path points. The continuity of the navigation path is optimized using local B-spline curve fitting to calculate the heading continuity. The distribution of path nodes is adjusted using the equidistant interpolation algorithm with the node spacing set to 0.5m to evenly distribute the path points to obtain the optimized navigation path.

[0096] See also Figure 2 In one embodiment, a real-time control module is also provided in an autonomous obstacle avoidance system of an underwater robot based on fuzzy control, including the following:

[0097] Control variable calculation submodule: Based on the optimized navigation path, it analyzes the path and heading change points, calculates the relative position of the navigation target point, analyzes the path deviation trend, extracts the navigation status parameters, calculates the navigation control variables, selects the key parameters for navigation adjustment, establishes the navigation adjustment plan, and obtains the navigation control parameters;

[0098] Heading correction submodule: Based on the navigation control parameters, it analyzes the deviation between the current heading and the target heading, calculates the heading angle adjustment amount, analyzes the impact of the heading correction on the propulsion force, selects the key points for heading adjustment, calculates the adjustment sequence of the heading change points, optimizes the navigation correction process, and obtains the heading correction instruction;

[0099] Propulsion adjustment submodule: Based on the heading correction instruction, it analyzes the propulsion torque demand, calculates the propulsion force adjustment range, matches the power distribution of different propulsion modes, analyzes the impact of propulsion adjustment on navigation stability, screens the propulsion control scheme, adjusts the propulsion system power, and obtains navigation execution instructions.

[0100] Specifically, the control variable calculation submodule: based on the optimized navigation path, the path heading change points are analyzed, the Douglas-Peucker algorithm is used to set the maximum tolerance to 1.5 degrees to identify the path change inflection points, the relative position of the navigation target point is calculated, the relative distance between the target point and the current position is calculated, the path deviation trend is analyzed, the Kalman filter state transfer matrix is set to the unit matrix, the measurement noise variance is set to 0.01 to calculate the path deviation change, the navigation state parameters are extracted, the inertial navigation system INS data is used to analyze the current speed, acceleration, and heading angle, the navigation control variables are calculated, the PID control algorithm is used to set the proportional gain Kp = 1.2, the integral gain Ki = 0.01, and the differential gain Kd = 0.5 to calculate the heading adjustment amount, the key parameters of the navigation adjustment are screened, the principal component analysis PCA is used to set the cumulative contribution rate to 90% to extract the key variables, the navigation adjustment plan is established, the fuzzy logic control is used to set the fuzzy set based on the rule base for adjustment calculation, and the navigation control parameters are obtained.

[0101] Heading correction submodule: Based on the navigation control parameters, the deviation between the current heading and the target heading is analyzed, and the heading angle adjustment amount is calculated. The fourth-order Runge-Kutta method is used with a step size of 0.02 to calculate the heading adjustment change rate. The impact of the heading correction on the propulsion force is analyzed using the propulsion torque calculation formula M=F×r, where F is set as the current propulsion force and r is set as the propulsion center distance to calculate the impact of the correction on the torque. The key points for heading adjustment are screened and the Bezier curve fitting control points are set to 3 to calculate the optimal adjustment point. The adjustment sequence of the heading change points is calculated using a dynamic programming algorithm to calculate the optimal adjustment path based on the shortest adjustment cost. The navigation correction process is optimized and the heading correction instruction is obtained.

[0102] Propulsion regulation submodule: Based on the heading correction instruction, the propulsion torque demand is analyzed and the propulsion force adjustment range is calculated. The modal analysis method is used to calculate the stability of the propulsion system under different thrusts. The power distribution of different propulsion modes is matched using the Pareto optimal allocation method to calculate the optimal power matching. The impact of propulsion adjustment on navigation stability is analyzed using the frequency domain analysis method to calculate the impact of propulsion power changes on stability. The particle swarm optimization algorithm is used to screen the propulsion control scheme, with the number of particles set to 100 and the inertia weight set to 0.7 to search for the optimal control parameters. The gradient descent algorithm is used to adjust the propulsion system power with a step size set to 0.01 to calculate the optimal power output and obtain the navigation execution instruction.

[0103] See also Figure 2In one embodiment, an autonomous obstacle avoidance system for an underwater robot based on fuzzy control also provides specific settings of a navigation adjustment module, including the following:

[0104] Deviation correction submodule: Based on the navigation execution instructions, it extracts real-time navigation data, analyzes the deviation between the heading angle and the target heading, calculates the impact of navigation speed on the deviation, analyzes the lateral drift caused by uneven thrust during navigation, determines the longitudinal position change trend, screens the heading deviation mutation points, calculates the correction force required for heading adjustment, adjusts the propulsion torque distribution, corrects the heading control error, and obtains the deviation correction parameters;

[0105] Path adjustment submodule: Based on the deviation correction parameters, extract the adjusted navigation path data, analyze the connection relationship between path segments, calculate the distance change of the path adjustment, detect the curvature change of the path change point, calculate the impact of the adjusted path on the speed, optimize the connection sequence of path segments, detect the path continuity and eliminate unstable routes, calculate the correction amplitude of the navigation path, adjust the navigation trajectory, and obtain the adjusted navigation path;

[0106] Waypoint fine-tuning submodule: Based on adjusting the navigation path, analyze the waypoint coordinate offset, calculate the impact of waypoint adjustment on path smoothness, measure the distance change between waypoints, screen the waypoint area that needs to be adjusted, calculate the waypoint fine-tuning range, adjust the waypoint position to match the navigation trajectory, correct the target positioning error, optimize the navigation direction, and obtain the navigation correction parameters.

[0107] Specifically, the deviation correction submodule extracts real-time navigation data based on the navigation execution instruction, uses the inertial navigation system (INS) with a sampling frequency of 100Hz to analyze the heading, velocity, and acceleration information, analyzes the deviation between the heading angle and the target heading, and uses the heading angle error calculation method to calculate the deviation of the current heading angle relative to the target heading. The influence of navigation speed on the deviation is calculated, and the velocity error propagation method is used to analyze the speed variation trend under different heading deviations. The lateral drift caused by uneven thrust during navigation is analyzed, and the lateral drift amplitude is detected based on the thrust difference. The longitudinal position variation trend is measured, and the time series analysis method is used to set the window size to 20 frames to analyze the longitudinal position variation trend over time. The heading deviation mutation points are screened, and the Canny edge detection is used to set high and low thresholds to detect the mutation points. The correction force required for heading adjustment is calculated, and the proportional, integral, and differential gain parameters are set by the control algorithm to calculate the correction force required for heading adjustment. The propulsion torque distribution is adjusted, and the fuzzy logic control is used to set the fuzzy set based on the rule base to calculate the propulsion torque adjustment strategy. The heading control error is corrected, and the Kalman filter is used to set the state transfer matrix and the observation noise variance for error correction to obtain the deviation correction parameters.

[0108] Path adjustment submodule: Based on the deviation correction parameters, the adjusted navigation path data is extracted and the number of control points is set by Bezier curve fitting to generate the adjusted path. The connection relationship between the path segments is analyzed and the Dijkstra shortest path algorithm is used to set the path smoothness as the weight to calculate the path connectivity. The distance change of the path adjustment is calculated and the path distance calculation method is used to analyze the change in path length before and after the adjustment. The curvature change of the path change point is detected and the curvature change detection method is used to set the calculation step size to analyze the path curvature change. The impact of the adjusted path on the speed is calculated and the speed change analysis method is used to analyze the impact of the path adjustment on the navigation speed. The path segment connection sequence is optimized and the dynamic programming algorithm is used to set the minimum adjustment cost to optimize the path segment sorting. The path continuity is detected and unstable routes are eliminated. The neighborhood radius and the minimum number of samples are set to screen stable paths. The correction amplitude of the navigation path is calculated and the mean square error calculation method is used to set the time window to analyze the path adjustment error. The navigation trajectory is adjusted and the particle filter algorithm is used to set the number of particles and the resampling threshold to optimize the trajectory smoothness to obtain the adjusted navigation path.

[0109] Waypoint fine-tuning submodule: Based on the adjustment of the navigation path, the waypoint coordinate offset is analyzed, and the least squares trajectory fitting is used to set the data window to calculate the waypoint offset. The impact of the waypoint adjustment on the path smoothness is calculated, and the Catmull-Rom spline interpolation is used to set the interpolation node to analyze the path smoothness. The spacing change between waypoints is measured, and the equidistant interpolation method is used to set the node spacing to calculate the spacing difference before and after adjustment. The waypoint area that needs to be adjusted is screened, and the K-Means clustering is used to set the number of clusters to screen the waypoint adjustment area. The waypoint fine-tuning range is calculated, and the coordinate distance calculation method is used to analyze the adjustment range. The waypoint position is adjusted to match the navigation trajectory, and the dynamic time warping (DTW) is used to set the window width to optimize the waypoint position. The target positioning error is corrected, and the extended Kalman filter is used to set the observation noise parameter to calculate the error correction amount, optimize the navigation direction, and control the rule-based adjustment gain parameter to calculate the heading fine-tuning and obtain the navigation correction parameter.

[0110] The above description is merely a preferred embodiment of the fuzzy-controlled autonomous obstacle avoidance system for an underwater robot disclosed herein and is not intended to limit the scope of protection of the embodiments herein. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the embodiments herein shall be included within the scope of protection of the embodiments herein.

[0111] It should be understood that, although the various module steps in the above-mentioned flow chart are shown in sequence according to the instructions of the arrows, these steps are not necessarily performed in sequence according to the order indicated by the arrows. Unless otherwise specified herein, the execution of these steps is not strictly limited in order, and these steps can be performed in other orders. Moreover, at least a portion of the steps in the above-mentioned flow chart may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily performed at the same time, but can be performed at different times, and the execution order of these sub-steps or stages is not necessarily performed in sequence, but can be performed in turn or alternately with at least a portion of other steps or sub-steps or stages of other steps.

Claims

1. An autonomous obstacle avoidance system for an underwater robot based on fuzzy control, characterized in that: include: Environmental perception module: acquires underwater sonar data, optical images, and inertial navigation information, analyzes echo signals, extracts target positions and corrects boundaries, performs image analysis, segments obstacle contours, calculates inertial data, corrects obstacle relative positions, and obtains dynamic obstacle information; State modeling module: Based on the dynamic information of the obstacles, the fuzzy K-nearest neighbor algorithm is used to perform terrain modeling, segment the terrain types to cluster targets and classify moving targets, and perform fuzzy weight assignment, calculate environmental impact factors, analyze trajectories, predict target states, and obtain environmental state characteristics; Obstacle avoidance strategy calculation module: Based on the environmental state characteristics, the Dijkstra algorithm is used to delineate the passage area, determine the feasible area, perform path interference calculation, analyze path conflicts, perform risk classification, evaluate the obstacle avoidance cost, and obtain the obstacle avoidance path weight matrix; Path optimization module: Based on the obstacle avoidance path weight matrix, it performs path evaluation, calculates comprehensive scores, calculates energy consumption, optimizes propulsion power, selects paths, adjusts path smoothness, and obtains an optimized navigation path; Real-time control module: Based on the optimized navigation path, it calculates control variables, adjusts propulsion torque, makes course corrections, compensates for navigation deviations, adjusts the propulsion system, matches power requirements, and obtains navigation execution instructions; Navigation adjustment module: Based on the navigation execution instructions, it performs deviation correction, adjusts the navigation path, suppresses errors, optimizes navigation stability and fine-tunes waypoints, adjusts target positions, and obtains navigation correction parameters.

2. The autonomous obstacle avoidance system of an underwater robot based on fuzzy control according to claim 1, characterized in that: The environment perception module includes: Target positioning submodule: This module collects underwater sonar data, optical images, and inertial navigation information through sensors, collects and filters echo signals, separates background noise, extracts target signals, calculates signal propagation delay, performs beamforming processing, analyzes echo energy distribution, determines target reflection areas, analyzes echo signal characteristics, matches target geometric parameters, corrects signal interference errors, and calculates target spatial coordinates. Obstacle parsing submodule: Based on the target spatial coordinates, it performs denoising and contrast enhancement on the optical image data, separates the background area, extracts the target contour features, performs edge detection and morphological analysis, segments the obstacle area, extracts the boundary point set, matches the structural parameters of the obstacle, calculates the connectivity of the target morphology, corrects the recognition error, and obtains the obstacle morphological contour; Dynamic calculation submodule: Based on the obstacle shape profile, it interpolates the inertial navigation information, compensates for sensor drift errors, analyzes the angular velocity and acceleration information in the inertial data, calculates the obstacle's motion trajectory, corrects the obstacle's relative position, associates the obstacle's historical motion state, calculates the trajectory change trend, establishes an obstacle motion parameter model, and obtains the obstacle's dynamic information.

3. The autonomous obstacle avoidance system of an underwater robot based on fuzzy control according to claim 1, characterized in that: The state modeling module includes: Terrain modeling submodule: Based on the dynamic information of the obstacles, collect and analyze terrain data, use ranging information to calculate regional depth, remove noise points and perform gradient analysis, identify terrain boundaries, calculate the terrain height change rate and screen steep areas, use the fuzzy K-nearest neighbor algorithm to classify terrain data, calculate the membership of sample points to different terrain types, perform terrain connectivity analysis, identify transition areas, calculate the terrain surface curvature and separate different terrain types, establish regional topological relationships, and obtain terrain structural characteristics; Target classification submodule: Based on the terrain structure characteristics, it screens target data, separates dynamic targets from static background, calculates the target's speed change rate and direction angle offset, extracts target motion trajectories and performs trajectory similarity analysis, clusters targets with different motion patterns, calculates target behavioral stability and performs category matching, assigns fuzzy weights, calculates environmental impact factors, and obtains target motion categories; State prediction submodule: Based on the target motion category, it traces back historical trajectory data, analyzes the change trend of the target position over time, calculates the target path deviation rate, identifies trajectory intersections and performs motion conflict analysis, establishes a motion state transition matrix, calculates the target's future motion trend and associates it with environmental influencing factors, predicts the target's subsequent state, and obtains environmental state characteristics.

4. The autonomous obstacle avoidance system of an underwater robot based on fuzzy control according to claim 3, characterized in that: The fuzzy K-nearest neighbor algorithm has the following formula: Where: u ij is the membership degree of terrain category j in the underwater robot environment, d ij is the Euclidean distance between the current position of the underwater robot and the centroid of terrain category j, d ik is the Euclidean distance between the current position of the underwater robot and the centroid of terrain category k, c is the total number of terrain categories in the environment, m is the fuzzy index, v ij is the terrain curvature change rate of the underwater robot at terrain category j, g ij is the gradient value of the underwater robot at terrain category j, h ij is the terrain height change of the underwater robot at terrain category j, w1, w2, w3 are weight coefficients, w n is the normalized weight, E n is information entropy, p n is the normalized probability distribution of parameter n.

5. The autonomous obstacle avoidance system of an underwater robot based on fuzzy control according to claim 1, characterized in that: The obstacle avoidance strategy calculation module includes: The passage area delineation submodule: Based on the environmental state characteristics, the Dijkstra algorithm is used to analyze the underwater terrain data, extract the boundary point set of fixed obstacles, calculate the obstacle height change, screen the steep areas and mark the inaccessible areas, combine the dynamic obstacle trajectory data, calculate the target activity range, perform regional connectivity detection, eliminate isolated and invalid areas, build a passage network model, delineate the passage range, and obtain the feasible navigation area; Path interference calculation submodule: Based on the feasible navigation area, it screens feasible paths, calculates the intersection points between path segments, analyzes the relative movement speed of obstacles, calculates the occupancy of navigation paths at different times, analyzes the time window of potential conflict points, calculates the interference intensity of navigation paths, screens high-risk conflict areas, adjusts path priorities, marks the degree of interference on paths, and obtains path conflict distribution; Obstacle avoidance cost evaluation submodule: Based on the path conflict distribution, calculate the energy consumption change required for path adjustment, analyze the impact of obstacle avoidance detour on navigation distance, analyze the time consumption of path adjustment, calculate the navigation stability of the obstacle avoidance path, extract the stability parameters of different obstacle avoidance schemes, construct a weight scoring system for obstacle avoidance adjustment schemes, screen path adjustment schemes, and obtain the obstacle avoidance path weight matrix.

6. The autonomous obstacle avoidance system of an underwater robot based on fuzzy control according to claim 5, characterized in that: The Dijkstra algorithm formula is as follows: C(n)=min(C(m)+p1·P(m,n)+p2·H(n)+p3·G(n)+p4·F(n)) Where: C(n) is the shortest path cost of the underwater robot from the starting position to the current node n, C(m) is the shortest path cost of the underwater robot from the starting position to the previous node m, P(m,n) is the basic path weight of the underwater robot from node m to node n, H(n) is the obstacle height change at the current node n, G(n) is the steep area gradient at the current node n, F(n) is the dynamic obstacle trajectory influence factor at the current node n, p1 is the adjustment coefficient of the path basic weight, p2 is the weight coefficient of the obstacle height change, p3 is the weight coefficient of the steep area gradient, p4 is the weight coefficient of the dynamic obstacle trajectory influence factor, E q is the information entropy of underwater environment characteristic data, r q is the normalized probability distribution of feature data, p q is the calculated feature weight value.

7. The autonomous obstacle avoidance system of an underwater robot based on fuzzy control according to claim 1, characterized in that: The path optimization module includes: Path scoring submodule: Based on the obstacle avoidance path weight matrix, it screens available path data, analyzes path traffic conditions, extracts obstacle distribution within the path, calculates path risk level, measures the density of path interference areas, analyzes the impact of path adjustment on navigation time, calculates connectivity between navigation paths, evaluates the stability of path changes, screens path traffic priority, and obtains path scoring results; Propulsion optimization submodule: Based on the path scoring results, calculate the hydrodynamic parameters of the path segment, analyze the impact of fluid resistance on propulsion efficiency, extract the depth changes along the path, calculate the navigation propulsion power requirement, analyze the adaptability of the propulsion system to different paths, adjust the propulsion torque distribution, optimize the navigation propulsion power, match the path energy consumption requirements, and obtain optimized propulsion parameters; Smoothing adjustment submodule: Based on the optimized propulsion parameters, analyze the path curvature changes, screen the low curvature path segments, calculate the heading adjustment angle, adjust the position of the heading change point, optimize the smoothness of the connection between path segments, eliminate the influence of mutation points, smooth the path curve, optimize the continuity of the navigation path, adjust the path node distribution, and obtain the optimized navigation path.

8. The autonomous obstacle avoidance system of an underwater robot based on fuzzy control according to claim 1, characterized in that: The real-time control module includes: Control variable calculation submodule: Based on the optimized navigation path, analyze the path heading change points, calculate the relative position of the navigation target point, analyze the path deviation trend, extract the navigation state parameters, calculate the navigation control variables, screen the key parameters for navigation adjustment, establish the navigation adjustment plan, and obtain the navigation control parameters; Heading correction submodule: Based on the navigation control parameters, it analyzes the deviation between the current heading and the target heading, calculates the heading angle adjustment amount, analyzes the impact of the heading correction on the propulsion force, selects the key points for heading adjustment, calculates the adjustment sequence of the heading change points, optimizes the navigation correction process, and obtains the heading correction instruction; Propulsion adjustment submodule: Based on the heading correction instruction, analyze the propulsion torque demand, calculate the propulsion force adjustment range, match the power distribution of different propulsion modes, analyze the impact of propulsion adjustment on navigation stability, screen the propulsion control scheme, adjust the propulsion system power, and obtain the navigation execution instruction.

9. The autonomous obstacle avoidance system of an underwater robot based on fuzzy control according to claim 1, characterized in that: The navigation adjustment module includes: Deviation correction submodule: Based on the navigation execution instruction, it extracts real-time navigation data, analyzes the deviation between the heading angle and the target heading, calculates the impact of navigation speed on the deviation, analyzes the lateral drift caused by uneven thrust during navigation, determines the longitudinal position change trend, selects the heading deviation mutation point, calculates the correction force required for heading adjustment, adjusts the propulsion torque distribution, corrects the heading control error, and obtains the deviation correction parameter; Path adjustment submodule: Based on the deviation correction parameters, extract the adjusted navigation path data, analyze the connection relationship between path segments, calculate the distance change of the path adjustment, detect the curvature change of the path change point, calculate the impact of the adjusted path on the speed, optimize the connection sequence of path segments, detect path continuity and eliminate unstable routes, calculate the correction amplitude of the navigation path, adjust the navigation trajectory, and obtain the adjusted navigation path; Waypoint fine-tuning submodule: Based on the adjusted navigation path, analyze the waypoint coordinate offset, calculate the impact of waypoint adjustment on path smoothness, measure the spacing change between waypoints, screen the waypoint area that needs to be adjusted, calculate the waypoint fine-tuning range, adjust the waypoint position to match the navigation trajectory, correct the target positioning error, optimize the navigation direction, and obtain the navigation correction parameters.

Images