Complex marine environment-oriented AUV (Autonomous Underwater Vehicle) motion planning and optimizing method and system

Abstract

The invention discloses an AUV (Autonomous Underwater Vehicle) motion planning and optimizing method and system for a complex marine environment, and the method comprises the steps: S1, environment perception and modeling: fusing collected multi-source environment data, and constructing a three-dimensional grid model; s2, initial path planning: performing initial path planning based on the three-dimensional grid model; s3, path optimization: constructing a multi-objective optimization function, and optimizing the initial path; s4, real-time dynamic adjustment: monitoring the change of the marine environment in real time, and triggering path re-planning when the change of the environment exceeds a set threshold value; and S5, energy management and motion control: constructing a dynamic energy consumption model, formulating a speed planning and attitude adjustment strategy of the AUV, and realizing energy and motion collaborative optimization through a reinforcement learning algorithm. According to the AUV motion planning and optimizing system, technologies such as environment perception, path planning, real-time adjustment and energy management are organically fused, all links are mutually coordinated and optimized, and a set of complete and efficient AUV motion planning and optimizing system is formed.

Description

Technical Field

[0001] This invention relates to the field of autonomous underwater vehicle technology, and in particular to a method and system for AUV motion planning and optimization in complex marine environments. Background Technology

[0002] With the deepening of ocean development and utilization, autonomous underwater vehicles (AUVs) are playing an increasingly crucial role as important tools for ocean exploration. AUVs need to perform various tasks in complex and ever-changing marine environments; however, these complex environments present numerous challenges to AUV motion planning.

[0003] First, the marine environment is characterized by complex hydrodynamic factors. Ocean currents exhibit spatiotemporal variations, with significant differences in speed and direction across different depths and regions. For example, in some strait areas, tidal currents can reach speeds of several knots, and their direction can change periodically over time. Furthermore, the presence of internal ocean waves also affects the movement of AUVs. The vertical and horizontal movements of water caused by internal waves can lead to deviations from the planned course and even affect their stability. When AUVs are performing long-duration, long-distance missions, the cumulative errors caused by hydrodynamic factors can prevent them from accurately reaching the target location, impacting mission completion.

[0004] Secondly, various obstacles exist in the ocean. In nearshore areas, there are numerous man-made obstacles such as fishing nets and shipwrecks; in deep-sea areas, the seabed topography is complex and varied, with natural obstacles such as seamounts, trenches, and reefs. During operation, AUVs need to perceive the location and shape of these obstacles in real time and plan a safe navigation path. Traditional sensors, such as sonar, can detect obstacles, but they suffer from limited detection range and complex data processing. Moreover, factors such as water turbidity and acoustic signal attenuation in the marine environment further reduce the detection accuracy and reliability of sensors.

[0005] Furthermore, the energy limitations of AUVs themselves are also a significant issue. AUVs typically carry limited battery power, and in complex marine environments, they may need to frequently adjust their attitude and speed to avoid obstacles and cope with hydrodynamic disturbances. This leads to increased energy consumption, thereby shortening the AUV's operating time and range. Therefore, how to balance path safety, timeliness, and energy consumption to achieve optimal motion planning during the motion planning process is a problem that urgently needs to be solved.

[0006] Currently, existing AUV motion planning methods have certain limitations when dealing with complex marine environments. Some traditional path planning algorithms, such as Dijkstra's algorithm and A* algorithm, while capable of planning optimal paths in static environments, struggle to adapt quickly to dynamic marine environments due to high computational complexity and poor real-time performance. Therefore, there is an urgent need for a method that can effectively address complex marine environments and achieve efficient, safe, and energy-saving AUV motion planning. Summary of the Invention

[0007] To address the above problems, the present invention provides an AUV motion planning and optimization method and system for complex marine environments, thereby solving the problems of poor real-time performance, easy getting trapped in local optima, and high energy consumption in the existing AUV motion planning in complex marine environments.

[0008] According to a first aspect of the present invention, an AUV motion planning and optimization method for complex marine environments is provided, comprising: S1. Environmental perception and modeling: Collect multi-source environmental data, fuse the collected multi-source environmental data through extended Kalman filter algorithm, and construct a three-dimensional mesh model including obstacle area, safe navigation area and hydrodynamic parameters. S2. Initial Path Planning: Based on the 3D mesh model, an improved fast expanding random tree algorithm is used, and the target gravity function and obstacle repulsion function are introduced to perform initial path planning; S3. Path Optimization: Based on the initial path planning, a multi-objective optimization function is constructed, and a path optimization method based on simulated annealing algorithm is used to optimize the initial path; S4. Real-time dynamic adjustment: By constructing obstacle change detection index, hydrodynamic change detection index and comprehensive environmental change index, the marine environment changes are monitored in real time. When the environmental change exceeds the set threshold, path replanning based on incremental fast expansion random tree algorithm is triggered. S5. Energy Management and Motion Control: Construct a dynamic energy consumption model, formulate speed planning and attitude adjustment strategies for AUVs based on the model, and achieve coordinated optimization of energy and motion through reinforcement learning algorithms.

[0009] In the above scheme, the multi-source data collected in step S1 includes: point cloud data of obstacles, attitude data of AUV, angular velocity of AUV, linear acceleration of AUV, current depth of AUV, seawater temperature, seawater salinity and electrical conductivity of seawater.

[0010] In the above scheme, in step S1, by combining seawater temperature, seawater salinity and seawater conductivity, an empirical orthogonal function analysis method is used to establish a spatiotemporal distribution model of hydrodynamic parameters and label it into a three-dimensional mesh model.

[0011] In the above scheme, in step S2, the target gravity function and obstacle repulsion function are combined with the spatiotemporal distribution model of hydrodynamic parameters to perform initial path planning; The target gravity function guides the growth of the random tree based on the distance and direction between the AUV and the target point, while the obstacle repulsion function adjusts the growth direction of the random tree based on the distance between the obstacle and the random tree node.

[0012] In the above scheme, the multi-objective optimization function in step S3 includes: path length, path curvature, and energy consumption; The path optimization method based on simulated annealing algorithm is adopted. The optimization of the initial path includes: generating new path solutions by randomly perturbing the path nodes; calculating the difference between the multi-objective optimization function of the new path and the multi-objective optimization function of the initial path; determining whether to accept the new path solution based on the difference of functions according to the Metropolis criterion; if accepted, the new path solution is used as the initial path for the next round; if not accepted, the current initial path is retained for the next round of optimization, until the optimization is completed.

[0013] In the above scheme, in step S4, the obstacle change detection index is calculated based on Hausdorff distance; the hydrodynamic change detection index is constructed by combining the AUV's own motion state; and the comprehensive environmental change index is constructed by combining the obstacle change detection index and the hydrodynamic change detection index.

[0014] In the above scheme, during step S4, parallel computing technology is used to synchronously expand multiple branches of the random tree during path replanning.

[0015] In the above scheme, step S5 includes the dynamic energy consumption model, which includes the thruster power model, the attitude adjustment mechanism power model, and the sensor system power model.

[0016] In the above scheme, in step S5, the reinforcement learning algorithm uses the energy state, position, velocity, attitude and environmental information of the AUV as the state space, and velocity adjustment and attitude adjustment as the action space, and learns the optimal energy and motion control strategy by designing a reward function.

[0017] According to a second aspect of the present invention, an AUV motion planning and optimization system for complex marine environments is provided, the system being used to implement the method described in any one of the above solutions, the system comprising: Environmental perception and modeling module: used to collect multi-source environmental data, and fuse the collected multi-source environmental data through extended Kalman filter algorithm to construct a three-dimensional mesh model including obstacle area, safe navigation area and hydrodynamic parameters; Initial path planning module: Based on a 3D mesh model, it uses an improved fast expanding random tree algorithm, introduces the target gravity function and obstacle repulsion function, and combines the hydrodynamic model to calculate the resultant force for initial path planning; Path optimization module: Used to construct a multi-objective optimization function and optimize the initial path using a path optimization method based on simulated annealing algorithm; Real-time dynamic adjustment module: It is used to monitor changes in the marine environment in real time by constructing obstacle change detection index, hydrodynamic change detection index and comprehensive environmental change index. When the environmental change exceeds the set threshold, it triggers path replanning based on incremental fast expansion random tree algorithm. Energy Management and Motion Control Module: This module is used to build a dynamic energy consumption model, formulate speed planning and attitude adjustment strategies for the AUV based on the model, and achieve coordinated optimization of energy and motion through reinforcement learning algorithms.

[0018] The beneficial effects of this invention are: In the environmental perception and modeling stage, multi-source heterogeneous sensor fusion technology, combined with the extended Kalman filter algorithm, enabled high-precision acquisition and processing of information such as obstacles and hydrodynamics in complex marine environments. The constructed 3D mesh model can identify obstacles and safe zones, and an empirical orthogonal function analysis was used to establish a spatiotemporal distribution model of hydrodynamics, predicting hydrodynamic parameters in different regions at different times. This provides a reliable data foundation for AUV motion planning, and compared with traditional methods, it captures environmental information more comprehensively and accurately, improving the AUV's adaptability to complex marine environments.

[0019] The initial path planning adopts an improved fast expanding random tree algorithm, which introduces the target gravity function, obstacle repulsion function, and considers the hydrodynamic influence. This solves the problems of blind search and slow convergence in the traditional fast expanding random tree algorithm, enabling the random tree to grow quickly toward the target and avoid obstacles. It also fully considers the hydrodynamic influence of the marine environment, and the planned path is more in line with actual needs, improving the efficiency and rationality of path planning.

[0020] In the path optimization phase, a multi-objective optimization model incorporating path length, curvature, and energy consumption is constructed based on the simulated annealing algorithm. By weighted combination of the comprehensive objective functions, a holistic optimization of the path is achieved, overcoming the limitations of traditional single-objective path planning. During the optimization process, random perturbations and the Metropolis criterion are used to avoid getting trapped in local optima, achieving efficient, smooth, and energy-saving paths, thereby improving the AUV's navigation performance.

[0021] In the real-time dynamic adjustment section, this invention proposes a mechanism based on multi-source data fusion and dynamic threshold determination. By constructing obstacle change detection indicators, hydrodynamic change detection indicators, and a comprehensive environmental change indicator, Hausdorff distance is used to calculate obstacle changes, and hydrodynamic change indicators are constructed in conjunction with the AUV's motion state. Environmental changes are determined through dynamic thresholds. When environmental changes exceed the threshold, path replanning based on an incremental fast-expanding random tree algorithm combined with parallel computing is triggered, enabling the AUV to quickly respond to dynamic changes in the complex marine environment. Compared to traditional fixed-path planning methods, this enhances the safety and reliability of AUVs in complex and dynamic marine environments.

[0022] In terms of energy management and motion control, a dynamic energy consumption model encompassing the thrusters, attitude adjustment mechanisms, and sensor systems was constructed. This model fully considers the impact of various factors, such as the AUV's motion state and environmental hydrodynamics, on energy consumption, achieving high accuracy. Based on the motion control strategy formulated using this model, energy efficiency is achieved while meeting mission requirements through speed planning and attitude adjustment optimization. Furthermore, reinforcement learning algorithms are introduced to establish an energy and motion co-optimization mechanism, enabling the AUV to autonomously learn the optimal control strategy under different environmental and mission conditions. This solves the problems of energy waste and inflexible motion control in traditional methods, extends the AUV's endurance and working range, and enhances its ability to perform missions in complex marine environments.

[0023] This invention organically integrates technologies such as environmental perception, path planning, real-time adjustment, and energy management, with each component working in synergy and mutual optimization to form a complete and efficient AUV motion planning and optimization system. Compared to existing single or fragmented technical solutions, this invention comprehensively improves the overall performance of AUVs in complex marine environments, enabling its application in various fields such as marine resource exploration, underwater environmental monitoring, and seabed facility inspection, demonstrating significant theoretical importance and promising application prospects. Attached Figure Description

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

[0025] Figure 1 This is a flowchart of the AUV motion planning and optimization method for complex marine environments according to the present invention.

[0026] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0027] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this disclosure. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this disclosure as detailed in the appended claims.

[0028] The terms "first," "second," etc., used in this disclosure are for distinguishing similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such use of data can be interchanged where appropriate so that the embodiments of this disclosure described herein can be implemented, for example, in orders other than those illustrated or described herein.

[0029] Furthermore, the terms “comprising” and “having”, and any variations thereof, are intended to cover non-exclusive inclusion, such that a process, method, system, product, or apparatus that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or units that are not explicitly listed or that are inherent to such process, method, product, or apparatus.

[0030] Multiple, including two or more.

[0031] And / or, it should be understood that, for the purposes of this disclosure, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent three situations: A exists alone, A and B exist simultaneously, and B exists alone.

[0032] like Figure 1 As shown, according to a first aspect embodiment of the present invention, an AUV motion planning and optimization method for complex marine environments is provided, comprising: S1. Environmental perception and modeling: Collect multi-source environmental data, fuse the collected multi-source environmental data through extended Kalman filter algorithm, and construct a three-dimensional mesh model including obstacle area, safe navigation area and hydrodynamic parameters. S2. Initial Path Planning: Based on the 3D mesh model, an improved fast expanding random tree algorithm is used, and the target gravity function and obstacle repulsion function are introduced to perform initial path planning; S3. Path Optimization: Based on the initial path planning, a multi-objective optimization function is constructed, and a path optimization method based on simulated annealing algorithm is used to optimize the initial path; S4. Real-time dynamic adjustment: By constructing obstacle change detection index, hydrodynamic change detection index and comprehensive environmental change index, the marine environment changes are monitored in real time. When the environmental change exceeds the set threshold, path replanning based on incremental fast expansion random tree algorithm is triggered. S5. Energy Management and Motion Control: Construct a dynamic energy consumption model, formulate speed planning and attitude adjustment strategies for AUVs based on the model, and achieve coordinated optimization of energy and motion through reinforcement learning algorithms.

[0033] Step S1 is as follows: The AUV is equipped with a multi-beam sonar, an inertial navigation system, and a multi-source heterogeneous sensor system including a temperature, salinity, and depth sensor. Each sensor collects data at a set frequency to obtain multi-source environmental data. Specifically, the multi-beam sonar collects obstacle point cloud data, the inertial navigation system outputs attitude, angular velocity, and linear acceleration information, and the depth sensor and temperature, salinity, and depth sensor collect depth, temperature, salinity, and conductivity data.

[0034] Specifically, multibeam sonar transmits broadband acoustic signals and uses the time delay and angle information of the echo signals to construct three-dimensional point cloud data of obstacles. ,in Indicating the first point cloud The three-dimensional coordinates of a point in the AUV coordinate system The number of points in the point cloud data.

[0035] The inertial navigation system outputs the AUV's attitude quaternion using high-frequency sampling. angular velocity and linear acceleration .

[0036] Depth sensor, temperature, salinity and depth sensor to measure the current depth of AUV in real time Collect seawater temperature and conductivity And based on the current depth Seawater temperature and conductivity Real-time salinity calculation .

[0037] Furthermore, to fuse multi-source sensor data, a fusion algorithm based on extended Kalman filtering is employed. Let the system state vector be... ,in For AUV location, Let be the AUV speed. The state transition equation is: (1) in, To take into account hydrodynamics and propulsion acceleration model, This is the transformation matrix from quaternions to rotation matrices, specifically in the form: (2) The observation equations are defined according to different sensors. For multibeam sonar, after transforming the point cloud data into a global coordinate system, the observation values ​​of the AUV's location are obtained by matching the data with the established map through an iterative nearest-point algorithm. The attitude and acceleration information output by the inertial navigation system is used as the observation value. The observations consist of data from depth sensors and temperature-salinity-depth sensors. The Extended Kalman Filter (EKF) algorithm modifies the state vector through two steps: prediction and update. Perform optimal estimation to achieve the fusion of multi-source data.

[0038] When constructing the 3D mesh model, the space surrounding the AUV is divided into areas of size [missing information]. The grid cells, each grid cell is represented as Its state is The status is determined based on multibeam sonar point cloud data: if point cloud data exists within the grid, then... (Obstacle area); otherwise (Safe zone). Simultaneously, combining data from temperature, salinity, and depth sensors, and utilizing empirical orthogonal function analysis methods, hydrodynamic parameters, such as ocean current velocity, are established. Spatiotemporal distribution model: (3) in, For time coefficient, For spatial mode functions, The number of modes selected.

[0039] Spatial mode functions were determined by principal component analysis of historical hydrodynamic data. and time coefficient The calculated hydrodynamic parameters of each grid cell are labeled into a 3D mesh model. This model allows for the prediction of hydrodynamic parameters of different grid cells at different times, providing accurate environmental information for subsequent path planning.

[0040] Step S1 utilizes multi-source heterogeneous sensor fusion technology combined with an extended Kalman filter algorithm to achieve high-precision acquisition and processing of information on obstacles and hydrodynamics in complex marine environments. The constructed 3D mesh model can identify obstacles and safe zones, and an empirical orthogonal function analysis is used to establish a spatiotemporal distribution model of hydrodynamics, predicting hydrodynamic parameters in different regions at different times. This provides a reliable data foundation for AUV motion planning, and compared to traditional methods, it captures environmental information more comprehensively and accurately, improving the AUV's adaptability to complex marine environments.

[0041] Step S2 is as follows: Set target gravity coefficient Obstacle repulsion coefficient Random tree expansion step size Maximum number of expansion nodes Let the current position of the AUV be... The target location is There is a set of obstacles in the AUV's movement space. ,in Indicates the first An obstacle, This represents the total number of obstacles.

[0042] To guide the random tree to grow rapidly toward the target point, a target gravity function is defined. for: (4) in, The gravitational coefficient is a positive parameter set according to the AUV's performance and mission requirements, used to adjust the strength of the target's gravity. This is a unit vector pointing from the current position to the target position, ensuring that the direction of gravity always points towards the target point. This function causes nodes in the random tree to tend to move closer to the target point during expansion, thus accelerating the algorithm's convergence speed.

[0043] To prevent random tree nodes from entering obstacle areas, an obstacle repulsion function is defined. For each obstacle First calculate the current position of the AUV. To the obstacle shortest distance point It can be achieved by solving optimization problems. Obtained. Then the obstacle. The repulsive force on the AUV is: (5) in, is the repulsion coefficient, a positive parameter used to control the magnitude of the repulsion force; This causes the magnitude of the repulsive force to be inversely proportional to the distance, with the repulsive force increasing as the distance decreases. When multiple obstacles exist, the total repulsive force from the obstacles is... .

[0044] Since ocean currents and other hydrodynamic forces affect the movement of AUVs, let the hydrodynamic velocity vector at the current position be... Then the net force on the AUV for: (6) in, For the quality of AUV, The term represents the hydrodynamic force on the AUV. By incorporating hydrodynamic factors into the random tree expansion decision, the planned path is made to better reflect the motion characteristics of the AUV in the actual marine environment.

[0045] During the random tree expansion process, at each iteration: (1) Randomly generate a sampling point within the AUV's motion space. .

[0046] (2) Calculate the relationship between the random tree and the random tree. nearest node The KD-Tree data structure is used to accelerate nearest neighbor search.

[0047] (3) Calculate the resultant force ,in For AUV quality, the hydrodynamic velocity at the current location is obtained based on the 3D mesh model. .

[0048] (4) Along Direction, by step from Expand to generate new nodes ,Right now .

[0049] (5)Judgment Whether it is in an obstacle area: when the corresponding grid cell state If the obstacle is not present and the AUV speed and acceleration limits are met, then it will be regenerated. Add to the random tree and establish a connection with... The connecting edge.

[0050] (6) Repeat the above steps until the random tree node enters a certain radius of the target location to generate the initial path.

[0051] In complex marine environments, the traditional fast expanding random tree algorithm for AUV path planning suffers from problems such as high search blindness, slow convergence speed, and difficulty in adapting to dynamic hydrodynamic disturbances and complex obstacle distributions. Step S2 of this invention improves the fast expanding random tree algorithm by introducing a target gravity function and an obstacle repulsion function, and combining them with a marine hydrodynamic model. This solves the problems of blind search and slow convergence in the traditional fast expanding random tree algorithm, enabling the random tree to grow rapidly towards the target and avoid obstacles, resulting in a planned path that better meets actual needs.

[0052] Step S3 is as follows: S31. Construction of the objective function Let the initial path consist of a series of nodes, denoted as . ,in Indicates the first path The three-dimensional coordinates of each node in the global coordinate system This represents the total number of path nodes.

[0053] Path length is an important indicator of path quality; a shorter path reduces AUV travel time and potential risks. Define the objective function for path length. for: (7) This formula calculates and sums the Euclidean distances between adjacent nodes to obtain the total length of the entire path.

[0054] AUVs consume more energy and experience reduced stability when turning; a smooth path reduces curvature changes and improves navigation efficiency. To quantify the smoothness of the path, a path curvature objective function is defined. First, calculate the angle between three adjacent path segments. Using the dot product formula: (8) in, , The objective function for path curvature is then: The summation of the angles between each path segment reflects the overall curvature of the path; the smaller the value, the smoother the path.

[0055] Considering the impact of hydrodynamics and AUV motion states on energy consumption in complex marine environments, an energy consumption model is established. Assume the AUV is at path nodes... The velocity at that point is The attitude adjustment angle is The hydrodynamic influence coefficient is The thruster power model is The power model of the attitude adjustment mechanism is The sensor's operating power is Travel time between adjacent nodes Then the energy consumption objective function The sum of propulsion energy consumption for all segments, attitude adjustment energy consumption for all nodes, and sensor energy consumption throughout the entire flight: (9) By weighting and combining the three objective functions mentioned above, we obtain the comprehensive objective function. : (10) Among them, path length weight Path curvature weight Energy consumption weight ,and The weights are adjusted according to different task requirements to achieve targeted optimization of the path.

[0056] S32. Simulated Annealing Algorithm Optimization Process Initial temperature Cooling coefficient Maximum number of iterations , the initial path As the current solution Calculate its comprehensive objective function value (11) Each iteration includes the following steps: (1) For the current path Perform random perturbation to generate a new path The perturbation method can be to randomly move the position of path nodes within a certain range, insert or delete nodes, etc. (2) Calculate the new path The comprehensive objective function value And calculate the difference in objective function values. ; (3) According to the Metropolis criterion, with probability Decide whether to accept the new solution. If so, then the new solution will be accepted directly. , ;like Then according to probability Accept new solutions. For example, generate a random number between [0, 1]. ,like If the new solution is found, then accept the new solution; otherwise, retain the current solution. (4) Lower the temperature. ; (5) Determine whether the maximum number of iterations has been reached. If not, return to step (1) to continue iterating. If it has been reached, stop the algorithm and output the final optimized path.

[0057] In complex marine environments, initial paths often suffer from problems such as excessive length, frequent turns, and insufficient consideration of energy consumption, failing to meet the requirements for efficient and energy-saving operation of AUVs. Step S3 of this invention constructs a multi-objective optimization model based on simulated annealing, incorporating path length, curvature, and energy consumption. By weighted combination of the comprehensive objective functions, a comprehensive optimization of the path is achieved. Compared with traditional single-objective path planning methods, the path generated by this invention exhibits superior path length, smoothness, and energy consumption, thus improving the navigation efficiency of AUVs.

[0058] Step S4 is as follows: According to step S1, the environmental data collected in real time by the AUV includes: obstacle point cloud data detected by multibeam sonar. ,in, express Moment in the cloud The three-dimensional coordinates of the points The point cloud point count at that moment; the attitude quaternion output by the inertial navigation system. angular velocity and linear acceleration Temperature collected by the temperature-salinity depth sensor ,salinity and conductivity At the same time, record the environmental data from the previous moment as... , , , , , , .

[0059] S41. Construction of Environmental Change Monitoring Indicators To determine whether new obstacles have appeared or the positions of existing obstacles have changed, the difference between the obstacle point cloud data at the current time and the previous time is calculated. Using the Hausdorff distance-based algorithm, two point sets are first defined. and One-way Hausdorf distance: (12) The bidirectional Hausdorff distance is: (13) To account for differences in point cloud density, a weighting coefficient is introduced. The degree of difference in obstacle changes is obtained: (14) in, Adaptive adjustments are made based on sensor accuracy and historical data statistics. Exceeding the obstacle change threshold When the obstacle environment changes, it is determined that the obstacle environment has changed.

[0060] Based on temperature, salinity, and depth sensor data and a spatiotemporal distribution model of hydrodynamics, the change in hydrodynamic velocity between the current moment and the previous moment is calculated. Let the hydrodynamic velocity vector at the current moment be... The previous moment was ,but: (15) Incorporating the AUV's own motion state, a velocity change sensitivity coefficient is introduced. Construct a comprehensive index of hydrodynamic changes: (16) when Exceeding the hydrodynamic change threshold At that time, it was determined that the hydrodynamic environment had changed.

[0061] S42, Path Replanning Trigger Mechanism Overall obstacle variation degree Comprehensive indicators of hydrodynamic changes Define comprehensive indicators of environmental change : (17) Among them, obstacle change weight Hydrodynamic change weight It is dynamically adjusted based on task priority. When hour, The overall threshold triggers a path replanning mechanism.

[0062] S43, Fast Path Replanning Algorithm After triggering path replanning, an improved algorithm based on an incrementally expanding random tree is employed. This is based on the current AUV position. As the new starting point, the target location Keeping the historical path planning information and the existing environmental model unchanged, the expansion region of the random tree is limited to: (18) in, The maximum extended radius is determined based on the AUV's power performance and remaining energy. This represents a newly detected obstacle region. During the expansion process, a new path is rapidly generated by combining the target's gravitational function, the obstacle's repulsive function, and hydrodynamic influences. Furthermore, to further improve efficiency, parallel computing techniques are employed to simultaneously expand multiple branches of the random tree, shortening the path generation time.

[0063] In complex marine environments, sudden changes in ocean current speeds and the emergence of new obstacles occur frequently, making traditional fixed-path planning insufficient to meet the safe and efficient operation requirements of AUVs. Step S4 of this invention constructs obstacle change detection indicators, hydrodynamic change detection indicators, and a comprehensive environmental change indicator to quickly and accurately perceive dynamic changes in the marine environment. When environmental changes exceed a set threshold, an improved algorithm based on incrementally rapidly expanding random trees is triggered, and parallel computing technology is used to achieve rapid path replanning. This enables AUVs to react quickly and adjust their paths promptly in the face of sudden environmental changes, enhancing the safety and reliability of AUVs in complex and dynamic marine environments compared to traditional fixed-path planning methods.

[0064] Step S5 is as follows: S51, Energy Consumption Model Construction The energy consumption of an AUV mainly comes from the operation of the thrusters, the operation of the attitude adjustment mechanism, and the operation of the sensor system. Let the AUV at time... The energy consumption power is ,but This can be expressed as thruster power. Power of attitude adjustment mechanism With sensor system power The sum is: (19) The thruster power is closely related to the AUV's speed and hydrodynamic drag. Considering factors such as ocean currents and eddies in the marine environment, let the AUV's velocity vector relative to the seawater be... Its speed is set by the AUV itself. With environmental hydrodynamic velocity Synthesis, i.e. .

[0065] According to the principles of fluid mechanics, the power required for a thruster to overcome hydrodynamic resistance It can be represented as: (20) in, The density of seawater can be calculated from data from temperature, salinity, and depth sensors combined with the seawater state equation. The drag coefficient is related to factors such as the shape and surface roughness of the AUV and is determined in advance through wind tunnel experiments or numerical simulations. The projected area of ​​the AUV perpendicular to the direction of motion.

[0066] The attitude adjustment mechanism is used to change the yaw, pitch, and roll attitude of the AUV. Let the attitude adjustment angle vector be... , respectively representing the circumference , , Rotation angle of the shaft. Power of the attitude adjustment mechanism. Related to the adjustment angle, adjustment speed, and the mechanism's own parameters, the following model can be established: (twenty one) in, Adjust the angular velocity vector for attitude control. , The coefficients related to the characteristics of the attitude adjustment mechanism were determined by fitting experimental data.

[0067] Sensor system power It mainly depends on the operating mode and number of sensors, assuming the AUV is equipped with... The first type of sensor, the first Such sensors at all times The operating power is ,but: (twenty two) S52, Motion Control Strategy Formulation Based on the optimized path and energy consumption model, a motion control strategy is formulated to achieve energy-efficient utilization of the AUV while meeting mission requirements. According to the path planning results, the path is divided into several segments, each with a length of [missing information]. The estimated time to pass is For each path segment, the objective function is optimized by combining hydrodynamic conditions and energy consumption models. Determine the optimal speed : (twenty three) in, , This is a weighting factor used to balance navigation efficiency and energy consumption. It can be increased when the mission has high time requirements. When energy is limited, increase By finding the minimum value of the objective function, the optimal speed on each path segment can be obtained.

[0068] To reduce energy consumption during path turning or obstacle avoidance, an attitude adjustment optimization model is established based on path curvature and obstacle distribution. Let the curvature of the path at a certain point be... The target attitude is The current posture is Then the attitude adjustment optimization objective function for: (twenty four) in, , The weighting coefficients are used to determine the optimal attitude adjustment angle and speed scheme by solving for the minimum value of the objective function, so that the AUV can reduce the energy consumption of the attitude adjustment mechanism while meeting the path requirements.

[0069] S53, Energy and Motion Co-optimization Mechanism To achieve synergy between energy management and motion control, a reinforcement learning algorithm is introduced. The energy state, position, velocity, attitude, and environmental information of the AUV are used as the state space. The speed adjustment, attitude adjustment and other control commands are used as the motion space. Design reward function as follows: (25) in, This represents the distance an AUV travels toward a target per unit of time, reflecting the progress of the mission. Energy consumption per unit of time. , This is the reward weighting coefficient.

[0070] Reinforcement learning algorithms learn optimal energy and motion control strategies by continuously interacting with the environment and based on a reward function. In each decision, the algorithm learns from the action space... Choose one action Execute, observe the new state s' and reward from environmental feedback. And update the Q-value function according to the Q-learning algorithm: (26) in, This serves as a discount factor to balance current and future rewards. Through multiple iterations of learning, the AUV can autonomously select the optimal energy management and motion control strategies under different environmental and mission conditions, maximizing energy efficiency and mission execution capabilities.

[0071] In complex marine environments, AUVs face challenges such as hydrodynamic interference and frequent path adjustments, making it difficult for traditional energy management and motion control methods to achieve a balance between energy efficiency and mission efficiency. Step S5 of this invention constructs a dynamic energy consumption model encompassing the thrusters, attitude adjustment mechanisms, and sensor systems, based on the AUV's motion characteristics and marine environmental features. This model fully considers the impact of various factors, including the AUV's motion state and environmental hydrodynamics, on energy consumption, achieving high accuracy. Based on the motion control strategy formulated using this model, energy efficiency is achieved through speed planning and attitude adjustment optimization, meeting mission requirements while maintaining high energy efficiency. Furthermore, a reinforcement learning algorithm is introduced to establish an energy and motion co-optimization mechanism, enabling the AUV to autonomously learn the optimal control strategy under different environmental and mission conditions. This solves the problems of energy waste and inflexible motion control inherent in traditional methods, extending the AUV's endurance and working range, and enhancing its ability to perform missions in complex marine environments.

[0072] According to a second aspect of the present invention, an AUV motion planning and optimization system for complex marine environments is provided, the system being used to implement the method described in any one of the above solutions, the system comprising: Environmental perception and modeling module: used to collect multi-source environmental data, and fuse the collected multi-source environmental data through extended Kalman filter algorithm to construct a three-dimensional mesh model including obstacle area, safe navigation area and hydrodynamic parameters; Initial path planning module: Based on a 3D mesh model, it uses an improved fast expanding random tree algorithm, introduces the target gravity function and obstacle repulsion function, and combines the hydrodynamic model to calculate the resultant force for initial path planning; Path optimization module: Used to construct a multi-objective optimization function and optimize the initial path using a path optimization method based on simulated annealing algorithm; Real-time dynamic adjustment module: It is used to monitor changes in the marine environment in real time by constructing obstacle change detection index, hydrodynamic change detection index and comprehensive environmental change index. When the environmental change exceeds the set threshold, it triggers path replanning based on incremental fast expansion random tree algorithm. Energy Management and Motion Control Module: This module is used to build a dynamic energy consumption model, formulate speed planning and attitude adjustment strategies for the AUV based on the model, and achieve coordinated optimization of energy and motion through reinforcement learning algorithms.

[0073] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0074] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0075] Through the above description of the embodiments, those skilled in the art can clearly understand that the above implementation methods can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes several instructions to cause a terminal (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0076] The embodiments of the present invention have been described above with reference to the accompanying drawings. However, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of the present invention without departing from the spirit and scope of the claims. All of these forms are within the protection scope of the present invention.

Claims

1. A method for AUV motion planning and optimization in complex marine environments, characterized in that, include: S1. Environmental perception and modeling: Collect multi-source environmental data, fuse the collected multi-source environmental data, and construct a three-dimensional mesh model that includes obstacle areas, safe navigation areas, and hydrodynamic parameters. S2. Initial Path Planning: Based on the 3D mesh model, an improved fast expanding random tree algorithm is used, and the target gravity function and obstacle repulsion function are introduced to perform initial path planning; S3. Path Optimization: Based on the initial path planning, a multi-objective optimization function is constructed, and a path optimization method based on simulated annealing algorithm is used to optimize the initial path; S4. Real-time dynamic adjustment: Based on the optimized path, the marine environment changes are monitored in real time by constructing environmental monitoring indicators. When the environmental changes exceed the set threshold, the optimized path is replanned based on the incremental fast expansion random tree algorithm. S5. Energy Management and Motion Control: Construct a dynamic energy consumption model, formulate speed planning and attitude adjustment strategies for AUVs based on the model, and achieve coordinated optimization of energy and motion through reinforcement learning algorithms.

2. The AUV motion planning and optimization method for complex marine environments according to claim 1, characterized in that, The multi-source data collected in step S1 includes: point cloud data of obstacles, attitude data of AUV, angular velocity of AUV, linear acceleration of AUV, current depth of AUV, seawater temperature, seawater salinity and seawater conductivity, and the collected multi-source environmental data are fused by extended Kalman filter algorithm.

3. The AUV motion planning and optimization method for complex marine environments according to claim 2, characterized in that, In step S1, by combining seawater temperature, seawater salinity and seawater conductivity, an empirical orthogonal function analysis method is used to establish a spatiotemporal distribution model of hydrodynamic parameters, which is then labeled into a three-dimensional mesh model.

4. The AUV motion planning and optimization method for complex marine environments according to claim 1, characterized in that, In step S2, the target gravity function and obstacle repulsion function are combined with the spatiotemporal distribution model of hydrodynamic parameters to perform initial path planning; The target gravity function guides the growth of the random tree based on the distance and direction between the AUV and the target point, while the obstacle repulsion function adjusts the growth direction of the random tree based on the distance between the obstacle and the random tree node.

5. The AUV motion planning and optimization method for complex marine environments according to claim 1, characterized in that, In step S3, the multi-objective optimization function includes: path length, path curvature, and energy consumption; The path optimization method based on simulated annealing algorithm is adopted. The optimization of the initial path includes: generating new path solutions by randomly perturbing the path nodes; calculating the difference between the multi-objective optimization function of the new path and the multi-objective optimization function of the initial path; determining whether to accept the new path solution based on the difference of functions according to the Metropolis criterion; if accepted, the new path solution is used as the initial path for the next round; if not accepted, the current initial path is retained for the next round of optimization, until the optimization is completed.

6. The AUV motion planning and optimization method for complex marine environments according to claim 1, characterized in that, In step S4, the environmental monitoring indicators include: obstacle change monitoring indicators, hydrodynamic change monitoring indicators, and comprehensive environmental change indicators. The obstacle change monitoring indicators are calculated based on Hausdorff distance; the hydrodynamic change monitoring indicators are constructed by combining the AUV's own motion state; and the comprehensive environmental change indicators are constructed by combining the obstacle change monitoring indicators and the hydrodynamic change indicators.

7. The AUV motion planning and optimization method for complex marine environments according to claim 1, characterized in that, In step S4, during the optimized path replanning, parallel computing technology is used to synchronously expand multiple branches of the random tree.

8. The AUV motion planning and optimization method for complex marine environments according to claim 1, characterized in that, In step S5, the dynamic energy consumption model includes the thruster power model, the attitude adjustment mechanism power model, and the sensor system power model.

9. The AUV motion planning and optimization method for complex marine environments according to claim 1, characterized in that, In step S5, the reinforcement learning algorithm uses the energy state, position, velocity, attitude, and environmental information of the AUV as the state space, and velocity adjustment and attitude adjustment as the action space. It learns the optimal energy and motion control strategy by designing a reward function.

10. An AUV motion planning and optimization system for complex marine environments, characterized in that, The system is used to implement the method as described in any one of claims 1-9, the system comprising: Environmental perception and modeling module: used to collect multi-source environmental data, fuse the collected multi-source environmental data, and construct a three-dimensional mesh model that includes obstacle areas, safe navigation areas and hydrodynamic parameters; Initial path planning module: Used for initial path planning based on a 3D mesh model, employing an improved fast expanding random tree algorithm, and introducing target gravity function and obstacle repulsion function; Path optimization module: Used to construct a multi-objective optimization function based on the initial path planning, and to optimize the initial path using a path optimization method based on simulated annealing algorithm; Real-time dynamic adjustment module: Based on the optimized path, it monitors changes in the marine environment in real time by constructing environmental monitoring indicators. When the environmental changes exceed the set threshold, it triggers the replanning of the optimized path based on the incremental fast-expanding random tree algorithm. Energy Management and Motion Control Module: This module is used to build a dynamic energy consumption model, formulate speed planning and attitude adjustment strategies for the AUV based on the model, and achieve coordinated optimization of energy and motion through reinforcement learning algorithms.

Images