AUV (Autonomous Underwater Vehicle) energy optimization path planning method and device

Abstract

The AUV energy optimization path planning method comprises the following steps: initializing a three-dimensional planning environment, and constructing a first random tree with a starting point as a root node and a second random tree with an end point as the root node; flow field information of the three-dimensional planning environment is obtained, if unknown, the flow field information is sensed in real time through a sensor, and if known, a pre-stored flow field model is called; calculating sampling probability distribution based on the flow field information, and correcting the sampling probability through the consistency of the flow field and the target, the distance weight and the target bias strategy; alternately expanding the first random tree and the second random tree, and introducing an artificial potential field to carry out flow field adaptive optimization on node positions in the expansion process; performing collision detection on a path segment between a newly generated node and a father node of the newly generated node, and if the path segment is safe, adding a random tree and performing reconnection optimization; and when the first random tree and the second random tree are successfully connected, extracting a global path, and outputting an energy optimization path. According to the method, a downstream energy-saving mechanism can be fully utilized, so that the length of an AUV planned path is shorter, and the calculation efficiency is higher.

Description

Technical Field

[0001] This invention relates to the field of underwater vehicle technology, and in particular to an AUV energy optimization path planning method and apparatus. Background Technology

[0002] An autonomous underwater vehicle (AUV) is a mission controller that integrates artificial intelligence and other advanced computing technologies. It is a type of unmanned underwater vehicle capable of autonomously operating according to pre-programmed instructions or adapting to environmental changes to complete tasks. It can be applied to marine scientific research, resource exploration, and environmental monitoring. Path planning technology, as a key manifestation of the AUV's autonomous decision-making capability, directly determines its mission execution efficiency, navigation safety, and energy utilization efficiency. Unlike terrestrial or aerial environments, AUVs operate in complex three-dimensional unstructured underwater spaces, requiring them not only to avoid static obstacles but also to cope with the continuous and spatiotemporally variable influence of ocean currents. Furthermore, AUVs carry extremely limited energy. Therefore, path planning is not a simple geometric shortest path problem but a multi-objective optimization problem that requires a comprehensive trade-off between path length, energy consumption, and navigation safety. Against this backdrop, effectively utilizing ocean current information to plan energy-efficient routes has become a key challenge in improving the endurance and operational capabilities of AUVs.

[0003] Sampling-based path planning algorithms, represented by the rapidly-exploring random tree (RRT) and its asymptotically optimal variant RRT*, generally possess probabilistic completeness and asymptotic optimality. RRT* algorithms do not require environmental modeling and are suitable for high-dimensional spaces, but suffer from drawbacks such as blind searching and slow convergence. Subsequent research, such as InformedRRT*, has improved convergence efficiency by restricting the sampling region, but it still does not effectively utilize environmental information to guide the search direction.

[0004] The Artificial Potential Algorithm (APF) works by constructing a potential function that allows a moving object to move within a potential field based on the forces acting upon it. The target point exerts an attractive force on the moving object, while obstacles exert a repulsive force. This method is simple to implement but suffers from the problem of unreachable local minima. Current research incorporates the APF algorithm into the heuristic function of Bi-RRT (Bi-RRT) to improve its efficiency as a global path and uses a dynamic window method for local path planning to handle dynamic obstacles, thus achieving path planning in dynamic environments.

[0005] While traditional RRT* algorithms possess probabilistic completeness, their uniform random sampling strategy suffers from high search blindness and high energy consumption in path planning. Directly applying them to complex marine environments with dynamic flow fields such as ocean currents presents significant challenges in terms of search time and energy efficiency. Existing methods combining RRT* with artificial potential fields also fail to consider the impact of flow fields on AUV energy consumption. Summary of the Invention

[0006] To address the problems existing in the prior art, this application provides a method, apparatus, computing device, computer storage medium, and product containing a computer program for AUV energy optimization path planning, which can fully utilize the downstream energy-saving mechanism to make the AUV planning path shorter and the calculation efficiency higher.

[0007] In a first aspect, embodiments of this application provide an AUV energy optimization path planning method, comprising: initializing a three-dimensional planning environment, establishing a grid map, setting a start point, an end point, and an obstacle region, and constructing a first random tree and a second random tree, wherein the first random tree has the start point as the root node and the second random tree has the end point as the root node; acquiring the flow field information of the three-dimensional planning environment, if unknown, sensing it in real time through sensors, and if known, calling a pre-stored flow field model; calculating the sampling probability distribution based on the flow field information, and correcting the sampling probability through flow field and target consistency, distance weight, and target bias strategies; alternately expanding the first random tree and the second random tree using a bidirectional RRT* algorithm, introducing an artificial potential field during the expansion process to adaptively optimize the flow field at the node positions; performing collision detection on the path segments between newly generated nodes and their parent nodes, and if safe, adding them to the random tree and performing reconnection optimization; extracting the global path when the first random tree and the second random tree are successfully connected, and outputting the energy-optimized path.

[0008] In some possible implementations, the calculation of the sampling probability distribution based on flow field information includes: calculating the consistency metric between the flow field direction and the direction pointing to the target in each grid; sorting and assigning flow field weights according to the consistency metric; calculating the adaptive distance weight centered on the node in the current tree that is closest to another tree; the total grid weight is the product of the distance weight and the flow field weight, which is normalized to obtain the final sampling probability.

[0009] In some possible implementations, the calculation of the consistency measure between the flow field direction in each grid cell and the direction pointing towards the target is performed using the following formula:

[0010]

[0011] In the formula, The center point of the grid The velocity vector of the flow field at that location, To pre-determine the destination of the route planning; according to The values ​​are sorted in descending order for all candidate points, and their flow field weights are determined by their positions in the sequence, as expressed by the formula:

[0012]

[0013] In the formula, To control the parameters of the probability distribution, and have , where n is the number of grid cells and j is the j-th grid cell.

[0014] In some possible implementations, the sampling probability is corrected by distance weighting, including: setting an adaptive sampling radius r with the midpoint of the line connecting the closest pair of nodes in the first random tree and the second random tree as the center of the sphere. ,in, The attenuation coefficient is... For the minimum radius, The minimum distance between the two trees; based on the center and radius of the sphere, the distance weight function is defined as follows:

[0015]

[0016] In the formula, For sampling points, The center of the sphere is within the spherical range. and It is a positive number, and .

[0017] In some possible implementations, the sampling probability is corrected through a target bias strategy, including: generating a sampling random number at each sampling time; when the sampling random number is greater than a set threshold, performing random sampling according to the sampling probability distribution; when the sampling random number is not greater than the set threshold, sampling the endpoint of the first random tree and sampling the starting point of the second random tree.

[0018] In some possible implementations, the introduction of an artificial potential field for adaptive flow field optimization at the node locations is described by the total potential function of the artificial potential field.

[0019] In the formula, The attractive potential generated at the target point The repulsive potential generated by the obstacle The potential for the introduced flow field.

[0020] In some possible implementations, the flow potential is defined as

[0021] In the formula, It is an adaptive function for flow field gain. For the direction weight adaptive function, The distance between the first random tree and the second random tree. This represents the degree of matching between the current direction of motion and the direction of the flow field. This represents the degree of matching between the flow field direction and the direction of the line connecting the current position and the target. For in position The water flow velocity at that location.

[0022] In some possible implementations, the adaptive flow field gain function is: In the formula, This is the flow field potential intensity adjustment coefficient; when the vehicle is downstream, ,have When the aircraft is sailing against the current, ,have ,coefficient .

[0023] In some possible implementations, the cost function of the path jointly optimizes the length and energy consumption, expressed as: Where r is the complete path from the starting point to the ending point. This is the path length coefficient. The total energy consumption coefficient is... Path length The total energy consumption is the sum of the Euclidean distances between all adjacent nodes. It represents the sum of the energy consumed to overcome water resistance and the additional energy consumed to turn.

[0024] Secondly, embodiments of this application provide an AUV energy optimization path planning device, comprising: an initialization module, used to initialize a three-dimensional planning environment, establish a grid map, set a start point, an end point, and an obstacle area, and construct a first random tree and a second random tree, wherein the first random tree has the start point as the root node and the second random tree has the end point as the root node; an acquisition module, used to acquire the flow field information of the three-dimensional planning environment, which is sensed in real time by sensors if unknown, and called a pre-stored flow field model if known; a processing module, used to calculate the sampling probability distribution based on the flow field information, and correct the sampling probability through flow field and target consistency, distance weight, and target bias strategies; the processing module is also used to alternately expand the first random tree and the second random tree using a bidirectional RRT* algorithm, and introduce an artificial potential field to adaptively optimize the flow field at the node position during the expansion process; the processing module is also used to perform collision detection on the path segment between the newly generated node and its parent node, and if safe, add it to the random tree and perform reconnection optimization; the processing module is also used to extract the global path and output the energy-optimized path when the first random tree and the second random tree are successfully connected.

[0025] Thirdly, embodiments of this application provide a computer-readable storage medium including computer-readable instructions that, when read and executed by a computer, cause the computer to perform the method as described in any of the first aspects.

[0026] Fourthly, embodiments of this application provide a computing device, including a processor and a memory, wherein the memory stores computer program instructions, which, when executed by the processor, perform the method as described in any of the first aspects.

[0027] Fifthly, embodiments of this application provide a product comprising a computer program that, when the computer program product is run on a processor, causes the processor to perform the method as described in any of the first aspects. Attached Figure Description

[0028] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0029] Figure 1 This is a flowchart illustrating an AUV energy optimization path planning method provided in an embodiment of this application;

[0030] Figure 2 This is a flowchart illustrating an AUV energy optimization path planning method provided in an embodiment of this application;

[0031] Figure 3 This is a real terrain and flow field map provided in the embodiments of this application;

[0032] Figure 4 This is a schematic diagram illustrating the effect of Bi-RRT* path planning provided in an embodiment of this application;

[0033] Figure 5 This is a schematic diagram illustrating the path planning effect using the AUV energy optimization path planning method provided in an embodiment of this application;

[0034] Figure 6 This is a schematic diagram of the structure of an AUV energy optimization path planning device provided in an embodiment of this application. Detailed Implementation

[0035] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0036] In this article, the term "and / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. The symbol " / " in this article indicates that the related objects are in an "or" relationship; for example, A / B means A or B.

[0037] The terms "first" and "second," etc., used in the specification and claims herein are used to distinguish different objects, not to describe a specific order of objects. For example, "first response message" and "second response message," etc., are used to distinguish different response messages, not to describe a specific order of response messages.

[0038] In the embodiments of this application, the terms "exemplary" or "for example" are used to indicate that something is an example, illustration, or description. Any embodiment or design that is described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design. Specifically, the use of the terms "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.

[0039] In the description of the embodiments of this application, unless otherwise stated, "multiple" means two or more, for example, multiple processing units means two or more processing units, multiple elements means two or more elements, etc.

[0040] To facilitate understanding of the embodiments of this application, the following will provide further explanation and description with reference to the accompanying drawings and specific embodiments. These embodiments do not constitute a limitation on the embodiments of the present invention.

[0041] This application provides an AUV energy optimization path planning method to generate high-quality paths that are collision-free, low-energy, and conform to flow field characteristics.

[0042] For example, Figure 1 A flowchart illustrating an AUV energy optimization path planning method provided in an embodiment of this application is shown. Figure 1 As shown, the method may include the following steps:

[0043] S11: Initialize the 3D planning environment, establish a grid map, set the start point, end point and obstacle area, and construct the first random tree and the second random tree, wherein the first random tree is rooted at the start point and the second random tree is rooted at the end point.

[0044] In this embodiment, a three-dimensional spatial model is performed on the operating area of ​​the autonomous underwater vehicle (AUV). Specifically, the planned space is divided into a uniform or non-uniform three-dimensional grid map, with each grid cell representing the accessibility status of that spatial location. Obstacle areas are identified based on prior environmental information or real-time detection data. These obstacle areas include, but are not limited to, seabed protrusions, shipwrecks, underwater facilities, or other static obstacles that prohibit passage. All grid cells marked as obstacles are considered impassable during subsequent path search. The starting and ending points of the path planning are pre-defined in the three-dimensional grid map. The starting point is denoted as... The destination is recorded as Using the start and end points as root nodes respectively, construct a first random tree and a second random tree, setting parameters such as the maximum number of iterations. The tree will be constructed using the start point... The first random tree with root node denoted as , will be the end The second random tree with the root node is denoted as . The first and second random trees grow together using a bidirectional expansion strategy to accelerate path convergence and improve search efficiency. Initially, each tree contains only its corresponding root node, and all nodes store their spatial coordinates, parent node pointers, cumulative path costs, and energy consumption information related to the flow field.

[0045] S12: Obtain flow field information. If the information is unknown, it will be sensed in real time by sensors. If the information is known, a pre-stored flow field model will be called.

[0046] In this embodiment, after initializing the 3D planning environment, ocean current field information within the current task area is acquired. When acquiring ocean current field information, a determination is made as to whether the current field data for the planning area is known. If the current field information has been pre-acquired and stored (e.g., from historical ocean observation data, numerical ocean models, or high-precision reanalysis datasets), the corresponding pre-stored current field model is directly invoked. The pre-stored current field model can be represented as a 3D vector field, containing the magnitude and direction of the water flow velocity at each grid location. If the current field information is unknown, the current field sensing module is activated. Using sensors mounted on the AUV, the estimated local current field state is acquired in real time, and a local or global current field vector map is constructed. These sensors may include a Doppler log, an inertial navigation system, and an acoustic current profiler.

[0047] S13: Calculate the sampling probability distribution based on flow field information, and correct the sampling probability through flow field and target consistency, distance weight, and target bias strategies.

[0048] In this embodiment, after obtaining the three-dimensional flow field vector field, the random sampling process in the traditional RRT algorithm is modified by comprehensively considering the consistency guidance between the flow field and the target, adaptive distance weights, and algorithm completeness, in order to improve path search efficiency and reduce energy consumption. Specifically, an adaptive sampling probability distribution that integrates flow field characteristics and target guidance is constructed for subsequent expansion of the first and second random trees. The modification of the random sampling process includes flow field and target consistency, distance weights, and target bias strategies.

[0049] Specifically, in a flow field environment, to improve the path search efficiency of Bi-RRT*, priority should be given to sampling regions where the flow direction aligns with the direction pointing towards the target, thus guiding the algorithm to sample points that better meet energy-saving goals. For a grid location in a grid network, the grid center point... Instead, define a measure of the consistency between the flow field at that point and the target direction. for: In the formula, For point The velocity vector of the flow field at that location. The range of its value is [-1, 1]. When When , it indicates that the flow field direction is completely consistent with the target direction. When the flow direction is completely opposite to the target direction, energy consumption is at its highest. When , it indicates that the direction of the flow field is perpendicular to the target direction.

[0050] In order to base on consistency metrics (Right now Assigning flow field weights, i.e., sampling probabilities determined by the influence of the flow field, first assigns all grid cells according to their values. Sort from largest to smallest, the first-ranked grid cell has the largest... Value. Let the sorted rank number be... ( Corresponding to the highest to the lowest If the value is 0, then the ranking is 0. Flow field weights of the grid Defined as: In the formula, To control the parameters of the probability distribution, and have , where n is the number of grid cells and j is the j-th grid cell.

[0051] This approach allows highly consistent rasters to have a higher sampling probability, thus guiding the algorithm to converge quickly to an energy-efficient path. Simultaneously, considering that regions with different flow directions are scattered throughout the map, if a path crosses multiple consecutive flow regions with different directions, energy consumption will increase. Therefore, the sampling process needs to focus more on the area near the current tree growth front, avoiding unnecessary long-distance exploration, while maintaining global search capabilities. To this end, a distance weighting function is designed to adaptively adjust the sampling range, increasing the sampling probability of nearby areas and decreasing the sampling probability of distant areas.

[0052] Define the distance weight function as an adaptive radius A function that is uniformly sampled inside and outside a spherical region bounded by: .in, For sampling points, The center of the sphere is within the spherical range. and It is a positive number, and However, the difference between the two should not be too large, so that the algorithm still has sufficient completeness. The first random tree and the second random tree each use the node closest to the other tree as the center of the distance weight function to guide the sampling to focus on the connection direction between the two trees, as follows: The minimum distance between two trees is: The adaptive sampling radius decreases as the distance between the two trees approaches, avoiding excessive shrinkage. This encourages broad sampling when the two trees are far apart, while focusing on local optimization and encouraging direct connections between the two trees when they are close. .in, The attenuation coefficient is... The minimum radius is given; the above parameters should be set according to actual needs.

[0053] For grid , assuming it is according to The ranking after sorting the values ​​is Then the total weight of the grid is the product of the distance weight and the flow field weight: After normalization, the final sampling probability is obtained as follows: .

[0054] Finally, a target point bias strategy is introduced to improve the algorithm's target orientation by increasing the probability of the target region being sampled. That is, when sampling random numbers Greater than the set threshold When the target point is in the target tree, random sampling is performed according to the sampling probability distribution; otherwise, the target point is sampled directly from the starting tree, and the starting point is sampled directly from the target tree. This ensures that the algorithm always has the opportunity to move directly towards the target, even if the target is in reverse flow. In this way, when the AUV needs to move in reverse flow, the sampling will actively cover the reverse flow path.

[0055] S14: The bidirectional RRT* algorithm is used to alternately expand the first random tree and the second random tree. During the expansion process, an improved artificial potential field is introduced to adaptively optimize the flow field at the node position.

[0056] In this embodiment, after constructing the adaptive sampling probability distribution, a bidirectional RRT algorithm is initiated to alternately expand the first and second random trees to efficiently search for a collision-free, low-energy path from the starting point to the ending point. In each iteration, a sampling point is first randomly generated based on the generated sampling probability distribution. Subsequently, in the current random tree to be expanded (initially...), a sampling point is generated randomly. The next round will switch to The algorithm searches for the nearest node in the path (and so on) and generates a new node along that direction with a preset step size. Because the sampling probability differs from traditional average random sampling, this algorithm considers both path length and flow field energy consumption in its cost function when selecting a parent node for the new expanded node and optimizing reconnection. Simultaneously, during the expansion process, the algorithm introduces an artificial potential field method to correct the target point. This potential field method further transforms the flow field force into a potential field gradient, generating a potential field force on the randomly sampled point or expanded node, fine-tuning it to make the local path segment better conform to the flow field and achieve energy consumption optimization.

[0057] Specifically, the traditional artificial potential field method forms a composite potential field that guides the movement of a body by superimposing the attractive potential of the target and the repulsive potential of the obstacle. Let the position vector of the moving body in three-dimensional space be... The target point position vector is The nearest point position vector on the obstacle surface is Target attraction potential It is the pulling force that causes the moving body to move towards the target point, expressed as: .in, The attraction potential coefficient determines the strength of the attraction potential. Let be the Euclidean distance between the AUV and the target point. The attraction vector generated by the attraction potential is the negative gradient of the attraction potential field, i.e.: The direction of attraction always points towards the target point, and its magnitude increases linearly with the distance between the moving object and the target point.

[0058] The repulsive potential of an obstacle generates a thrust that moves the moving object away from the obstacle. To avoid the repulsive potential having an excessive influence in the global space, it is usually constructed using a nonlinear function that only affects the vicinity of the obstacle, and its expression is:

[0059]

[0060] In the formula, The repulsive potential coefficient, The shortest distance between the moving object and the surface of the obstacle. This is the critical distance. When the shortest distance between the moving object and the obstacle exceeds... When the repulsive potential fails, the repulsive force generated by the repulsive potential is expressed as:

[0061]

[0062] The repulsive force always moves away from the obstacle, and its magnitude increases sharply as the distance between the moving body and the obstacle decreases, which can effectively prevent the moving body from colliding with the obstacle.

[0063] The total potential field of a traditional artificial potential field is the superposition of attractive and repulsive potentials, that is... The total force acting on the moving body is the negative gradient of the total potential field. Based on the traditional artificial potential field method, this embodiment introduces an adaptive flow field potential to address the issue of the influence of flow field forces on the motion state of a moving body in complex flow field environments. This approach not only considers the consistency between the vehicle's motion direction and the flow field direction but also integrates the target direction with the flow field direction, ensuring that while fully utilizing the downstream energy-saving mechanism, it avoids falling into local optima due to upstream flow. For still water conditions, no flow field potential is generated. Let... The potential generated by the flow field is then the total potential field. for Define the flow potential function. for: .in, It is an adaptive function for flow field gain. For the direction weight adaptive function, The distance between the first random tree and the second random tree. This represents the degree of matching between the current direction of motion and the direction of the flow field. This represents the degree of matching between the flow field direction and the direction of the line connecting the current position and the target. For in position The water flow velocity at that location. Among them, In the formula, This is the current motion direction vector of the AUV, i.e., the direction connecting the current tree to the sampling point. When the vehicle moves downstream, When the potential energy is at its minimum, movement in that direction is encouraged; conversely, when the vehicle moves against the current, it is discouraged from moving in that direction. When the potential energy is at its maximum, it inhibits movement in that direction. In the formula, This is the current position of the AUV, i.e., the direction vector connecting the node closest to the sampling point in the current tree to the target point. This term is used to ensure global convergence, making the path eventually approach the target point. To achieve adaptive adjustment of the flow field potential function, a dual adaptive mechanism is designed. First, an adaptive function with direction weights is designed. for In the formula, , It is a constant, and has When the distance between two trees is relatively large, Larger Flow field potential function The focus is more on optimizing the direction of movement, prioritizing fine-tuning along the flow to reduce energy consumption in the current movement segment. When the distance between two trees is relatively close... Smaller Flow field potential function It focuses more on guiding the path in the direction of the target, thus accelerating path convergence.

[0064] When flowing downstream, fine-tuning along the flow field direction can achieve energy-saving effects. However, when flowing upstream, if adjustments along the flow field direction are still considered, it may prevent the vehicle from traveling upstream, hindering path planning when the target point is in the upstream direction. Therefore, this embodiment proposes an adaptive flow field gain function: In the formula, This is the flow field potential intensity adjustment coefficient; when the vehicle is downstream, ,but Make full use of energy conservation with the current; when the aircraft is sailing against the current, ,but To reduce the influence of the flow field potential on the entire potential function and avoid excessive deviation from the target caused by following the countercurrent, the coefficients are used. Adjusting the magnitude of the influence of the flow field potential during countercurrent.

[0065] The force induced by the flow potential is obtained by taking the negative gradient of the flow potential: The total force generated by the improved artificial potential field is: This potential field design ensures the robustness of the algorithm in complex flow field environments, making full use of environmental information to achieve energy-saving optimization while guaranteeing the completeness and safety of path planning.

[0066] S15: Perform collision detection on the path segments between the newly generated node and its parent node. If it is safe, add it to the random tree and perform reconnection optimization.

[0067] In this embodiment, after a new node is obtained through flow field adaptive optimization of the node position, the path formed between this node and its parent node needs to be detected to ensure that the planned path meets the collision-free constraint. If the line segment is entirely within the passable grid and does not intersect with any preset obstacle area, it is determined to be a collision-free path segment, indicating safety. If the line segment crosses any obstacle grid or enters a prohibited navigation area, it is determined to have a collision risk, the current new node is discarded, and this expansion terminates. After the path segment passes the collision detection, the optimized new node is added to the currently expanded random tree, and its spatial coordinates, parent node pointer, cumulative path length, and cumulative energy consumption cost are recorded. Then, reconnection optimization is performed, and the node with the optimal path cost is taken as the new parent node, and the path cost of all nodes in its subtree is iteratively updated.

[0068] Joint optimization of path cost function: Length and energy consumption Where r is the complete path from the starting point to the ending point. This is the path length coefficient. The total energy consumption coefficient is... Path length The sum of the Euclidean distances between all adjacent nodes is expressed by the formula: Total energy consumption It represents the sum of the energy consumed to overcome water resistance and the additional energy consumed during turning, expressed by the formula: Where A is a constant related to the hydrodynamic drag coefficient of the aircraft, water density, the aircraft's frontal area, and the aircraft's mass; B is the aircraft's turning energy consumption coefficient. A and B are set according to actual conditions. This refers to the vehicle's speed relative to the ground. The turning angle between adjacent node segments satisfies That is, taking the angle of the minimum turning direction, where, The heading angle. According to the vector composition rule, the object's velocity relative to the ground... for ,in, The three-dimensional velocity vector of the water flow at the location of the vehicle. This is the three-dimensional velocity vector of the vehicle relative to the water.

[0069] S16: When the first random tree and the second random tree are successfully connected, extract the global path and output the energy-optimized path.

[0070] In this embodiment, the distance between the first and second random trees is monitored in real time during each iteration. The algorithm terminates when the first and second random trees are successfully connected, i.e., when the minimum distance between a newly added node in this tree and a node in the other tree is less than the connection threshold. This is achieved by backtracking the random trees from the connection point. The first random tree is backtracked from the connection point. until the starting point This yields the first half of the path. The second random tree is then traced back from the connection point. until the finish line This yields the second half of the path. By concatenating the first and second halves of the path, a globally optimal path from the starting point to the destination can be obtained.

[0071] Figure 2 A flowchart illustrating the AUV energy optimization path planning method provided in an embodiment of this application is shown. Figure 2 As shown, the execution flow of this method is as follows: First, the planning environment is initialized. Then, the algorithm determines whether the current flow field information is known. If unknown, the flow field information sensing module is activated to estimate local flow field data online using sensor data; if known, the pre-built flow field model is directly called. Based on the acquired flow field information, the sampling probability of flow field correction is calculated. This strategy aims to reduce the sampling probability of countercurrent and high-energy-consumption areas, while increasing the sampling probability of downstream and favorable areas, thereby guiding the search direction from a global perspective and reducing blindness. Next, Bi-RRT* with probabilistic random sampling is used for alternating expansion and connection of two trees. Since the sampling probability is different from the traditional average random sampling, the cost function of this algorithm comprehensively considers path length and flow field energy consumption when selecting parent nodes for new expansion nodes and optimizing reconnection. At the same time, during the expansion process, the algorithm introduces an artificial potential field method called "flow field potential" to correct the target point. This potential field method, based on the traditional artificial potential field method that considers obstacle avoidance, further transforms the flow field force into a potential field gradient. This gradient exerts attractive or repulsive forces on randomly sampled points or extended nodes, fine-tuning them to make local path segments better conform to the flow field and achieve energy consumption optimization. After determining new random extended points, collision detection is performed on the new path segments generated after correction by the artificial potential field. If a collision occurs, the path is discarded; if safe, the node is formally added to the random tree, and Bi-RRT* parent node reconnection optimization is performed. Sampling continues, and the algorithm terminates its iteration when two trees are successfully connected (i.e., the minimum distance between a newly added node in one tree and a node in another tree is less than the connection threshold). Finally, by backtracking the random tree from the connection point to the starting and ending points respectively, a globally optimal or suboptimal path from the starting point to the ending point can be obtained.

[0072] The pseudocode of the algorithm is shown in Table 1.

[0073] Table 1

[0074]

[0075] To verify the effectiveness of this method, a simulation comparison experiment was conducted. The obstacle model used multiple rectangular obstacles. Real terrain and flow field data were obtained from the SODA (Simple Ocean Data Assimilation) dataset, with a range of 112°E ~ 114°E, 16°N ~ 18°N. Figure 3 As shown, latitude, longitude, and elevation are used as the three-dimensional coordinates of the grid. The main hyperparameters are shown in Table 2 below.

[0076] Table 2

[0077]

[0078] Path planning is performed on the real terrain flow field, and only Bi-RRT* and our method are compared. Starting point [112.473, 16.5962, -10], target point [112.65, 17.75, -25]. Figure 4 This diagram illustrates the effect of Bi-RRT* path planning. Figure 5 A schematic diagram illustrating the path planning effect of this method is shown. For example... Figure 4 and Figure 5 As shown in the figure, the map color represents the terrain elevation. Table 3 compares the path energy consumption, path length, and total cost of Bi-RRT* and this method in a real terrain flow field environment. It can be seen that this method can also achieve more efficient path searching in a real flow field, demonstrating advantages in various evaluation metrics.

[0079] Table 3

[0080]

[0081] The above is an introduction to the AUV energy-optimized path planning provided in the embodiments of this application. By combining the global search capability of bidirectional RRT* with the improved artificial potential field method, the algorithm effectively reduces path length and energy consumption while maintaining probabilistic completeness and using the flow field potential field to adaptively guide the sampling process. Simulation experiments show that this method exhibits superior performance compared to the traditional Bi-RRT*: shorter path length, higher computational efficiency, and full utilization of the downstream energy-saving mechanism. In non-downstream environments, the planned path can also avoid upstream deviation, demonstrating the effectiveness of this method and providing a new approach for AUV energy-saving path planning in ocean currents.

[0082] It is understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application. Furthermore, in some possible implementations, each step in the above embodiments may be selectively executed according to actual circumstances; it may be partially or fully executed, without limitation here. All or part of any feature of any embodiment of this application can be freely and arbitrarily combined without contradiction. The combined technical solutions are also within the scope of this application.

[0083] Based on the methods in the above embodiments, this application also provides an AUV energy optimization path planning device. Figure 6 A schematic diagram of an AUV energy optimization path planning device provided in an embodiment of this application is shown. Figure 6 As shown, the AUV energy optimization path planning device 600 includes an initialization module 601, an acquisition module 602, and a processing module 603.

[0084] The initialization module 601 is used to initialize the three-dimensional planning environment, establish a grid map, set the start point, end point and obstacle area, and construct a first random tree and a second random tree. The first random tree is rooted at the start point and the second random tree is rooted at the end point.

[0085] The acquisition module 602 is used to acquire the flow field information of the three-dimensional planning environment. If the information is unknown, it is sensed in real time by sensors; if the information is known, it calls the pre-stored flow field model.

[0086] The processing module 603 is used to calculate the sampling probability distribution based on the flow field information and correct the sampling probability through the consistency between the flow field and the target, distance weight, and target bias strategy.

[0087] The processing module 603 is also used to perform alternating expansion using a bidirectional RRT* algorithm, and to introduce an artificial potential field to adaptively optimize the flow field at the node position during the expansion process.

[0088] The processing module 603 is also used to perform collision detection on the path segments between the newly generated node and its parent node. If it is safe, it is added to the random tree and reconnection optimization is performed.

[0089] The processing module 603 is also used to extract the global path and output the energy-optimized path when the first random tree and the second random tree are successfully connected.

[0090] It should be understood that the above-described device is used to execute the methods in the above embodiments. The implementation principle and technical effect of the corresponding program modules in the device are similar to those described in the above methods. The working process of the device can be referred to the corresponding process in the above methods, and will not be repeated here.

[0091] Based on the methods in the above embodiments, this application provides a computer-readable storage medium storing a computer program that, when run on a processor, causes the processor to execute the methods in the above embodiments.

[0092] Based on the methods in the above embodiments, this application provides a computer program product that, when run on a processor, causes the processor to execute the methods in the above embodiments.

[0093] It is understood that the processor in the embodiments of this application may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. A general-purpose processor may be a microprocessor or any conventional processor.

[0094] The method steps in the embodiments of this application can be implemented in hardware or by a processor executing software instructions. The software instructions can consist of corresponding software modules, which can be stored in random access memory (RAM), flash memory, read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), registers, hard disks, portable hard disks, CD-ROMs, or any other form of storage medium known in the art. An exemplary storage medium is coupled to the processor, enabling the processor to read information from and write information to the storage medium. Of course, the storage medium can also be a component of the processor. The processor and the storage medium can reside in an ASIC.

[0095] In the above embodiments, implementation can be achieved entirely or partially through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented entirely or partially in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted through the computer-readable storage medium. The computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid-state disk (SSD)).

[0096] It is understood that the various numerical designations used in the embodiments of this application are merely for descriptive convenience and are not intended to limit the scope of the embodiments of this application.

Claims

1. An AUV energy optimization path planning method, characterized in that, The method includes: Initialize the 3D planning environment, establish a grid map, set the start point, end point and obstacle area, and construct a first random tree and a second random tree, wherein the first random tree is rooted at the start point and the second random tree is rooted at the end point; The flow field information of the three-dimensional planning environment is obtained. If it is unknown, it is sensed in real time by sensors. If it is known, a pre-stored flow field model is called. The sampling probability distribution is calculated based on the flow field information, and the sampling probability is corrected by the consistency between the flow field and the target, distance weight, and target bias strategy. The bidirectional RRT* algorithm is used to alternately expand the first random tree and the second random tree. During the expansion process, an artificial potential field is introduced to adaptively optimize the flow field at the node position. Collision detection is performed on the path segments between the newly generated node and its parent node. If the path is safe, it is added to the random tree and reconnection optimization is performed. When the first random tree and the second random tree are successfully connected, the global path is extracted and the energy-optimized path is output.

2. The method according to claim 1, characterized in that, The calculation of the sampling probability distribution based on flow field information includes: Calculate the consistency measure between the flow field direction and the direction pointing to the target in each grid cell; Flow field weights are assigned based on consistency metrics; Calculate the adaptive distance weight centered on the node in the current tree that is closest to the node in the other tree; The total grid weight is the product of the distance weight and the flow field weight, and the final sampling probability is obtained after normalization.

3. The method according to claim 2, characterized in that, The formula for calculating the consistency metric between the flow field direction and the target direction in each grid cell is as follows: In the formula, The center point of the grid The velocity vector of the flow field at that location, To pre-determine the destination of the route planning; according to The values ​​are sorted in descending order for all candidate points, and their flow field weights are determined by their positions in the sequence, as expressed by the formula: In the formula, To control the parameters of the probability distribution, and have n is the number of grid cells, and j is the j-th grid cell. This represents the position in the sequence.

4. The method according to claim 1, characterized in that, The sampling probability is corrected by distance weighting, including: Using the midpoint of the line connecting the closest pair of nodes in the first random tree and the second random tree as the center of the sphere, the adaptive sampling radius r is set as... ,in, The attenuation coefficient is... For the minimum radius, This represents the minimum distance between two trees. Based on the center and radius of the sphere, the distance weighting function is defined as follows: In the formula, For sampling points, The center of the sphere is within the spherical range. and It is a positive number, and .

5. The method according to claim 1, characterized in that, The sampling probability is corrected through a target bias strategy, including: Generate a random number for each sample. When the number of random samples exceeds the set threshold, random sampling is performed according to the sampling probability distribution; When the sampled random number is not greater than the set threshold, the end point of the first random tree is sampled, and the starting point of the second random tree is sampled.

6. The method according to claim 1, characterized in that, The introduction of an artificial potential field for adaptive flow field optimization at node locations, wherein the total potential function of the artificial potential field is expressed as follows: In the formula, The attractive potential generated at the target point The repulsive potential generated by the obstacle The potential for the introduced flow field.

7. The method according to claim 6, characterized in that, The flow field potential is defined as In the formula, It is an adaptive function for flow field gain. For the direction weight adaptive function, The distance between the first random tree and the second random tree. This represents the degree of matching between the current direction of motion and the direction of the flow field. The degree of matching between the flow field direction and the direction of the line connecting the current position and the target. For in position The water flow velocity at that location.

8. The method according to claim 7, characterized in that, The adaptive function for the flow field gain is: In the formula, This is the flow field potential intensity adjustment coefficient; when the vehicle is downstream, ,have When the aircraft is sailing against the current, ,have ,coefficient .

9. The method according to claim 1, characterized in that, The cost function of the path jointly optimizes the length and energy consumption, expressed as: Where r is the complete path from the starting point to the ending point. This is the path length coefficient. The total energy consumption coefficient is... Path length The total energy consumption is the sum of the Euclidean distances between all adjacent nodes. It represents the sum of the energy consumed to overcome water resistance and the additional energy consumed to turn.

10. An AUV energy optimization path planning device, characterized in that, The device includes: The initialization module is used to initialize the 3D planning environment, establish a grid map, set the start point, end point and obstacle area, and construct a first random tree and a second random tree, wherein the first random tree is rooted at the start point and the second random tree is rooted at the end point. The acquisition module is used to acquire the flow field information of the three-dimensional planning environment. If the information is unknown, it is sensed in real time by sensors; if the information is known, a pre-stored flow field model is called. The processing module is used to calculate the sampling probability distribution based on the flow field information, and correct the sampling probability through the consistency between the flow field and the target, distance weight, and target bias strategy. The processing module is also used to alternately expand the first random tree and the second random tree using a bidirectional RRT* algorithm, and to introduce an artificial potential field to adaptively optimize the flow field at the node position during the expansion process. The processing module is also used to perform collision detection on the path segments between the newly generated node and its parent node. If it is safe, it is added to the random tree and reconnection optimization is performed. The processing module is further configured to extract the global path and output the energy-optimized path when the first random tree and the second random tree are successfully connected.

Images