Data-driven shallow water area underwater robot path planning method
Through CFD numerical simulation and current data model, combining global sampling and local directed expansion, a flow field simulation database is established, and physically infeasible areas are identified and corrected, which solves the problem of insufficient physical feasibility of traditional path planning in complex flow fields, and realizes the problem of insufficient physical feasibility of traditional path planning in complex flow fields. More efficient path planning.
Patent Information
- Application Number
- CN202410993490.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-24
- Publication Date
- 2024-11-12
AI Technical Summary
Traditional underwater robot path planning algorithms ignore the interference of water flow in complex flow fields, resulting in the robot being unable to overcome the countercurrent in strong flow areas with high flow velocities, and the physical feasibility of path planning is insufficient.
Using a data-driven method, combining CFD numerical simulation and current data model, a flow field simulation database is established through global sampling and locally directed expansion path planning methods, and physically infeasible areas are identified and corrected to ensure the physical feasibility of the path.
It improves the efficiency and physical feasibility of path planning, avoids the unpassable strong current areas, reduces the number of iterations, and significantly improves the accuracy and efficiency of path planning.
Smart Images

Figure CN118936473A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of underwater robot path planning, and in particular to a data-driven shallow water underwater robot path planning method. Background Art
[0002] Underwater robot operations are sensitive to water flow environments, and different strategies are often required in different waters. The water flow in the deep sea environment is relatively stable, and the water quality has good light transmittance, which has relatively little impact on underwater operations; inland waters are mostly shallow waters, and the hydrological characteristics vary due to the influence of topography and climate, or the water flow is turbulent and changeable, or the water quality is turbid, and the types of underwater tasks are also relatively wide. Therefore, operations in shallow waters require robots to have not only excellent operating performance, but also strong environmental adaptability.
[0003] Path planning is an important part of underwater robot operations. Traditional underwater robot path planning algorithms are mostly borrowed from the path planning algorithms of drones. In complex flow fields, the impact of water flow on robot movement cannot be ignored. Traditional algorithms mostly consider obstacle avoidance, energy consumption, etc., but ignore the interference of water flow. In strong flow areas with high flow rates, there is a tendency for countercurrents that underwater robots cannot overcome. Therefore, path planning needs to consider the physical feasibility of the path. Summary of the invention
[0004] In view of this, the purpose of the present invention is to provide a data-driven path planning method for a shallow water underwater robot, which utilizes CFD numerical simulation combined with ocean current data model as prior information, and a path planning method combining global sampling with local directed expansion to ensure the physical feasibility of path planning and improve the efficiency of path planning.
[0005] To achieve the above object, the present invention adopts the following technical solution: a data-driven shallow water underwater robot path planning method, comprising the following steps:
[0006] Step S1: Model the underwater vehicle and its environment according to the mission requirements, so that it can output the position observation value of the vehicle and the environmental state at the next moment;
[0007] Step S2: Obtain the flow field boundary velocity vector information in the environment, use it as the velocity boundary value to perform CFD numerical simulation, and establish a simulation database of flow field velocity information under all working conditions;
[0008] Step S3: In the actual scenario, real-time hydrological and meteorological information is collected and imported into the flow field simulation database to obtain flow velocity information of all areas in the flow field;
[0009] Step S4: obtaining the current position of the underwater robot from the environment model, performing global sampling and path optimization for the set target point, so as to obtain a theoretically feasible initial path;
[0010] Step S5: Perform feasibility test on the initial path obtained in step 4 to analyze physically feasible areas and physically infeasible areas;
[0011] Step S6: Perform local directed expansion on the physically infeasible area and correct it to a physically feasible path;
[0012] Step S7: Output the optimal reachable path according to the path length.
[0013] In a preferred embodiment, the specific implementation of step S1 is as follows: setting basic information of the environment, mainly including: the boundary size of the environment horizontal plane range, the grid size of the environment, the location information of obstacles, the deployment time of the path planning task, the interaction frequency between the environment and the underwater robot, and defining its action space and observation space;
[0014] When an underwater robot is operating around an underwater structure, it should move according to instructions near obstacles to avoid collision with obstacles; There is an obstacle R : and time-varying flow field in and represents the time interval considered, all position information in the flow field is observable; the underwater robot runs at In the flow field, the path planning task is defined as: Find a reachable safe path ε from the starting point to the destination, which is composed of a set of subpaths connected by control nodes, that is,
[0015] If the water flow velocity in the environment is The speed of the underwater robot in still water is It can be seen that the actual speed of the underwater robot in the environment is:
[0016]
[0017] The area with higher velocity in the convection field:
[0018] When the water flow speed is greater than the speed of the underwater robot, the water flow in the area is considered to be strong, otherwise it is called weak flow;
[0019] When strong flow areas and weak flow areas exist simultaneously in the environment, the flow field is called a mixed flow field;
[0020] The path length is taken as the main research indicator; suppose the path consists of a set of nodes T = {T1 ,T 2 ,...T N}, where N is the number of nodes; let the distance cost between two adjacent nodes Ti and Ti+1 be τ i =||T i+1 -T i || 2 , define the cost function as:
[0021]
[0022] sC 1 :
[0023] C 2 :T 1 =T start ,T N =T goal
[0024] C 3 :
[0025] In the above formula, ||·|| represents the Euclidean distance; C 1 Represents node T i Must be located in flow field D; C 2 Indicates that the initial point and final point of the function must be the starting point and target point of the plan; C 3 Indicates that the task time must be within the specified time interval T.
[0026] In a preferred embodiment, the specific implementation method of step S2 is: using the historical data of the flow field in the environment, numerically simulating the flow field through CFD, and recording the flow velocity and direction in the flow field in the form of a grid; building a simulation database of the flow field for all working conditions;
[0027] The velocity vectors at adjacent grid points in the flow field are usually similar. Assume that within the robot deployment time, the flow field is divided into a finite number of cells The union of all cells is the flow field Let F 0 : Represents the velocity information of all discrete grid points in the flow field within the time interval T. Then any unit R in the grid i The speed information is Represents the position x and time The velocity vector on ;
[0028] right and represents the flow field, and Respectively represent the magnitude and direction of the velocity vector. The uncertainty of the velocity vector is expressed as:
[0029]
[0030] In a preferred embodiment, the specific implementation of step S3 is: before deploying the underwater robot, the real-time boundary flow velocity information of the flow field in the environment is obtained, and the flow field simulation database is imported to obtain the real-time flow field velocity vector information.
[0031] In a preferred embodiment, the specific implementation method of step S4 is: obtaining the initial position of the underwater robot from the environment, setting the target position that the underwater robot needs to reach, so as to improve the rapid exploration random number algorithm and obtain a theoretically feasible initial path to reach the target point.
[0032] In a preferred embodiment, the specific implementation of step S5 is as follows: for the flow field grids through which the initial path passes, based on the velocity vector information in each grid, determine whether the path is physically feasible; and separate physically feasible areas from physically infeasible areas.
[0033] In a preferred embodiment, the specific implementation method of step S6 is: for the path in the physically infeasible area, the starting point and the end point of the path entering the infeasible area are recorded, and the physically feasible path is searched according to the flow velocity vector information in the grid. When the search reaches the end point, the path part in the infeasible area is replaced with the newly corrected path part.
[0034] In a preferred embodiment, the specific implementation of step S7 is: calculating the total length of the path, and when the path is completely physically feasible and the path distance is the shortest, outputting the path as the optimal path.
[0035] Compared with the prior art, the present invention has the following beneficial effects:
[0036] 1. Combining CFD technology with the ocean current prediction model (ROMS) to build a flow field database for the operating environment, guide the underwater robot to plan the path, and avoid strong current areas that cannot be passed through.
[0037] 2. During path planning, an initial solution is quickly obtained through global sampling, and an optimization method is proposed to optimize the path, which can avoid the time-consuming process caused by a large number of iterations.
[0038] 3. Design a local correction module based on the flow field database to obtain a physically feasible solution for the local strong flow areas in the flow field by identifying and correcting them. BRIEF DESCRIPTION OF THE DRAWINGS
[0039] Figure 1 A flowchart of a preferred embodiment of the present invention;
[0040] Figure 2 A feasible range diagram of paths provided for a preferred embodiment of the present invention;
[0041] Figure 3 A pseudo code diagram of the method provided for the preferred embodiment of the present invention;
[0042] Figure 4 A schematic diagram of global sampling and optimization provided for a preferred embodiment of the present invention;
[0043] Figure 5 A local directed expansion graph provided for a preferred embodiment of the present invention;
[0044] Figure 6 A graph showing different step lengths and computational efficiency results provided by a preferred embodiment of the present invention;
[0045] Figure 7 A simulated flow field provided for a preferred embodiment of the present invention;
[0046] Figure 8 A path planning result diagram provided for a preferred embodiment of the present invention, wherein (a) is a planning task in the downstream direction, and (b) is a planning task in the upstream direction. DETAILED DESCRIPTION
[0047] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.
[0048] It should be noted that the following detailed descriptions are illustrative and are intended to provide further explanation of the present application. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present application belongs.
[0049] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application; as used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, they indicate the presence of features, steps, operations, devices, components and / or their combinations.
[0050] A path planning method based on flow field data drive, reference Figure 1-8 , the method comprises the following steps:
[0051] Step S1: Model the underwater vehicle and its environment according to the mission requirements, so that it can output the position observation value of the vehicle and the environmental state at the next moment;
[0052] In this case, we first establish the problem description in the shallow water environment as follows:
[0053] When an underwater robot is operating around an underwater structure, it should move according to instructions near obstacles to avoid collision with obstacles. There is an obstacle R : and time-varying flow field in and represents the time interval considered, and all position information in the flow field is observable. The underwater robot runs in D, and its path planning task is defined as: finding a reachable and safe path ε from the starting point to the destination in the flow field D. The path consists of a set of subpaths connected by control nodes, that is,
[0054] If the water flow velocity in the environment is The speed of the underwater robot in still water is It can be seen that the actual speed of the underwater robot in the environment is:
[0055]
[0056] The area with higher velocity in the flow field is defined as follows:
[0057] Definition 1: When the water flow speed is greater than the speed of the underwater robot, we consider the water flow in this area to be a strong flow, otherwise it is called a weak flow.
[0058] Definition 2: When there are strong flow areas and weak flow areas in the environment at the same time, the flow field is called a mixed flow field.
[0059] Evaluation of path quality involves multiple aspects such as path length and time consumption. Generally, a specific cost function can be designed as the goal of optimizing the path. In this study, we use path length as the main research indicator. Suppose the path consists of a set of nodes T = {T 1 ,T 2 ,...T N}, where N is the number of nodes. Let the distance cost between two adjacent nodes Ti and Ti+1 be τ i =||T i+1 -T i || 2 , define our cost function as:
[0060]
[0061] sC 1 :
[0062] C 2 :T 1 =T start ,TN =T goal
[0063] C 3 :
[0064] In the above formula, ||·|| represents the Euclidean distance; C1 means that the node Ti must be located in the flow field D; C2 means that the initial point and the final point of the function must be the starting point and the target point of the plan; C3 means that the task time must be within the specified time interval T.
[0065] Step S2: Obtain the flow field boundary velocity vector information in the environment, use it as the velocity boundary value to perform CFD numerical simulation, and establish a simulation database of flow field velocity information under all working conditions, specifically:
[0066] The velocity vectors of the flow field at adjacent grid points are usually similar, so for the flow field Assuming that within the robot deployment time, the flow field can be divided into a finite number of cells The union of all cells is the flow field Let F 0 : Represents the velocity information of all discrete grid points in the flow field within the time interval T, then the velocity information of any unit Ri in the grid is Represents the position x and time The velocity vector on .
[0067] The velocity information can be obtained by predicting the historical data of the flow field. For tidal estuaries and oceans, it can also be obtained through the ocean current prediction model. is not a definite value. In order to deal with this uncertainty, a method is designed to and represents the flow field, and Respectively represent the magnitude and direction of the velocity vector. The uncertainty of the velocity vector is expressed as:
[0068]
[0069] The method of obtaining flow field information should be determined according to the actual situation, including two typical situations:
[0070] Case 1: The free area far away from obstacles is usually large in area, and the overall flow field variation range is not large. Traditional prediction methods such as ocean current models can be used to obtain better data.
[0071] Case 2: The area near the obstacle is affected by the specific obstacle boundary. The velocity gradient of this part of the flow field is large, and there are local velocity mutations or turbulence. Traditional prediction methods are difficult to obtain accurate information, so other methods are needed.
[0072] As an important means of flow field analysis, CFD (Computational Fluid Dynamics) has been widely used in the shape optimization of underwater robots, but it is rarely used in the path planning of underwater robots. This study attempts to build a local flow field database simulated by CFD to provide small-scale flow field information for underwater robot path planning, which is also a motivation for this study.
[0073] In tidal estuaries or offshore areas, a complete cycle of tidal historical data can be used as the boundary conditions for local area simulation, and a simulation database can be established according to a given time interval T. This process can be completed on the PC as preliminary work before deploying the underwater robot. The Regional Ocean Modeling System (ROMS) is widely used in the simulation of ocean circulation, estuaries, waves, marine sediment and other processes. The tidal data required in this paper is also provided by ROMS.
[0074] Step S3: In actual scenarios, real-time hydrological and meteorological information is collected and imported into a flow field simulation database to obtain flow velocity information of all areas in the flow field;
[0075] Step S4: Obtain the current position of the underwater robot from the environment model, perform global sampling and path optimization for the set target point to obtain a theoretically feasible initial path, specifically:
[0076] This paper proposes an optimization algorithm Curtate-RRT, which requires first obtaining a feasible initial solution ε 0 : This solution is obtained by the classic RRT algorithm and is a broken line formed by a finite number of nodes with a distance d. t is the path of the initial solution after t optimizations. The optimization process of the algorithm is divided into two key parts: the first part is sampling the global range to obtain a feasible theoretical optimal solution, which includes two main modules: path contraction and corner point optimization. The second part is to perform feasibility testing on the theoretical optimal solution. If there is an infeasible part, it is locally corrected through directed expansion. The pseudo code of the algorithm is as follows: Figure 3 shown.
[0077] from Figure 3It can be seen that the first half of the algorithm (the first to the eleventh line) is consistent with the steps of the RRT algorithm. A point Xrand is randomly sampled in the workspace S, the node Xnearst closest to Xrand is searched on the tree, and a new point Xnew is generated at a fixed step size. A collision check is performed on the point Xnew. If no collision occurs, it is added to the tree and its parent node is recorded. Given a small threshold Calculate the Euclidean distance Δd between the target point Xgoal and Xnew = ||X goal -X new ||, if Δd≤δ, then the target point is found. The thirteenth line function T←LineCurate(X start ,X goal ,T,obstacle) is a path contraction module, which optimizes the initial solution to a transition solution close to the obstacle. This process will be completed in a very short time, and then the corner optimization module T←AngleOptimization(X start ,X goal ,T,obstacle) iteratively optimizes the corner points of the new path until the optimal solution is obtained. Lines 18 to 20 are local path corrections. If an infeasible part appears, T←LocalRewise(S,T,obstacle) is executed for local replanning. Otherwise, the program ends and returns the planning result.
[0078] The main modules of Curtate-RRT algorithm include:
[0079] (1) Path contraction
[0080] like Figure 4 As shown in the left figure, at the adjacent node P i With P i+ 1, with a step size of d 1 Equidistant uniform interpolation will add new points Add to the path point set, where d 1 <d. Renumber the expanded point set in order, starting from the starting point P start Start by connecting all remaining visible nodes, performing a collision check each time you connect. Take the last collision-free node as the next starting point, and search for the remaining nodes until you find the target point p. goal .
[0081] (2) Corner point optimization
[0082] like Figure 4 As shown in the figure on the right, suppose any node P except the starting point and the end point i (x,y), called P i is the middle node, and its previous point P i-1is the left node, and the next node is the right node. Use step size d 2 For point P i Front and back line segments and Equidistant interpolation, obtain two interpolation point sets and Point Set From the starting point Start by connecting all the points in another point set in sequence Each connection performs a collision check until a collision occurs. If the node before the collision Become the left node, Become an intermediate node, for the line segment and Repeat the previous steps until the middle node coincides with the right node, and the search ends. The shortest path after segmentation is the optimized path.
[0083] Step S5: Perform feasibility test on the initial path obtained in step 4 to analyze physically feasible areas and physically infeasible areas;
[0084] This step is aimed at the situation where the optimized path passes through the strong current area, such as Figure 5 As shown in the figure, after searching the cell water flow information on the path point one by one, determine whether it enters the strong flow area. If it does not pass through the strong flow area, the algorithm ends; otherwise, determine the entry point S0 and exit point Sn of the strong flow area. For the part in the strong flow area, detect the flow field cells passed by the path in turn to determine whether the path direction is within the feasible range of the cell, and analyze according to the following two situations:
[0085] Case 1): If all the directions of the path satisfy the feasible region of the cell, the path is considered to be a feasible optimal solution and the algorithm ends.
[0086] Case 2): If there is a part that does not conform to the feasible interval of the cell, then S0 and Sn are the starting point and target point respectively, record the reachable point set {S1,…Sn}, n∈R+ of the current point in the next step, and search the point set that can be reached by each point in the point set in the next step in turn. Nodes beyond the boundary of the strong flow area are not included. When the target point appears in the reachable point set, it is considered that a feasible solution has been found. At the end of the traversal, all feasible solutions are compared, and the solution with the shortest distance is taken as the final solution, and the algorithm ends.
[0087] Step S7: output the optimal reachable path according to the path length.
Claims
1. A data-driven shallow water underwater robot path planning method, characterized in that: The steps include: Step S1: Model the underwater vehicle and its environment according to the mission requirements, so that it can output the position observation value of the vehicle and the environmental state at the next moment; Step S2: Obtain the flow field boundary velocity vector information in the environment, use it as the velocity boundary value to perform CFD numerical simulation, and establish a simulation database of flow field velocity information under all working conditions; Step S3: In the actual scenario, real-time hydrological and meteorological information is collected and imported into the flow field simulation database to obtain flow velocity information of all areas in the flow field; Step S4: Obtain the current position of the underwater robot from the environment model, perform global sampling and path optimization for the set target point, and obtain a theoretically feasible initial path; Step S5: Perform feasibility test on the initial path obtained in step 4 to analyze physically feasible areas and physically infeasible areas; Step S6: Perform local directed expansion on the physically infeasible area and correct it to a physically feasible path; Step S7: Output the optimal reachable path according to the path length.
2. A data-driven shallow water underwater robot path planning method according to claim 1, characterized in that: The specific implementation of step S1 is as follows: setting basic information of the environment, mainly including: the boundary size of the environment horizontal plane range, the grid size of the environment, the location information of obstacles, the deployment time of the path planning task, the interaction frequency between the environment and the underwater robot, and defining its action space and observation space; When an underwater robot is operating around an underwater structure, it should move according to instructions near obstacles to avoid collision with obstacles; There are obstacles and time-varying flow field in and represents the time interval considered, all position information in the flow field is observable; the underwater robot runs at In the flow field, the path planning task is defined as: Find a reachable safe path ε from the starting point to the destination, which is composed of a set of subpaths connected by control nodes, that is, If the water flow velocity in the environment is The speed of the underwater robot in still water is It can be seen that the actual speed of the underwater robot in the environment is: The area with higher velocity in the convection field: When the water flow speed is greater than the speed of the underwater robot, the water flow in the area is considered to be strong, otherwise it is called weak flow; When there are strong flow areas and weak flow areas in the environment at the same time, the flow field is called a mixed flow field; The path length is taken as the main research indicator; suppose the path consists of a set of nodes T = {T1, T2, ...T N }, where N is the number of nodes; let the distance cost between two adjacent nodes Ti and Ti+1 be τ i =||T i+1 -T i ||2, define the cost function as: C2:T1=T start ,T N =T goal In the above formula, ||·|| represents the Euclidean distance; C1 represents the node T i It must be located in the flow field D; C2 means that the initial point and final point of the function must be the starting point and target point of the plan; C3 means that the task time must be within the specified time interval T.
3. A data-driven shallow water underwater robot path planning method according to claim 1, characterized in that: The specific implementation method of step S2 is: using the historical data of the flow field in the environment, numerically simulating the flow field through CFD, and recording the flow velocity and direction in the flow field in the form of a grid; building a simulation database of the flow field for all working conditions; The velocity vectors at adjacent grid points in the flow field are usually similar. Assume that within the robot deployment time, the flow field is divided into a finite number of cells The union of all cells is the flow field set up Represents the velocity information of all discrete grid points in the flow field within the time interval T. Then any unit R in the grid i The speed information is Represents the position x and time The velocity vector on ; right and represents the flow field, and Respectively represent the magnitude and direction of the velocity vector; the uncertainty of the velocity vector is expressed as:
4. A data-driven shallow water underwater robot path planning method according to claim 1, characterized in that: The specific implementation method of step S3 is: before deploying the underwater robot, obtain the real-time boundary flow velocity information of the flow field in the environment, import it into the flow field simulation database, and obtain the real-time flow field velocity vector information.
5. A data-driven shallow water underwater robot path planning method according to claim 1, characterized in that: The specific implementation method of step S4 is: obtaining the initial position of the underwater robot from the environment, setting the target position that the underwater robot needs to reach, so as to improve the rapid exploration random number algorithm and obtain a theoretically feasible initial path to reach the target point.
6. A data-driven shallow water underwater robot path planning method according to claim 1, characterized in that: The specific implementation of step S5 is as follows: for the flow field grids that the initial path passes through, based on the flow velocity vector information in each grid, determine whether the path is physically feasible; and separate physically feasible areas from physically infeasible areas.
7. A data-driven shallow water underwater robot path planning method according to claim 1, characterized in that: The specific implementation method of step S6 is: for the path in the physically infeasible area, record the starting point and the end point of the path entering the infeasible area, search for the physically feasible path according to the flow velocity vector information in the grid, and when searching to the end point, replace the path part in the infeasible area with the newly corrected path part.
8. A data-driven shallow water underwater robot path planning method according to claim 1, characterized in that: The specific implementation of step S7 is: calculating the total length of the path, and when the path is completely physically feasible and the path distance is the shortest, outputting the path as the optimal path.
