Underwater environment autonomous exploration method, device and system, and storage medium

Abstract

The invention discloses an underwater environment autonomous exploration method, device and system, and a storage medium, and the method comprises the steps: S1, constructing a random tree in an occupation grid map constructed by an underwater vehicle through a multi-beam sonar through a boundary point detector, so as to search a boundary point; s2, clustering the detected boundary points through a filter; and S3, the clustered boundary points are evaluated through a utility function and are allocated to the aircraft, and the aircraft starts to move towards the boundary points for exploration. By adopting the technical scheme of the invention, the problems that an omnidirectional method is easy to stop in an underwater environment and the exploration efficiency is low are solved.

Description

Technical Field

[0001] This invention belongs to the field of marine emergency search and rescue technology, and in particular relates to a method, device, system, and storage medium for autonomous underwater environment exploration. Background Technology

[0002] With the rapid development of my country's marine economy, marine accidents are occurring frequently. Therefore, improving marine emergency search and rescue capabilities has become crucial for ensuring marine safety. Since prior information about the target waters is often lacking when conducting underwater emergency search and rescue missions, technologies for sensing and exploring unknown underwater environments have become key to enhancing marine emergency search and rescue capabilities.

[0003] Employing rapidly developing underwater vehicles, including remotely operated underwater vehicles (ROVs) and autonomous underwater vehicles (AUVs), offers numerous advantages such as low cost, high efficiency, and strong maneuverability. Simultaneously, by carrying multibeam sonar, the underwater environment of the target sea area can be perceived, acquiring hydrological parameters and the distribution of stranded personnel—essential search and rescue conditions—to assist subsequent search and rescue operations. The key technologies of this method are as follows:

[0004] The underwater vehicle is equipped with a multibeam sonar. The sonar emits sound waves towards the target area via a transducer. These sound waves propagate through the water, reflect off objects, and generate echoes. The sonar receives these echoes, processes the data, and obtains the position and shape information of underwater objects to construct an underwater two-dimensional grid map, enabling underwater environmental perception. Based on the environmental information provided by the local underwater two-dimensional grid map, the underwater vehicle uses a random tree algorithm to obtain a set of exploration target points at its current location. A utility function is used to filter the target point set, obtaining the optimal target point and assigning it to the underwater vehicle. The underwater vehicle then rapidly moves to the optimal target point, autonomously exploring the target area and ultimately obtaining a global underwater two-dimensional grid map of the target area.

[0005] Traditional underwater mapping algorithms use underwater images collected by visual cameras for mapping. When the target water area is turbid or encounters severe weather, the underwater visibility decreases, which reduces the quality of the underwater images collected by the visual cameras. This affects the mapping distance, quality, and accuracy to a certain extent, making mapping failure more likely and preventing underwater vehicles from completing perception and exploration tasks.

[0006] Compared to lidar, forward-looking multibeam sonar has a fan-shaped detection area rather than a circle. However, this fan-shaped perception brings a unique problem: the underwater vehicle's current position (i.e., the center of the fan) is often identified as a boundary point. Since this point is zero distance from itself, it is often selected as the optimal target by the target assigner, causing the vehicle to remain stationary. If this point is simply removed, the vehicle's previous position may be repeatedly selected as the target point, resulting in the vehicle moving back and forth. This movement not only reduces exploration efficiency but may also make the vehicle unstable or even out of control.

[0007] In existing underwater exploration missions, ground stations use path planning algorithms to solve for the trajectory route based on the real-time environment of the target water area, and then transmit the information to the underwater vehicle. This causes the underwater vehicle to rely on the trajectory route specified by the host computer at the ground station, lacking autonomy and thus reducing the efficiency of the exploration mission. At the same time, the real-time environment of the target water area collected by the ground station has a certain lag, which makes it impossible to guarantee the timeliness of the trajectory route solved by the host computer. As a result, it is difficult for the underwater vehicle to navigate safely in complex and ever-changing target water areas. Summary of the Invention

[0008] The technical problem to be solved by the present invention is to provide a method, device, system, and storage medium for autonomous exploration of underwater environments.

[0009] To achieve the above objectives, the present invention adopts the following technical solution:

[0010] A method for autonomous underwater environment exploration, comprising:

[0011] Step S1: Using a boundary point detector, construct a random tree in the occupancy grid map built by the underwater vehicle using multibeam sonar to search for boundary points;

[0012] Step S2: Cluster the detected boundary points using filters;

[0013] Step S3: Using the target assigner, the clustered boundary points are evaluated using a utility function and assigned to the vehicle. The vehicle then begins to move towards the boundary points to explore.

[0014] Preferably, the boundary point detector includes a local boundary point detector and a global boundary point detector. The local boundary point detector uses the Fast Exploration Sector Tree (RST) algorithm to extend forward within the fan-shaped perception range of the UUV, guiding the detected boundary points to be distributed in the front and side areas. The global boundary point detector uses the Sparse Exploration Random Tree (SRT) algorithm to enable the UUV to conduct a comprehensive exploration of the entire area.

[0015] Preferably, in step S3, when allocating tasks, the priority of local boundary points is higher than that of global boundary points; the priority between local and global points is determined by the utility function.

[0016] The present invention also provides an autonomous underwater environment exploration device, comprising:

[0017] The first processing module is used to build a random tree in the occupancy grid map constructed by the underwater vehicle using multibeam sonar to search for boundary points by using a boundary point detector.

[0018] The second processing module is used to cluster the detected boundary points through filters;

[0019] The third processing module, through the target allocator, evaluates the clustered boundary points using a utility function and assigns them to the vehicle. The vehicle then begins to move towards the boundary points to explore.

[0020] Preferably, the boundary point detector includes a local boundary point detector and a global boundary point detector. The local boundary point detector uses the Fast Exploration Sector Tree (RST) algorithm to extend forward within the fan-shaped perception range of the UUV, guiding the detected boundary points to be distributed in the front and side areas. The global boundary point detector uses the Sparse Exploration Random Tree (SRT) algorithm to enable the UUV to conduct a comprehensive exploration of the entire area.

[0021] Preferably, when allocating tasks, the third processing module prioritizes local boundary points over global boundary points; the priority between local and global points is determined by a utility function.

[0022] The present invention also provides an autonomous underwater environment exploration system, comprising: a memory and a processor, wherein the memory stores a computer program executed by the processor, and the computer program executes an autonomous underwater environment exploration method when executed by the processor.

[0023] The present invention also provides a storage medium storing a computer program, which executes an autonomous underwater environment exploration method when running.

[0024] This invention fully considers the detection range and signal characteristics of forward-looking sonar, solving the problems of omnidirectional methods easily getting stuck and having low exploration efficiency in underwater environments. This invention combines global and local strategies, improving exploration efficiency and completeness while achieving obstacle avoidance and exploration functions in enclosed underwater environments. Attached Figure Description

[0025] 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 embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0026] Figure 1 This is a flowchart of the underwater environment autonomous exploration method according to an embodiment of the present invention; Figure 2 A two-dimensional schematic diagram of the detection effect of multibeam sonar; Figure 3 This is a schematic diagram of information gain at different locations. Detailed Implementation

[0027] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0028] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0029] Example 1:

[0030] like Figure 1 As shown, this embodiment of the invention provides a method for autonomous exploration of underwater environments, including:

[0031] Step S1: Using boundary point detectors, construct a random tree in the occupancy grid map built by the underwater vehicle using multibeam sonar to search for boundary points; wherein, the boundary point detectors include: local boundary point detectors and global boundary point detectors;

[0032] Step S2: Cluster the detected boundary points using filters;

[0033] Step S3: The clustered boundary points are evaluated using a utility function by the target assigner and assigned to the aircraft; the aircraft then begins to move toward the boundary points to explore; during the exploration process, the airborne forward-looking multibeam sonar will continuously scan and explore the surrounding environment within its perception range.

[0034] Furthermore, the local boundary point detector focuses on the area near the vehicle's current position to accelerate the exploration process; while the global boundary point detector scans the entire exploration space to ensure the integrity of the exploration; the filter is responsible for deleting invalid and outdated boundary points to ensure the dynamic feasibility of the boundary points; the target assigner receives the clustered boundary points output by the filter and assigns them to the vehicle for exploration.

[0035] As one embodiment of the present invention, the working process of the local boundary point detector in step S1 is as follows:

[0036] To address the limitations of sonar sensing, this invention proposes a Rapidly-exploring Sector-shaped Tree (RST) algorithm. This algorithm extends forward within the UUV's sector-shaped sensing range, guiding the distribution of detected boundary points in the forward and lateral regions. An overview of the RST algorithm is shown below. Whenever a new target point is detected or the UUV's position changes, the RST tree is reset to ensure that the tree remains within the UUV's current sensing area.

[0037] The algorithm flow for quickly exploring sector trees is shown below:

[0038] ●Step 1: Initialize the vertex set and edge set, let V = x init , E = φ.

[0039] ●Step 2: Randomly sample a point from free space, denoted as x. rand ←SampleSector

[0040] ●Step 3: Find the distance x in the current graph G(V,E) rand The most recent existing node, obtained

[0041] x nearest ←Nearest(G(V,E),x rand )

[0042] ●Step 4: Based on the growth step size parameter η, from x nearest Chao X rand New extended nodes are generated in the direction

[0043] x new ←Steer(x nearest ,x rand ,η)

[0044] ●Step 5: For (x) nearest ,x new Perform a voxel check (GridCheck) on the path between the voxels: If the check result is -1, meaning the voxel belongs to an unknown region, then perform a point publishing operation.

[0045] PublishedPoint(x new )

[0046] ●Step 6: If the check result is 1, meaning the voxel belongs to the known region, then perform the following operation:

[0047] x new Add to node collection:

[0048] V←V∪{x new}

[0049] edge(x) nearest ,x new Add to edge set:

[0050] E←E∪{(x nearest ,x new )}

[0051] ●Step 7: Repeat Step 2 to Step 6 until the exploration termination condition is met.

[0052] in:

[0053] SampleSector function:

[0054] This function is used to sample random points independently and identically distributed within the sonar detection area.

[0055] Quickly explore a sector tree with an initial vertex V = x init Initially, the set of edges is E = φ.

[0056] In each iteration, random points The definition is as follows:

[0057]

[0058] This method ensures that the sampling points are located within the sonar sector detection region V. The SampleSector function ensures that the RST tree grows within region V.

[0059] Nearest function:

[0060] This function takes a graph G = (V, E) and a point x in free space as input, and returns a point v ∈ V such that:

[0061]

[0062] If multiple points v have the same minimum distance to x, return the first point encountered.

[0063] Steer function:

[0064] The function takes two points Given input, return a point. Make:

[0065]

[0066] Where η>0 is the tree growth step size. A larger η value means faster tree expansion.

[0067] GridCheck function:

[0068] This function uses a map The input consists of two points x and y in free space. If there are obstacle pixels between the two points, the function returns 0; if there are unknown pixels, it returns -1; otherwise, it returns 1, indicating that the two points are located in the known free space.

[0069] PublishPoint function:

[0070] This function sends the detected boundary points to the filtering module for processing.

[0071] As one embodiment of the present invention, in step S1, the global boundary point detector operates as follows:

[0072] To ensure the completeness of the exploration, this invention proposes a Sparse-Exploring Random Tree (SRT) algorithm to construct a global boundary point detector, thereby ensuring that the UUV can fully explore the entire region.

[0073] The steps of the Sparse-Exploring Random Tree (SRT) algorithm are as follows:

[0074] ●Step 1: Initialize the vertex set v←x init That is, it only contains the initial position point; the edge set E←φ is initialized to be empty.

[0075] ●Step 2: Enter the loop process and continue to build the exploration tree until the external stopping conditions are met (such as time limit, exploration completion, etc.).

[0076] ●Step 3: Call the function SampleFree to randomly sample a point from free space, denoted as x. rand .

[0077] ●Step 4: Call the function Nearest(G(V,E),x rand Find the distance x in the existing tree structure. rand The nearest vertex, denoted as xnearest .

[0078] ●Step 5: Call the function Steer(x) nearest ,x rand ,η), from x nearest To x rand Expand the direction by a step size η to generate a new point x. new .

[0079] ●Step 6: Call the function SparseCheck(map, x nearest ,x new Perform sparsity detection on the path:

[0080] ○ If the return value is -1, it indicates that the path area is an unknown area, then the function PublishedPoint(x) will be executed. new ), and publish this point as a candidate boundary point;

[0081] ○ If the return value is 1, it means that the path is entirely within a known free region or sparse region, then perform the following operation:

[0082] ■Step 6.1: Add the new point to the vertex set: V←V∪{x new};

[0083] ■Step 6.2: Add the connecting edge to the edge set: E←E∪{(x nearest ,x new )}.

[0084] ●Step 7: Return to Step 2 and continue building the exploration tree.

[0085] Compared with the RST algorithm, the SRT algorithm has the following three main differences:

[0086] First, the SRT algorithm does not require resetting the tree structure. The SRT tree it generates continues to grow on the map throughout the exploration process, thus ensuring coverage of all explored areas.

[0087] Secondly, the SRT algorithm replaces the SampleSector function in RST with the SampleFree function, enabling the algorithm to detect boundary points located behind UUVs.

[0088] SampleFree function:

[0089] This function is used to extract data from a map. Sampling random points in free space with moderate probability, i.e. returning random sampling points that are independent and identically distributed in free space.

[0090] Furthermore, the SRT algorithm replaces the GridCheck function in RST with the SparseCheck function, allowing the spanning tree to continue growing in sparse regions, such as... Figure 2 As shown.

[0091] Because sonar signals gradually become sparser with increasing distance during propagation, known and unknown regions are interspersed along the sonar detection boundary. To enable the SRT tree to continue growing in these sparse regions, the SparseCheck function is introduced.

[0092] SparseCheck function:

[0093] This function takes a map as input. And two points x and y in free space,

[0094] ● If there is an obstacle pixel between the two points, the return value is 0;

[0095] ● If there exists an unknown region with an area greater than n (i.e., the number of unknown pixels equals n), then return -1;

[0096] ●If the area between two points is a known free space or an unknown region with an area less than n, then return 1.

[0097] This mechanism ensures that the SRT algorithm can grow in sparse regions that are not entirely known, thereby improving the overall exploration efficiency.

[0098] In one embodiment of the present invention, the working process of the filter in step S2 is as follows:

[0099] The filter receives boundary points detected jointly by the local and global boundary point detectors. It first clusters all boundary points, retaining only the center of each cluster and discarding the rest. This clustering and discarding process is crucial for reducing the number of boundary points, as the global and local detectors may generate a large number of boundary points that are very close to each other. Sending these points directly to the target assigner would result in unnecessary computational resource consumption without providing additional map information. Furthermore, the filter removes infeasible or outdated boundary points in each iteration to ensure the validity of the target points.

[0100] In one embodiment of the present invention, in step S3, the target allocator operates as follows: the target allocator receives boundary points from the filter and assigns them to a UUV for exploration. During task allocation, local boundary points have a higher priority than global boundary points. The priority between local and global points is determined by a utility function.

[0101] For mobile robot platforms and drones, navigation costs are typically the norm of the distance between the current vehicle position and the exploration point. However, for unmanned underwater vehicles (UUVs), since their operating environment is water, the angle between the current vehicle position and the exploration point must also be considered. Because water is denser and has higher drag than air, UUVs consume more energy when turning, resulting in relatively lower turning accuracy. Furthermore, the propulsion and control mechanisms of UUVs limit their minimum turning radius; sharp turns or frequent maneuvers can cause the UUV to tilt or even become unstable.

[0102] Therefore, in rapidly exploring a sector tree, to more realistically measure the turning cost for a given target point... and the current location of the UUV The embodiments of this invention propose the following new utility function:

[0103]

[0104] Wherein, parameter λ is used to increase the importance of information gain relative to navigation cost in the evaluation; g(x T ,x R () indicates hysteresis gain:

[0105]

[0106] Among them, positive number g rad It is set based on experience, I s (x t ,ψ t ),like Figure 3 As shown:

[0107] The definition is as follows:

[0108]

[0109] Where r represents the sonar's detection range, α represents the sonar's detection angle; S unknown It is the area of ​​the unknown region within the range of sonar detection.

[0110] Let P = (x1, x2) be an unknown point on map M. If the condition is satisfied... Then point P is located at S. unknown within the area.

[0111] Where d(x) T ,x R ) represents the target point x T With current UUV position x R The Euclidean distance between them is defined as follows:

[0112]

[0113] θ(x T ,x R ,ψ t ) represents vector (x) T ,x R ) and the current yaw angle ψ of the UUV t The included angle between them is defined as follows:

[0114] θ(x T ,x R ,ψ t )=φ(x T ,x R )-ψ t

[0115] For sparse exploratory random trees, the utility function is defined similarly to that of mobile robotic platforms or drones, aiming to explore every possible region to the greatest extent possible. Its utility function is defined as follows:

[0116]

[0117] Among them, I c (x T ) is defined as:

[0118]

[0119] Among them, C unknown This represents the area of ​​the unknown region within the circular detection area.

[0120] The underwater vehicle selects the boundary point with the highest utility as the target point based on the utility function, and calculates the path to the target point using the A* algorithm, thereby moving towards the boundary point for exploration (the output is the navigation target of the underwater vehicle).

[0121] Example 2:

[0122] This invention also provides an autonomous underwater environment exploration device, comprising:

[0123] The first processing module is used to build a random tree in the occupancy grid map constructed by the underwater vehicle using multibeam sonar to search for boundary points by using a boundary point detector.

[0124] The second processing module is used to cluster the detected boundary points through filters;

[0125] The third processing module, through the target allocator, evaluates the clustered boundary points using a utility function and assigns them to the vehicle. The vehicle then begins to move towards the boundary points to explore.

[0126] As one embodiment of the present invention, the boundary point detector includes: a local boundary point detector and a global boundary point detector; wherein, the local boundary point detector expands forward within the fan-shaped perception range of the UUV through the Fast Exploration Sector Tree (RST) algorithm, guiding the detected boundary points to be distributed in the front and side regions; the global boundary point detector enables the UUV to conduct a comprehensive exploration of the entire region through the Sparse Exploration Random Tree (SRT) algorithm.

[0127] In one embodiment of the present invention, when the third processing module allocates tasks, the priority of local boundary points is higher than that of global boundary points; the priority between local points and global points is determined by a utility function.

[0128] Example 3:

[0129] This invention also provides an autonomous underwater environment exploration system, comprising: a memory and a processor, wherein the memory stores a computer program executed by the processor, and the computer program executes an autonomous underwater environment exploration method when run by the processor.

[0130] Example 4:

[0131] This invention also provides a storage medium storing a computer program that executes an autonomous underwater environment exploration method during runtime.

[0132] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A method for autonomous exploration of underwater environments, characterized in that, include: Step S1: Using a boundary point detector, construct a random tree in the occupancy grid map built by the underwater vehicle using multibeam sonar to search for boundary points; Step S2: Cluster the detected boundary points using filters; Step S3: Using the target assigner, the clustered boundary points are evaluated using a utility function and assigned to the vehicle. The vehicle then begins to move towards the boundary points to explore.

2. The underwater environment autonomous exploration method as described in claim 1, characterized in that, Boundary point detectors include local boundary point detectors and global boundary point detectors. The local boundary point detector uses the Fast Exploration Sector Tree (RST) algorithm to extend forward within the fan-shaped perception range of the UUV, guiding the detected boundary points to be distributed in the front and side areas. The global boundary point detector uses the Sparse Exploration Random Tree (SRT) algorithm to enable the UUV to conduct a comprehensive exploration of the entire area.

3. The underwater environment autonomous exploration method as described in claim 1, characterized in that, In step S3, when assigning tasks, local boundary points have a higher priority than global boundary points; the priority between local and global points is determined by the utility function.

4. An autonomous underwater environment exploration device, characterized in that, include: The first processing module is used to build a random tree in the occupancy grid map constructed by the underwater vehicle using multibeam sonar to search for boundary points by using a boundary point detector. The second processing module is used to cluster the detected boundary points through filters; The third processing module is used to evaluate the clustered boundary points through a utility function using a target allocator and assign them to the vehicle. The vehicle then begins to move towards the boundary points to explore.

5. The underwater environment autonomous exploration device as described in claim 4, characterized in that, Boundary point detectors include local boundary point detectors and global boundary point detectors. The local boundary point detector uses the Fast Exploration Sector Tree (RST) algorithm to extend forward within the fan-shaped perception range of the UUV, guiding the detected boundary points to be distributed in the front and side areas. The global boundary point detector uses the Sparse Exploration Random Tree (SRT) algorithm to enable the UUV to conduct a comprehensive exploration of the entire area.

6. The underwater environment autonomous exploration device as described in claim 5, characterized in that, When allocating tasks, the third processing module prioritizes local boundary points over global boundary points; the priority between local and global points is determined by the utility function.

7. An autonomous underwater environment exploration system, characterized in that, include: A memory and a processor, wherein the memory stores a computer program executed by the processor, the computer program executing the underwater environment autonomous exploration method as described in any one of claims 1-3 when run by the processor.

8. A storage medium, characterized in that, The storage medium stores a computer program, which executes the underwater environment autonomous exploration method as described in any one of claims 1-3 when it runs.

Images