An active mapping method for unknown underwater environment based on hovering underwater vehicle
By combining a hovering underwater vehicle with sensor and planning control modules, environmental data is acquired in real time, the best viewpoint is determined, and a path is planned. This solves the problems of accuracy and autonomy in underwater environmental mapping, and enables efficient and rapid underwater environmental mapping and dynamic updates.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- HARBIN ENG UNIV
- Filing Date
- 2023-04-24
- Publication Date
- 2026-07-10
Smart Images
Figure CN116878507B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to an active mapping method for unknown underwater environments based on hovering underwater vehicles, belonging to the field of robotic environment exploration technology. Background Technology
[0002] In recent years, with the continuous development and widespread application of advanced technologies such as intelligent control, navigation and communication, and detection and identification, underwater vehicles have demonstrated outstanding capabilities in tasks such as seabed mineral resource exploration, underwater structure monitoring, and underwater environmental exploration. Most existing underwater vehicles use onboard sensors to completely cover a target sea area for tasks such as seabed mapping and target search. However, this method is no longer the optimal choice when higher-precision environmental modeling of certain special areas is required.
[0003] On the one hand, to avoid rugged seabed terrain and considering the limitations of underwater vehicles' obstacle avoidance capabilities, existing methods require underwater vehicles to maintain a horizontal position at a certain height above the seabed, making optical sensors almost unusable. On the other hand, existing underwater environment modeling methods passively collect and store sensor data for map construction, lacking sensor feedback, resulting in uncertain data quality and potentially leading to repeated mission execution, increasing overall costs. Furthermore, existing underwater environment modeling methods largely rely on manually preset paths, lacking sensor feedback and decision-making behavior towards better viewpoints during mission execution, thus limiting the autonomy and intelligence of underwater vehicles. For example, existing underwater environment modeling methods collect data according to a preset path. If the sensor data collection task is completed when the path is 60% complete, the underwater vehicle will still complete the remaining 40% of the path to continue collecting redundant data; if the path is 100% complete, although there is still a lack of sensor data, the underwater vehicle will stop the mission without human guidance.
[0004] Furthermore, in the current process of active mapping of unknown underwater environments using view planning methods, the most commonly used frontier exploration methods lead to the underwater vehicle constantly moving towards new viewpoints, increasing the uncertainty of the constructed environment model and thus reducing mapping accuracy. Moreover, when determining the optimal next viewpoint, most existing methods design utility functions based on information gain theory. Currently, the utility function in these methods is mostly calculated using the distance and angle difference between the underwater vehicle and the viewpoint. However, such methods are prone to causing the underwater vehicle's path planning to get trapped in local optima, ultimately becoming a greedy algorithm.
[0005] Furthermore, while optical sensors can assist in active mapping of unknown underwater environments, acquiring better underwater environmental data at a lower cost, current optical mapping often involves a second dive based on an existing acoustic occupancy map to collect optical images of targets or obstacles. Due to underwater drift and poor underwater positioning accuracy, the point A reached by the vehicle based on the acoustic occupancy map may deviate from the actual point A, resulting in a mismatch between the acquired optical and acoustic images. Severe drift may necessitate multiple dives. Moreover, to avoid unsafe underwater collisions, the vehicle operates at a slow speed and has a short endurance. If a mission fails, it may need to recharge and dive again, consuming significant human and material resources.
[0006] In summary, based on the needs of underwater structure exploration and marine environment exploration, as well as the shortcomings of existing mapping methods, this invention aims to provide a method for underwater unknown environment exploration and active mapping based on hovering underwater vehicles.
[0007] It should be noted that the above content falls within the inventor's technical knowledge and does not necessarily constitute prior art. Summary of the Invention
[0008] To address the problems existing in the prior art, this invention provides an active mapping method for unknown underwater environments based on hovering underwater vehicles. This method can efficiently, quickly, and accurately construct and dynamically update maps, thereby assisting underwater vehicles in navigating target areas.
[0009] The present invention achieves the above objectives by adopting the following technical solutions:
[0010] A method for active mapping of unknown underwater environments based on hovering underwater vehicles includes the following steps:
[0011] S1. Obtain the current position and attitude information of the underwater vehicle through the navigation module, obtain the exploration range information of the shore-based setup through the communication module, and obtain environmental data through the sensor module;
[0012] Among them, environmental data includes acoustic image data collected by forward-looking sonar within its sensing range;
[0013] S2. The map generator creates a 3D raster map starting from the initial position of the underwater vehicle. The size of the 3D raster map area is the exploration range set on the shore. Based on the collected environmental data, the cells of the 3D raster map are divided into occupied cells, empty cells, and unknown cells.
[0014] S3. Use the advanced planner to perform uniform sampling on the current 3D grid map to obtain multiple next viewpoints. Compare the information gain values of each next viewpoint to determine the best next viewpoint.
[0015] S4. After obtaining the best viewpoint position published by the high-level planner using the low-level planner, calculate the safe and feasible path to the best next viewpoint, and publish the waypoints in the path at a certain control frequency.
[0016] S5. After obtaining the current position of the underwater vehicle and the desired trajectory points issued by the low-level planner using the guidance module, the desired heading is generated. After obtaining the current heading and desired heading issued by the guidance module, the low-level control module calculates and issues the rudder angle, so that the underwater vehicle travels along the calculated path.
[0017] S6. The underwater vehicle arrives at the new viewpoint according to the waypoints. The navigation module publishes the current position as the new viewpoint. The map generator at the new viewpoint updates the map based on the environmental data obtained by the sensor module.
[0018] S7. After the map is updated, repeat steps S3-S6, iterating sequentially until the map of the preset exploration area is completed.
[0019] Preferably, step S3 specifically includes:
[0020] S3.1. Obtain m next viewpoints by uniformly sampling around the current position i. Using a viewpoint p as a node, continue uniform sampling to obtain n child viewpoints. Viewpoints located in the occupying grid need to be removed from the next viewpoint and child viewpoints;
[0021] S3.2 Calculate the information gain values of the next viewpoint and the child viewpoint. The information gain value of the next viewpoint is the difference between the information entropy of the next viewpoint and the information entropy of the current position. The information gain value of the child viewpoint is the difference between the information entropy of the child viewpoint and the information entropy of the corresponding next viewpoint. Information entropy is the ratio of unknown information to known information obtained within the sensor module's sensing range at the current viewpoint position.
[0022] S3.3 Construct a utility function based on the information gain value of the next viewpoint and the information gain value of the child viewpoint. The next viewpoint that is most efficient is calculated using the utility function and then determined as the optimal next viewpoint.
[0023] The information gain value of the next viewpoint p at the current position i is Let E(x) be the information entropy of the next viewpoint p. i ) represents the information entropy of the current position i;
[0024] The information gain value of the child viewpoint is The information entropy of the next viewpoint p, The information entropy of the child viewpoint q of the next viewpoint p.
[0025] Preferably, in step S3.2, known information refers to the total volume of grid cells marked as occupied and empty, and unknown information refers to the total volume of grid cells marked as unknown.
[0026] Preferably, in step S1, the environmental data also includes optical image data collected by the optical sensor, which is used to perform coverage detection on the detection area of the optical sensor.
[0027] Preferably, in step S2, after viewing the occupied grid in the acoustic image using the optical image, the grid is marked as a viewed grid.
[0028] Preferably, in step S4, the method for calculating a safe and feasible path to the optimal next viewpoint is as follows:
[0029] Starting from the initial position of the aircraft, a tree is formed by sequentially identifying the best next viewpoints. The newly determined best next viewpoint is then incorporated into the tree. Using the newly determined best next viewpoint as the center, all nodes in the tree that are adjacent to the newly determined best next viewpoint within a preset radius are identified. It is then checked whether the path between the newly determined best next viewpoint and each adjacent node intersects with obstacles. If the path does not intersect with obstacles, the cost of the newly determined best next viewpoint with the current adjacent node as the parent node is calculated. This cost is the path length from the starting node of the tree to the newly determined best next viewpoint. The path with the lowest cost is selected as the safe and feasible path to the newly determined best next viewpoint.
[0030] Preferably, in step S7, before each iteration, the task manager is used to determine the map coverage. If the map coverage does not reach the preset standard, the task continues. If the map coverage reaches the preset standard, the task is completed and the underwater vehicle stops the active mapping task.
[0031] When the map coverage does not reach the preset standard, but the advanced planner has not calculated the next viewpoint, i.e. there is no next viewpoint, the iteration also terminates, the map is output, and the underwater vehicle's active mapping task is completed.
[0032] The beneficial effects of this application include, but are not limited to:
[0033] This invention provides an active mapping method for unknown underwater environments based on hovering underwater vehicles. It utilizes a hovering underwater vehicle coupled with sensors to acquire environmental data of unknown ocean areas, determines the optimal next-viewpoint (NGP) based on information gain, and then determines a safe and feasible path to the NFP using the costs of each neighboring node in the current tree as its parent node. This real-time updating of the raster map enables efficient, rapid, and accurate map construction and dynamic updates, thereby assisting the underwater vehicle in navigation to target areas. It has broad application prospects in underwater environment exploration and underwater structure inspection tasks based on underwater vehicles.
[0034] The present invention provides an active mapping method for unknown underwater environments based on hovering underwater vehicles. It can obtain acoustic occupancy maps and optical coverage images in a single underwater mission, collect high-quality environmental data with low operating costs and time, and achieve effective marine environmental exploration.
[0035] The present invention provides an active mapping method for underwater unknown environments based on hovering underwater vehicles. Different planning algorithms and utility functions are executed on planners at different levels, thereby enabling underwater vehicles to explore underwater target areas safely and more efficiently. Attached Figure Description
[0036] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:
[0037] Figure 1 A diagram showing the working relationship between the modules in the active mapping method for underwater unknown environments provided by this invention.
[0038] Figure 2 A schematic diagram illustrating the method for determining the optimal next viewpoint;
[0039] Figure 3 A schematic diagram illustrating a method for determining a safe and feasible path to the optimal next viewpoint. Detailed Implementation
[0040] To clearly illustrate the technical features of this solution, the invention will be described in detail below through specific implementation methods and in conjunction with the accompanying drawings.
[0041] It should be noted that many specific details are set forth in the following description to provide a thorough understanding of the invention; however, the invention may also be implemented in other ways different from those described herein. Therefore, the scope of protection of the invention is not limited to the specific embodiments disclosed below.
[0042] like Figure 1As shown in the figure, the active mapping method for unknown underwater environments based on hovering underwater vehicles provided by the present invention includes the following steps:
[0043] S1. Obtain the current position and attitude information of the underwater vehicle through the navigation module, obtain the exploration range information of the shore-based setup through the communication module, and obtain environmental data through the sensor module;
[0044] The environmental data includes acoustic image data collected by forward-looking sonar within its sensing range. Furthermore, hovering underwater vehicles can hover at relatively low altitudes above the seabed, maintaining a safe and observable distance from other marine structures, thus enabling data acquisition by optical sensors. Therefore, in this embodiment, the environmental data also includes optical image data collected by optical sensors, which are used for coverage detection of the detection area.
[0045] S2. Using a map generator, a 3D raster map is created starting from the initial position of the underwater vehicle. The size of the 3D raster map area corresponds to the exploration range set on the shore. Based on the collected environmental data, the cells of the 3D raster map are divided into occupied cells, empty cells, and unknown cells. Occupied cells are then marked as viewed after being examined using optical image data acquired by the optical camera. In other words, the map generator determines whether a cell is occupied based on the acoustic image data from the forward-looking sonar, and then re-examines the cell based on more precise optical image data from the optical camera to obtain a more refined judgment.
[0046] Specifically, the map generator used in this embodiment is Octomap, a 3D map creation tool based on octrees. It can display a complete 3D graphic that includes barrier-free areas and unmapped areas, and sensor data based on occupancy grids can be fused and updated in multiple measurements.
[0047] S3. Use the advanced planner to perform uniform sampling on the current 3D raster map to obtain multiple next viewpoints. Compare the information gain values of each next viewpoint to determine the optimal next viewpoint.
[0048] The specific method for determining the best next viewpoint is as follows:
[0049] S3.1. Obtain m next viewpoints by uniformly sampling around the current position i. Using a viewpoint p as a node, continue uniform sampling to obtain n child viewpoints. Viewpoints located in the occupying grid need to be removed from the next viewpoint and child viewpoints;
[0050] S3.2 Calculate the information gain values of the next viewpoint and the child viewpoint. The information gain value of the next viewpoint is the difference between the information entropy of the next viewpoint and the information entropy of the current position. The information gain value of the child viewpoint is the difference between the information entropy of the child viewpoint and the information entropy of the corresponding next viewpoint. Information entropy is the ratio of unknown information to known information obtained within the sensor module's sensing range at the current viewpoint position.
[0051] In this embodiment, known information refers to the total volume of grid cells marked as occupied and empty, and unknown information refers to the total volume of grid cells marked as unknown.
[0052] S3.3 Construct a utility function based on the information gain value of the next viewpoint and the information gain value of the child viewpoint. The next viewpoint that is most efficient is calculated using the utility function and then determined as the optimal next viewpoint.
[0053] Wherein, the information gain value of the next viewpoint p at the current position i is Let E(x) be the information entropy of the next viewpoint p. i ) represents the information entropy of the current position i;
[0054] The information gain value of the child viewpoint is The information entropy of the next viewpoint p, The information entropy of the child viewpoint q of the next viewpoint p.
[0055] Specifically, such as Figure 2 As shown, the method for determining the best next viewpoint will be introduced using a plane in a 3D grid as an example.
[0056] Figure 2 In the diagram, position X represents the vehicle's current position, and the fan-shaped shaded area centered at position X represents the known range of forward-looking sonar detection at that current position; the circular shaded area represents obstacles.
[0057] Points A to H are multiple next viewpoints sampled around the current position. Point H is within an occupied grid, so point H is discarded.
[0058] Taking point G as an example, the information gain of point G is the ratio of the unknown information to the known information at point G. Specifically, the unknown information at point G refers to the predicted sonar detection range of point G minus the known grid area within the predicted sonar detection range. That is,
[0059] Multiple child viewpoints were obtained by sampling around point G, denoted as a to h. Points e and h are within the occupied grid, so point H is removed. Point b is the viewpoint that has been reached. The fan-shaped shadow area centered on point G is the known range of forward-looking sonar detection at point G.
[0060] Calculate the information gain values at points a, c, d, f, and g; the sector-shaped shaded area centered at point c represents the known range of forward-looking sonar detection at point c;
[0061] By summing the information gain values at points G, a, c, d, f, and g, the utility function I at point G is obtained. G =I a +I c +I d +I f +I g .
[0062] S4. After obtaining the optimal viewpoint position published by the high-level planner using the low-level planner, calculate a safe and feasible path to the next optimal viewpoint, and publish the waypoints along this path at a certain control frequency. The low-level control module needs time to react. If the waypoints are published too quickly, the low-level control module will not have time to react to the previous waypoint before receiving the next one, resulting in a significant deviation in the vehicle's trajectory. Therefore, it is necessary to publish the waypoints along this safe and feasible path at a certain control frequency.
[0063] The method for calculating a safe and feasible path to the optimal next viewpoint is as follows:
[0064] Starting from the initial position of the aircraft, a tree is formed by sequentially identifying the best next viewpoints. The newly determined best next viewpoint is then incorporated into the tree. Using the newly determined best next viewpoint as the center, all nodes in the tree that are adjacent to the newly determined best next viewpoint within a preset radius are identified. It is then checked whether the path between the newly determined best next viewpoint and each adjacent node intersects with obstacles. If the path does not intersect with obstacles, the cost of the newly determined best next viewpoint with the current adjacent node as the parent node is calculated. This cost is the path length from the starting node of the tree to the newly determined best next viewpoint. The path with the lowest cost is selected as the safe and feasible path to the newly determined best next viewpoint.
[0065] Specifically, such as Figure 3 As shown, the square shaded area represents an obstacle, and the best next viewpoint is X. rand X rand Let X be a node in a tree, with three neighboring nodes A, B, and C within a dashed circle of a predetermined radius. rand They have a relatively short Euclidean distance.
[0066] First, calculate X when the nearest node A is the parent node. rand The cost: When the nearest node A is the parent node, X is achieved. rand The path is X int O-OC-CA-AX randThe path does not intersect with any obstacles and is therefore safe. Calculate the length of the path.
[0067] Then, calculate X when the nearest node B is the parent node. rand The cost: Achieving X when neighboring node B is the parent node rand The path intersects with the obstacle, making this path infeasible.
[0068] Continue calculating X when the nearest node C is the parent node. rand The cost: Achieving X when the nearest node C is the parent node rand The path is X int O-OC-CX rand Calculations showed that the cost of this path was higher than that of X when neighboring node A was the parent node. rand If the cost is small, then assign the neighboring node C to X. rand The parent node, the path is a safe and feasible path to the newly determined best next viewpoint.
[0069] S5. After obtaining the current position of the underwater vehicle and the desired trajectory points issued by the low-level planner using the guidance module, the desired heading is generated. After obtaining the current heading and desired heading issued by the guidance module, the low-level control module calculates and issues the rudder angle, so that the underwater vehicle travels along the calculated path.
[0070] S6. The underwater vehicle arrives at the new viewpoint according to the waypoints. The navigation module publishes the current position as the new viewpoint. The map generator at the new viewpoint updates the map based on the environmental data obtained by the sensor module.
[0071] S7. After the map is updated, repeat steps S3-S6, iterating sequentially until the map of the preset exploration area is completed.
[0072] Before each iteration, the task manager is used to determine the map coverage. If the map coverage does not reach the preset standard, the task continues. For example, if the preset standard for map coverage is 80% of the exploration area, the task will stop when the coverage of the built map exceeds 80%. When the map coverage reaches the preset standard, the task is completed and the underwater vehicle stops actively building maps.
[0073] When the map coverage does not reach the preset standard, but the advanced planner has not calculated the next viewpoint, i.e. there is no next viewpoint, the iteration also terminates, the map is output, and the underwater vehicle's active mapping task is completed.
[0074] In practical applications, the parameters required for the exploration mission are set and published through the user interface of the shore-based software system. The airborne software system includes an active mapping module, a data management module, and other modules including a navigation module, a guidance module, external sensors, and a low-level controller.
[0075] It should be noted that the above specific embodiments should not be construed as limiting the scope of protection of the present invention. For those skilled in the art, any alternative improvements or modifications made to the embodiments of the present invention shall fall within the scope of protection of the invention.
[0076] Any aspects of this invention not described in detail are well-known to those skilled in the art.
Claims
1. A method for active mapping of unknown underwater environments based on hovering underwater vehicles, characterized in that, Includes the following steps: S1. Obtain the current position and attitude information of the underwater vehicle through the navigation module, obtain the exploration range information of the shore-based setup through the communication module, and obtain environmental data through the sensor module; Among them, environmental data includes acoustic image data collected by forward-looking sonar within its sensing range; S2. The map generator creates a 3D raster map starting from the initial position of the underwater vehicle. The size of the 3D raster map area is the exploration range set on the shore. Based on the collected environmental data, the cells of the 3D raster map are divided into occupied cells, empty cells, and unknown cells. S3. Use the advanced planner to perform uniform sampling on the current 3D grid map to obtain multiple next viewpoints. Compare the information gain values of each next viewpoint to determine the optimal next viewpoint. S4. After obtaining the best viewpoint position published by the high-level planner using the low-level planner, calculate the safe and feasible path to the best next viewpoint, and publish the waypoints in the path at a certain control frequency. S5. After obtaining the current position of the underwater vehicle and the desired trajectory points issued by the low-level planner using the guidance module, the desired heading is generated. After obtaining the current heading and desired heading issued by the guidance module, the low-level control module calculates and issues the rudder angle, so that the underwater vehicle travels along the calculated path. S6. The underwater vehicle arrives at the new viewpoint according to the waypoints. The navigation module publishes the current position as the new viewpoint. The map generator at the new viewpoint updates the map based on the environmental data obtained by the sensor module. S7. After the map is updated, repeat steps S3-S6, iterating sequentially until the map of the preset exploration area is completed. Step S3 is as follows: S3.1, at the current location m next viewpoints are obtained by uniform sampling around the perimeter. From the following perspective To obtain n child viewpoints, continue uniform sampling of the node. In the next viewpoint and child viewpoints, viewpoints located in the occupied grid need to be removed; S3.2 Calculate the information gain values of the next viewpoint and the child viewpoint. The information gain value of the next viewpoint is the difference between the information entropy of the next viewpoint and the information entropy of the current position. The information gain value of the child viewpoint is the difference between the information entropy of the child viewpoint and the information entropy of the corresponding next viewpoint. Information entropy is the ratio of unknown information to known information obtained within the sensor module's sensing range at the current viewpoint position. S3.3 Construct a utility function based on the information gain value of the next viewpoint and the information gain value of the child viewpoint. The next viewpoint with the highest efficiency is calculated through the utility function and determined as the best next viewpoint. Current location The next perspective Information gain value , For the next perspective Information entropy Current location Information entropy; The information gain value of the child viewpoint is , For the next perspective Information entropy For the next perspective offspring perspective Information entropy.
2. The active mapping method for unknown underwater environments based on hovering underwater vehicles according to claim 1, characterized in that, In step S3.2, known information refers to the total volume of grid cells marked as occupied and empty, and unknown information refers to the total volume of grid cells marked as unknown.
3. The active mapping method for unknown underwater environments based on hovering underwater vehicles according to claim 1, characterized in that, In step S1, the environmental data also includes optical image data collected by the optical sensor, which is used to perform coverage detection on the detection area of the optical sensor.
4. The active mapping method for unknown underwater environments based on hovering underwater vehicles according to claim 1, characterized in that, In step S2, after viewing the occupied grids in the acoustic image in conjunction with the optical image, they are marked as viewed grids.
5. The active mapping method for unknown underwater environments based on hovering underwater vehicles according to claim 1, characterized in that, In step S4, the method for calculating a safe and feasible path to the optimal next viewpoint is as follows: Starting from the initial position of the aircraft, a tree is formed by sequentially identifying the best next viewpoints. The newly determined best next viewpoint is then incorporated into the tree. Using the newly determined best next viewpoint as the center, all nodes in the tree that are adjacent to the newly determined best next viewpoint within a preset radius are identified. It is then checked whether the path between the newly determined best next viewpoint and each adjacent node intersects with obstacles. If the path does not intersect with obstacles, the cost of the newly determined best next viewpoint with the current adjacent node as the parent node is calculated. This cost is the path length from the starting node of the tree to the newly determined best next viewpoint. The path with the lowest cost is selected as the safe and feasible path to the newly determined best next viewpoint.
6. The active mapping method for underwater unknown environments based on hovering underwater vehicles according to claim 1, characterized in that, In step S7, before each iteration, the task manager is used to determine the map coverage. If the map coverage does not reach the preset standard, the task continues. When the map coverage reaches the preset standard, the task is completed and the underwater vehicle stops the active mapping task. When the map coverage does not reach the preset standard, but the advanced planner has not calculated the next viewpoint, i.e. there is no next viewpoint, the iteration also terminates, the map is output, and the underwater vehicle's active mapping task is completed.