Multi-source information mapping and path planning method and system for autonomous unmanned underwater vehicle

Abstract

The application discloses a kind of multi-source information mapping and path planning method and system of autonomous unmanned underwater vehicle, method includes loading initial map containing prior information and giving initial confidence distribution to grid, configure multi-source perception system containing sonar and optical camera;Formulate the data confidence acquisition mode and conflict fusion rule in effective sensing range;Risk value is mapped to grid confidence, and the comprehensive cost evaluation function is defined in combination with distance and risk;Global reference path is generated by cyclically selecting optimal grid, evaluating adjacent grid and updating path;Perception data is continuously acquired and grid confidence is updated during movement, and re-planning is triggered when high-risk confidence or high-unknown confidence grid appears in path.The application realizes accurate modeling of underwater dynamic uncertain environment through complementary advantages of multi-source perception and confidence fusion, and improves the safety and navigation efficiency of autonomous unmanned underwater vehicle underwater operation.

Description

Technical Field

[0001] This invention relates to the field of mapping and path planning technology for autonomous unmanned underwater vehicles (UUVs), and more specifically, to a method and system for mapping and path planning using multi-source information for autonomous UUVs. Background Technology

[0002] Autonomous underwater vehicles (UUVs) face typical environmental characteristics of high dynamism, strong uncertainty, and incomplete information during underwater inspection operations. These include environmental state drift caused by ocean current disturbances, temporary obstacles caused by biological activity, map blurring due to sonar sensor noise, and unknown areas resulting from a lack of prior information. These uncertainties make it difficult for traditional path planning methods to simultaneously characterize the safety, danger, and state uncertainty of a grid, leading to potential risks in actual implementation. Existing underwater sensing systems typically rely on a single sensor, making it difficult to simultaneously handle long-range environmental perception and close-range accurate identification. The blind spots of a single sensor in scenarios with incomplete information further exacerbate map uncertainty.

[0003] Multi-source sensor fusion technology provides a new technical means for underwater environmental perception by integrating the complementary advantages of different types of sensors. Sonar sensors have a long detection range but lower accuracy, making them suitable for large-scale obstacle detection and terrain contour recognition; optical cameras have a short detection range but higher accuracy, making them suitable for target confirmation and fine feature recognition. However, a systematic solution is still lacking for how to effectively fuse the detection results of multiple sensors within a unified framework and apply them to map building and dynamic updates in path planning.

[0004] Therefore, it is necessary to propose a mapping and path planning method for autonomous unmanned underwater vehicles based on multi-source information fusion, which can achieve complementarity between sonar and optical perception through a confidence fusion mechanism, so as to improve the safety and navigation efficiency of underwater operations. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method and system for multi-source information mapping and path planning of autonomous unmanned underwater vehicles.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] A method for multi-source information mapping and path planning of an autonomous unmanned underwater vehicle includes the following steps:

[0008] Step S1: The autonomous unmanned underwater vehicle loads an initial map containing prior layout information and assigns an initial confidence level to each grid cell of the initial map. At the same time, it configures a multi-source perception system containing sonar sensors and optical cameras.

[0009] Step S2: Formulate the data reliability acquisition method and the fusion processing mechanism for conflicts within the effective sensing range of the multi-source sensing system;

[0010] Step S3: Map the latest confidence level of each grid after merging to a unified risk cost value, and combine it with the comprehensive cost evaluation function defined by the route and risk to evaluate the quality of the grid.

[0011] Step S4: Based on the comprehensive cost evaluation function, a global reference path that takes into account both distance and risk is generated by iteratively selecting the optimal grid, evaluating adjacent grids, and updating the path.

[0012] In step S5, the autonomous unmanned underwater vehicle continuously acquires perception data and updates the grid confidence as it moves along the global reference path. When a grid with high risk confidence or high unknown confidence appears on the path, replanning is triggered.

[0013] Furthermore, in step S1, the initial confidence assignment is used... , and express, This indicates the confidence level that supports the grid being known to be safe based on prior layout information. This indicates the confidence level supporting the view that the grid is a known danger. This indicates the confidence level that supports the grid being of unknown security; the sum of the three is 1.

[0014] Furthermore, in step S2, the data reliability is obtained in the following ways: for grids that are only within the effective detection range of the sonar sensor but outside the effective detection range of the optical camera, data reliability is generated only using the detection data of the sonar sensor; for grids that are simultaneously within the effective detection range of both the sonar sensor and the optical camera, data reliability is generated using the detection data of both.

[0015] Furthermore, in step S2, the fusion processing mechanism for conflicts in multi-source data reliability includes: for each grid within the sensing range, using the initial reliability assignment from step S1 as the initial reliability. First, set the initial reliability Reliability of sonar sensor data The fusion is performed according to the reliability fusion rule. If the grid is also within the effective detection range of the optical camera, the fusion result is then compared with the reliability of the optical camera data. The fusion is performed according to the confidence fusion rules to obtain the current confidence of the updated raster.

[0016] Furthermore, the reliability fusion rule is as follows: for two reliability sources in the same raster... and The reliability after fusion The calculation formula is:

[0017]

[0018]

[0019]

[0020] Fusion normalization constant The calculation method is as follows:

[0021]

[0022] In the formula, For the security reliability after fusion; The hazard reliability after fusion; The unknown reliability after fusion; As a source of confidence The security reliability; As a source of confidence Risk reliability; As a source of confidence Unknown reliability; As a source of confidence The security reliability; As a source of confidence Risk reliability; As a source of confidence Unknown reliability;

[0023] When two reliability sources agree on the state of the same grid, the reliability is enhanced after fusion; when their judgments conflict, the conflicting reliability is transferred to the unknown reliability.

[0024] Furthermore, in step S3, the risk cost assessment includes: adjusting the current grid's security confidence level. Risk reliability and unknown reliability The value of risk is mapped to risk, and its expression is:

[0025]

[0026] In the formula, For the current grid; , and These are the weighting coefficients.

[0027] Furthermore, in step S3, a comprehensive cost function is constructed based on the starting and ending grids of the autonomous unmanned underwater vehicle, and its expression is:

[0028]

[0029]

[0030]

[0031]

[0032]

[0033] In the formula, From the starting point to the current grid The cumulative distance cost of the optimal path is known; From the starting point to the current grid The cumulative risk cost of the optimal path is known; For the current grid The Euclidean distance to the destination; To start from the current grid The estimated average risk cost to the destination; , , and These are adjustable weighting coefficients; From the starting point to the current grid All raster sequences along the path; For the current grid All raster sequences that the line connecting to the endpoint passes through; This represents the number of grid cells the line passes through. For grid The risk-value; Indicates the Euclidean distance between grid cells; Represents a grid The parent node on the path; For the current grid coordinates The coordinates of the endpoint; For raster indexes.

[0034] Furthermore, step S4 includes the following steps:

[0035] Step S41: Create an open list of priority queue structures with an initial state of emptiness, and a closed list with an initial state of emptiness.

[0036] Step S42: Calculate the comprehensive cost evaluation function of the confidence level of all reachable adjacent graticles of the starting point and add them to the open list; initialize the comprehensive cost evaluation function of the remaining graticles to infinity.

[0037] Step S43: Remove the grid cell with the smallest confidence comprehensive cost evaluation function value from the open list as the current grid cell and move it into the closed list;

[0038] Step S44: Check each reachable neighboring cell of the current cell in turn; if it is in the closed list, ignore it.

[0039] Step S45: For adjacent rasters that are not ignored and are not in the open list, add them to the open list, set their parent node as the current raster, and calculate their comprehensive confidence cost evaluation function value.

[0040] Step S46: For adjacent grids that have not been ignored and are already in the open list, if the obtainable comprehensive confidence cost evaluation function value is small when starting from the starting point and passing through the current grid, then update its parent node to the current grid and update its comprehensive confidence cost evaluation function value.

[0041] Step S47: Repeat steps S43 to S46 until the endpoint enters the open list. Then, start from the endpoint and trace back along the parent node to the starting point to generate a global reference path, or until the open list is empty.

[0042] Furthermore, step S5 includes the following steps:

[0043] Step S51: The autonomous unmanned underwater vehicle moves along the global reference path, and the multi-source sensing system continuously acquires the surrounding underwater environment data and updates the credibility of each grid within the detection range in real time according to the credibility fusion rules.

[0044] Step S52: If the updated current path has a danger confidence level higher than a preset threshold... Or the unknown confidence level is higher than the preset threshold If the grid is a certain type, path replanning is triggered, and the threshold is [not specified]. and threshold Preset positive real numbers;

[0045] Step S53: Starting from the current position, proceed to step S4 to re-execute the global path search;

[0046] Step S54: If no high-risk or high-unknown-confidence grid appears on the current path, continue moving along the current path; repeat steps S51 to S54 until the destination is reached.

[0047] This invention also provides a multi-source information mapping and path planning system for autonomous unmanned underwater vehicles, the system comprising:

[0048] A multi-source sensing system, including sonar sensors and optical cameras, is used to acquire real-time data on the surrounding underwater environment;

[0049] The reliability fusion module is used to generate real-time data reliability based on the detection data from the sonar sensor and the optical camera, and to fuse the current reliability with the real-time data reliability from the multi-source sensors step by step through the reliability fusion rules, thereby updating the safety reliability, danger reliability and unknown reliability of each grid.

[0050] The comprehensive cost assessment module maps the current confidence level of each grid cell to a uniform risk cost value and combines the route and risk to define the comprehensive cost assessment function;

[0051] The global path search module is used to generate a global reference path that takes into account both distance and risk based on the comprehensive cost evaluation function.

[0052] The dynamic replanning module is used when the path becomes dangerous and the confidence level exceeds a preset threshold. Or the unknown confidence level is higher than the preset threshold When dealing with a grid, the global path search module is invoked again to replan the path, using the current position as the new starting point; threshold and threshold Preset positive real numbers;

[0053] The map storage module is used to store the initial map and the updated raster reliability data.

[0054] The beneficial effects of this invention are:

[0055] 1. This invention constructs a multi-source information map based on confidence fusion by loading an initial map containing prior layout information and assigning an initial confidence level to each grid cell according to the prior layout information. It utilizes safety confidence, danger confidence, and unknown confidence levels to quantify the safety, uncertainty, and danger of each grid cell, achieving accurate modeling of the underwater dynamic and uncertain environment. By acquiring underwater environmental data in real time through a multi-source perception system including sonar sensors and optical cameras, and employing confidence fusion rules to progressively fuse the initial confidence level with the real-time data confidence from the multi-source sensors to update the grid confidence level, the map information is continuously and progressively optimized as the unmanned underwater vehicle moves. By constructing a comprehensive cost evaluation function that includes path cost and risk cost, high danger confidence and high unknown confidence are transformed into high path cost, achieving an effective balance between path safety and compactness. By introducing a continuous perception-driven map update and path replanning mechanism, replanning is triggered when a high danger confidence or high unknown confidence grid cell appears in the current path, enabling the unmanned underwater vehicle to adaptively adjust its path under environmental changes.

[0056] 2. This invention upgrades the traditional "shortest path" search problem into a search problem of finding the "most reliable path" in a field of uncertainty. By leveraging the complementary advantages of sonar sensors and optical cameras, and by using confidence fusion rules to properly handle conflicting evidence, it significantly improves the operational safety and navigation efficiency of autonomous unmanned underwater vehicles in dynamic and uncertain underwater environments, providing reliable technical support for tasks such as underwater resource exploration and pipeline inspection. Attached Figure Description

[0057] Figure 1 This is a flowchart of a multi-source information mapping and path planning method for an autonomous unmanned underwater vehicle in this embodiment.

[0058] Figure 2 This is a framework diagram of the multi-source information mapping and path planning system for the autonomous unmanned underwater vehicle in this embodiment.

[0059] Figure 3 This is the initial underwater confidence map based on confidence assignment in this embodiment;

[0060] Figure 4 This is the initial comprehensive risk cost map generated based on confidence-weighted mapping in this embodiment;

[0061] Figure 5 This is the first global reference path graph generated based on the comprehensive cost evaluation function search in this embodiment;

[0062] Figure 6 This is the underwater reliability map updated based on multi-source information from an autonomous unmanned underwater vehicle in this embodiment;

[0063] Figure 7 This is the second global reference path map generated after the autonomous unmanned underwater vehicle triggers replanning in this embodiment.

[0064] Figure labels: 1. Multi-source sensing system; 2. Reliability fusion module; 3. Comprehensive cost assessment module; 4. Global path search module; 5. Dynamic replanning module; 6. Map storage module. Detailed Implementation

[0065] 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.

[0066] Example: A multi-source information mapping and path planning method for an autonomous unmanned underwater vehicle (UUV). This method loads an initial map composed of several unit grids, each grid containing prior layout information. Based on the prior layout information, each grid is assigned an initial safety confidence level, a hazard confidence level, and an unknown confidence level, constructing an initial confidence map. Then, through a confidence-weighted mapping, the confidence level of each grid is transformed into a unified risk cost value. A comprehensive cost evaluation function is then constructed by combining distance cost and risk cost to evaluate the quality of grids during path search. A heuristic search algorithm iteratively executes optimal node selection, adjacent node evaluation, and path update to generate the first global reference path from the starting point to the destination. The UUV then moves along the path... During the process, the multi-source sensing system continuously acquires data on the surrounding underwater environment. This system includes a sonar sensor with a long detection range but low accuracy and an optical camera with a short detection range but high accuracy. The sonar sensor probes the surrounding area, updating unknown areas within its detection range to safe zones with a certain level of confidence. The optical camera performs high-precision confirmation of nearby grid lines, verifying the sonar detection results and correcting the confidence level, thus achieving progressive map updates and accuracy improvements. In this embodiment, the maximum effective detection range of the sonar sensor is set to 3 grids, and the effective detection range of the optical camera is set to 1 grid. The effective detection range can be adjusted adaptively according to the actual scenario. When encountering a new obstacle, the path is replanned from the current position as a new starting point. This method, through the complementary multi-source sensing and confidence fusion of sonar and optical cameras, upgrades the traditional shortest path search problem to finding the most reliable path in a confidence uncertainty field, effectively balancing path length and passage risk, and achieving adaptive path adjustment under environmental changes.

[0067] like Figure 1 As shown, the method specifically includes the following steps:

[0068] Step S1, Initial map loading and multi-source perception system configuration: The autonomous unmanned underwater vehicle loads an initial map consisting of several unit grids. Each grid contains prior layout information. Based on the prior layout information and allocation rules, an initial confidence assignment is given to each grid. The vehicle is equipped with a multi-source perception system to acquire surrounding underwater environmental data in real time. The multi-source perception system includes a sonar sensor with a long detection range but low accuracy and an optical camera with a short detection range but high accuracy.

[0069] This step aims to load an initial map containing prior layout information and assign initial confidence to each grid cell, while configuring a multi-source sensing system to support the real-time acquisition of subsequent environmental data.

[0070] Specifically, step S1 includes the following steps:

[0071] Step S11, Initial Map Loading: Load the initial map of the underwater environment. The initial map consists of several cell grids. Each cell contains prior layout information, which includes one or more of the following: aquaculture cage layout, fixed reef location, and historical operation channel. An initial confidence assignment is given to each cell according to the prior layout information and assignment rules.

[0072] Among them, the initial confidence allocation is used , and express, This indicates the confidence level that supports the grid being "known safe" given prior layout information. This indicates the confidence level supporting the view that the grid is "known dangerous". This indicates a confidence level supporting the raster's "security unknown". , and The sum of the three is 1.

[0073] The initial reliability assignment rule is: if a grid cell is located in a historically verified safe passage area, then assign it a reliability score. =0.8、 =0.1、 =0.1; if the grid is located in a known obstacle area, then assign... =0.8、 =0.1、 =0.1; if the grid is located in an unknown area where prior layout information is missing, then assign 0.1. =1.0、 =0、 =0.

[0074] Step S12, Multi-source perception system configuration: The autonomous unmanned underwater vehicle is equipped with a multi-source perception system, which includes a sonar sensor and an optical camera. The sonar sensor has a long detection range but low accuracy, and is suitable for long-distance obstacle detection and terrain contour recognition. The optical camera has a short detection range but high accuracy, and is suitable for close-range target confirmation and fine feature recognition.

[0075] Step S2, Multi-source sensor data reliability fusion rule formulation: Formulate data reliability fusion and update rules for the autonomous unmanned underwater vehicle's multi-source perception system within its effective perception range. The rules include the data reliability acquisition methods of different sensors within their respective effective detection areas, as well as the fusion processing mechanism when multi-source data reliability conflicts.

[0076] This step aims to establish unified rules for data reliability acquisition and fusion for multi-source sensing systems, enabling the fusion of detection data from sonar sensors and optical cameras within a unified framework.

[0077] Specifically, step S2 includes:

[0078] Effective sensing range definition: The effective sensing range is a circular area centered on the current position of the autonomous unmanned underwater vehicle and with the maximum effective detection range of the sonar sensor as the radius; for grids that are only within the effective detection range of the sonar sensor but outside the effective detection range of the optical camera, data reliability is generated only based on the detection data of the sonar sensor; for grids that are simultaneously within the effective detection range of both the sonar sensor and the optical camera, data reliability is generated separately based on the detection data of both.

[0079] Data reliability acquisition method: Sonar sensors and optical cameras collect real-time underwater environmental data within the sensing range, and generate real-time sonar data reliability based on the confidence level of sensor echo signal intensity or image features, respectively. and the reliability of optical camera data Real-time sonar data reliability and real-time optical camera data reliability All three include safety reliability, hazard reliability, and unknown reliability, and the sum of the three is 1.

[0080] The data reliability generation method for sonar sensors is as follows: when the echo intensity is lower than a set threshold, the sonar data security reliability is... Assigning high confidence values ​​to sonar data hazard confidence A low confidence value is assigned; when the echo intensity exceeds a set threshold, the sonar data's hazard confidence level is adjusted. Assigning high confidence values ​​to sonar data security confidence Assign low confidence values; when the echo signal is unclear or cannot be determined, sonar data Assign a high confidence value.

[0081] The data confidence level for optical cameras is generated as follows: the corresponding security confidence level for the optical camera is generated based on the classification confidence level output by the target recognition algorithm. Risk reliability and unknown reliability .

[0082] The fusion processing mechanism when multi-source data reliability conflicts occur: For each grid cell within the sensing range, the initial reliability assignment in step S1 is used as the initial reliability. The real-time reliability is determined by the real-time data reliability of each sensor falling within the grid area. The initial reliability is then... Reliability of sonar sensor data The reliability is obtained by fusing according to the reliability fusion rules. If the grid is also within the effective detection range of the optical camera, then the confidence level is... Reliability of data from optical cameras The fusion is performed according to the reliability fusion rules, and the new fused reliability is used as the current reliability of the updated raster.

[0083] The reliability fusion rule is as follows: for two reliability sources in the same raster... and The reliability after fusion The calculation formula is:

[0084]

[0085]

[0086]

[0087] Among them, the fusion normalization constant The calculation method is as follows:

[0088]

[0089] In the formula, For the security reliability after fusion; The hazard reliability after fusion; For the unknown reliability after fusion; As a source of confidence The security reliability; As a source of confidence The risk reliability; As a source of confidence Unknown reliability; As a source of confidence The security reliability; As a source of confidence The risk reliability; As a source of confidence Unknown reliability;

[0090] Based on the above fusion rules, when two confidence sources agree on the state of the same grid, the corresponding confidence is enhanced after fusion; when the two sources disagree, the conflicting confidence is transferred to the unknown confidence, making the grid state tend to be "unknown in terms of security".

[0091] Step S3, Comprehensive Cost Assessment: Map the latest reliability of each grid after fusion to a unified risk cost value, so that high-risk reliability and high unknown reliability both generate higher path costs after weighted mapping, and combine the route and risk definitions to use a comprehensive cost assessment function to evaluate the quality of the grid.

[0092] This step aims to transform the confidence of each grid into a unified risk value and construct a comprehensive cost evaluation function to provide an evaluation basis for subsequent path search.

[0093] Specifically, step S3 includes:

[0094] Risk cost assessment: assessing the security confidence of each grid cell. Risk reliability and unknown reliability The mapping is used to represent the risk cost suitable for path search, transforming high-risk confidence and high-unknown confidence into high-risk costs. Its expression is:

[0095]

[0096] In the formula, For the current grid; , and Each is the current grid Safety reliability, hazard reliability, and unknown reliability; , and These are the weighting coefficients.

[0097] As a preferred option, the weighting coefficient , and The value of satisfies > > This reflects the decision-making preference of avoiding known risks over being wary of uncertainty, and being wary of uncertainty over preferring safe areas.

[0098] Construction of the comprehensive cost evaluation function: The starting and ending grids of the autonomous unmanned underwater vehicle are defined, and the comprehensive cost function is constructed, with the following expression:

[0099]

[0100]

[0101]

[0102]

[0103]

[0104] In the formula, From the starting point to the current grid The cumulative distance cost of the optimal path is known; From the starting point to the current grid The cumulative risk cost of the optimal path is known; For the current grid The Euclidean distance to the destination; To start from the current grid The estimated average risk cost to the destination; , , and An adjustable weighting coefficient that is greater than or equal to zero; From the starting point to the current grid All raster sequences along the path; For the current grid All raster sequences that the line connecting to the endpoint passes through; This represents the number of grid cells the line passes through. For grid The risk-value; Indicates the Euclidean distance between grid cells; Represents a grid The parent node on the path; For the current grid coordinates The coordinates of the endpoint; For raster indexes.

[0105] Step S4, Global Reference Path Search: Based on the comprehensive cost evaluation function, a global reference path that balances distance and risk is generated by iteratively executing the operations of selecting the optimal grid, evaluating adjacent grids, and updating the path.

[0106] This step aims to use a heuristic search algorithm to generate a global reference path from the starting point to the destination based on a comprehensive cost evaluation function.

[0107] Specifically, step S4 includes the following steps:

[0108] Step S41, List creation and initialization: Create an open list with an initial empty priority queue structure and a closed list with an initial empty state. The open list is maintained in ascending order according to the value of the reliability comprehensive cost evaluation function of the raster. The closed list is used to store fully expanded rasters for which all feasible adjacent rasters have been evaluated at least once.

[0109] Step S42, Start point initialization: Calculate the comprehensive confidence cost evaluation function of all reachable neighboring grids of the start point and add them to the open list. At the same time, initialize the comprehensive confidence cost evaluation function of all grids except the start point and its reachable neighboring grids to infinity.

[0110] Step S43, Select the current node: Take the grid with the smallest confidence comprehensive cost evaluation function value from the open list as the current grid and move it into the closed list.

[0111] Step S44, check adjacent nodes: check each reachable adjacent grid cell of the current grid cell in turn, and ignore it if it is in the closed list.

[0112] Step S45, Generate new path: For adjacent rasters that are not ignored and are not in the open list, add them to the open list, set their parent node as the current raster, and calculate their confidence comprehensive cost evaluation function value.

[0113] Step S46, optimize existing paths: For adjacent rasters that are not ignored and are already in the open list, if a smaller confidence comprehensive cost evaluation function value can be obtained by starting from the starting point and passing through the current raster, then update its parent node to the current raster and update its confidence comprehensive cost evaluation function value.

[0114] Step S47, Global Reference Path Generation: Repeat steps S43 to S46 until the endpoint enters the open list, then start from the endpoint and trace back along the parent node to the starting point to generate a global reference path, or until the open list is empty.

[0115] Step S5, Dynamic Map Update and Path Replanning: As the autonomous unmanned underwater vehicle moves along the global reference path, the multi-source perception system continuously acquires surrounding underwater environment data and updates the confidence of relevant grids according to the confidence fusion rules in step S2. When a grid with high risk confidence or high unknown confidence appears on the current path, step S4 is re-executed to replan the global path, with the current position as the new starting point.

[0116] This step aims to enable autonomous unmanned underwater vehicles to continuously perceive environmental changes during movement, dynamically update map reliability, and replan their routes when necessary.

[0117] Specifically, step S5 includes the following steps:

[0118] Step S51, Continuous perception and confidence update: The autonomous unmanned underwater vehicle moves along the global reference path, and the multi-source perception system continuously acquires the surrounding underwater environment data, and updates the safety confidence, danger confidence and unknown confidence of each grid within the detection range in real time according to the confidence fusion rules in step S2.

[0119] Step S52, Replanning trigger condition judgment: If there is a danger confidence on the updated current path. Higher than the preset threshold or unknown reliability Higher than the preset threshold The grid, using the current position of the autonomous underwater vehicle as a new starting point, triggers path replanning, with a threshold... and threshold This is a preset positive real number, set by the system according to the task's security level.

[0120] Step S53, Global Path Replanning: Starting from the current position, proceed to step S4 to re-execute the global path search;

[0121] Step S54, iterative execution: If no high-risk confidence or high-unknown confidence grid appears on the current path, continue moving along the current path; repeat steps S51 to S54 until the autonomous unmanned underwater vehicle reaches the destination.

[0122] This embodiment also provides a multi-source information mapping and path planning system for autonomous unmanned underwater vehicles, such as... Figure 2 As shown, the system includes a multi-source perception system 1, a reliability fusion module 2, a comprehensive cost evaluation module 3, a global path search module 4, a dynamic replanning module 5, and a map storage module 6.

[0123] The multi-source sensing system 1 includes a sonar sensor with a long detection range but low accuracy and an optical camera with a short detection range but high accuracy, used to acquire underwater environmental data around the autonomous unmanned underwater vehicle in real time.

[0124] The reliability fusion module 2 is responsible for generating real-time data reliability based on the detection data from the sonar sensor and the optical camera, respectively. It then merges the current reliability in the map storage module 6 with the real-time data reliability from the multi-source sensors step by step through the reliability fusion rules, updates the safety reliability, danger reliability, and unknown reliability of each grid, and stores the updated reliability in the map storage module 6.

[0125] The comprehensive cost assessment module 3 is responsible for obtaining the current confidence level of each grid from the map storage module 6, converting the current confidence level of each grid into a unified risk cost value through weighted mapping, converting high risk confidence and high unknown confidence into high path cost, and defining a comprehensive cost assessment function based on risk cost value, starting point and ending point to evaluate the quality of grids during the path search process.

[0126] The global path search module 4 is responsible for constructing and initializing the open and closed lists. It performs operations such as selecting the grid with the smallest comprehensive cost evaluation function value as the current grid, evaluating its neighboring grids, and updating the path in a loop. Based on the current confidence in the map storage module 6, it generates a global reference path that takes into account both route and risk.

[0127] The dynamic replanning module 5 is responsible for continuously acquiring surrounding underwater environment data through the multi-source sensing system 1 during the movement of the autonomous unmanned underwater vehicle along the global reference path, triggering the confidence fusion module 2 to update the confidence of relevant grids, and responding to situations where the confidence level exceeds a preset threshold when a danger occurs on the current path. Or the unknown confidence level is higher than the preset threshold When retrieving a grid, the global path search module 4 is invoked again to replan the path, using the current position as the new starting point, with a threshold. and threshold It is a pre-defined positive real number.

[0128] Map storage module 6 is responsible for storing the initial map data and the raster reliability data updated by reliability fusion module 2.

[0129] The method proposed in this embodiment will be described in detail below, using a 12×12 grid underwater operation area as a simulation scenario.

[0130] First, according to step S1, load the initial map and assign initial confidence levels, such as... Figure 3 As shown, this embodiment loads an initial map consisting of 12×12 unit grids, each grid being 1 meter × 1 meter in size. Based on the prior layout information of the underwater environment, each grid is assigned an initial safety confidence, a hazard confidence, and an unknown confidence, the sum of which is 1.

[0131] In this embodiment, the outermost ring of the map is set as boundary obstacles, and assigned... =0.8、 =0.2 indicates a highly trustworthy danger zone. In historically verified safe passage zones, an assignment is made... =0.8、 =0.2. In the unknown region where prior information is missing, assign... =1.0. Meanwhile, this embodiment sets up five static obstacle areas to simulate underwater cages and reefs, all of which are assigned... =0.8、 =0.2. Figure 3 The initial confidence map is displayed using an RGB true color mapping method, where green corresponds to safe confidence (the greener the color, the safer the confidence), blue corresponds to unknown confidence (the bluer the color, the more uncertain the confidence), and red corresponds to dangerous confidence (the redder the color, the more dangerous the confidence).

[0132] Next, an initial comprehensive risk cost map is generated according to step S2, such as... Figure 4 As shown, in this embodiment, the confidence level of each grid cell is transformed into a uniform risk cost value through a weighted mapping, where the weight coefficient is set to... =1.0、 =0.5、 =0.2, reflecting the highest penalty for known dangerous obstacles, while imposing a medium penalty for unknown areas. Calculations show that the risk cost for known safe areas is 0.10, for boundaries and areas with known obstacles it is above 1.00, and for unknown areas it is 0.50. Figure 4 A color-coded risk cost map is used to display the overall risk, with colors ranging from blue to red indicating increasing risk. In this embodiment, the starting point is set to (2,2) and the ending point to (11,11). The weighting coefficients of the overall cost evaluation function are set to... =1.0、 =2.0、 =1.0、 =2.0, reflecting the high penalties for high-risk areas in order to prioritize safety.

[0133] Then, according to step S4, the first global reference path search and generation is performed. First, an empty open list and closed list are created, the starting point is added, and its comprehensive evaluation value is initialized. Then, the operations of selecting the node with the smallest comprehensive cost evaluation function value as the current node, checking its adjacent nodes, generating a new path, or optimizing an existing path are executed iteratively. When the endpoint is added to the open list, a global reference path is generated by tracing backwards from the endpoint along its parent node. For example... Figure 5 As shown, this embodiment successfully generated the first global reference path from the starting point (2,2) to the ending point (11,11). The path has 12 nodes, a total distance cost of 13.8995, and a total risk cost of 1.1000. The path effectively avoids all red high-risk areas, prioritizes green safe areas, and passes through the edge of blue unknown areas when complete avoidance is not possible, verifying the effectiveness of the proposed method in balancing distance and risk.

[0134] Subsequently, according to step S5, the autonomous unmanned underwater vehicle moves along the first global reference path, the multi-source perception system continuously acquires surrounding underwater environmental data, the sonar sensor detects the surrounding 3 grid range, updates the unknown area within the detection range to a safe area with a certain confidence level, and the optical camera performs high-precision confirmation of the surrounding grid, verifies the sonar detection results and corrects the confidence level. Figure 6 To explore the updated underwater confidence map, compared to the initial confidence map, a large number of unknown areas along the route and near the endpoint have been updated to known safe areas, and the map information is gradually enriched.

[0135] When the autonomous underwater vehicle moves to waypoint (6,5), it detects a new obstacle ahead at (7,6), triggering path replanning. For example... Figure 7 As shown, this embodiment re-executes the path search with the current position as the new starting point and successfully generates a second global reference path that bypasses the new obstacle. This verifies the effectiveness of the method proposed in this embodiment in achieving adaptive path adjustment through multi-source perception and confidence fusion in a dynamic and uncertain environment.

[0136] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A method for multi-source information mapping and path planning of an autonomous unmanned underwater vehicle, characterized in that, Includes the following steps: Step S1: The autonomous underwater vehicle loads an initial map containing prior layout information and assigns initial confidence assignments to each grid cell of the initial map. The initial confidence assignments are... , and express, This indicates the confidence level that supports the grid being known to be safe based on prior layout information. This indicates the confidence level supporting the view that the grid is a known danger. This indicates a confidence level that supports the grid as having unknown security, with the sum of the three factors being 1; it also includes a multi-source sensing system with sonar sensors and optical cameras. Step S2: Formulate the data reliability acquisition method and the fusion processing mechanism for conflicts within the effective sensing range of the multi-source sensing system; Step S3: Map the latest confidence scores of each grid after merging to a unified risk value. The mapping includes: mapping the current grid's safety confidence score... Risk reliability and unknown reliability The value of risk is mapped to risk, and its expression is: In the formula, For the current grid; , and These are the weighting coefficients; A comprehensive cost evaluation function, combining distance and risk definitions, is used to assess the quality of grids. This function is constructed based on the starting and ending grids of the autonomous unmanned underwater vehicle, and its expression is: In the formula, From the starting point to the current grid The cumulative distance cost of the optimal path is known; From the starting point to the current grid The cumulative risk cost of the optimal path is known; For the current grid The Euclidean distance to the destination; To start from the current grid The estimated average risk cost to the destination; , , and These are adjustable weighting coefficients; From the starting point to the current grid All raster sequences along the path; For the current grid All raster sequences that the line connecting to the endpoint passes through; This represents the number of grid cells the line passes through. For grid The risk-value; Indicates the Euclidean distance between grid cells; Represents a grid The parent node on the path; For the current grid coordinates The coordinates of the endpoint; For raster indexes; Step S4: Based on the comprehensive cost evaluation function, a global reference path that takes into account both distance and risk is generated by iteratively selecting the optimal grid, evaluating adjacent grids, and updating the path. In step S5, the autonomous unmanned underwater vehicle continuously acquires perception data and updates the grid confidence as it moves along the global reference path. When a grid with high risk confidence or high unknown confidence appears on the path, replanning is triggered.

2. The multi-source information mapping and path planning method for an autonomous unmanned underwater vehicle according to claim 1, characterized in that, In step S2, the data confidence level is obtained in the following ways: for grids that are only within the effective detection range of the sonar sensor but outside the effective detection range of the optical camera, data confidence level is generated only using the detection data of the sonar sensor; for grids that are simultaneously within the effective detection range of both the sonar sensor and the optical camera, data confidence level is generated using the detection data of both.

3. The multi-source information mapping and path planning method for an autonomous unmanned underwater vehicle according to claim 2, characterized in that, In step S2, the fusion processing mechanism for conflicts in multi-source data reliability includes: for each grid cell within the sensing range, the initial reliability assignment in step S1 is used as the initial reliability. First, set the initial reliability Reliability of sonar sensor data The fusion is performed according to the reliability fusion rule. If the grid is also within the effective detection range of the optical camera, the fusion result is then compared with the reliability of the optical camera data. The fusion is performed according to the confidence fusion rules to obtain the current confidence of the updated raster.

4. The multi-source information mapping and path planning method for an autonomous unmanned underwater vehicle according to claim 3, characterized in that, The reliability fusion rule is as follows: for two reliability sources in the same raster... and The reliability after fusion The calculation formula is: Fusion normalization constant The calculation method is as follows: In the formula, For the security reliability after fusion; The hazard reliability after fusion; For the unknown reliability after fusion; As a source of confidence The security reliability; As a source of confidence The risk reliability; As a source of confidence Unknown reliability; As a source of confidence The security reliability; As a source of confidence Risk reliability; As a source of confidence Unknown reliability; When two reliability sources agree on the state of the same grid, the reliability is enhanced after fusion; when their judgments conflict, the conflicting reliability is transferred to the unknown reliability.

5. The multi-source information mapping and path planning method for an autonomous unmanned underwater vehicle according to claim 1, characterized in that, Step S4 includes the following steps: Step S41: Create an open list of priority queue structures with an initial state of emptiness, and a closed list with an initial state of emptiness. Step S42: Calculate the comprehensive cost evaluation function of the confidence level of all reachable adjacent graticles of the starting point and add them to the open list; initialize the comprehensive cost evaluation function of the remaining graticles to infinity. Step S43: Remove the grid cell with the smallest confidence comprehensive cost evaluation function value from the open list as the current grid cell and move it into the closed list; Step S44: Check each reachable neighboring cell of the current cell in turn; if it is in the closed list, ignore it. Step S45: For adjacent rasters that are not ignored and are not in the open list, add them to the open list, set their parent node as the current raster, and calculate their comprehensive confidence cost evaluation function value. Step S46: For adjacent grids that have not been ignored and are already in the open list, if the obtainable comprehensive confidence cost evaluation function value is small when starting from the starting point and passing through the current grid, then update its parent node to the current grid and update its comprehensive confidence cost evaluation function value. Step S47: Repeat steps S43 to S46 until the endpoint enters the open list. Then, start from the endpoint and trace back along the parent node to the starting point to generate a global reference path, or until the open list is empty.

6. The multi-source information mapping and path planning method for an autonomous unmanned underwater vehicle according to claim 1, characterized in that, Step S5 includes the following steps: Step S51: The autonomous unmanned underwater vehicle moves along the global reference path, and the multi-source sensing system continuously acquires the surrounding underwater environment data and updates the credibility of each grid within the detection range in real time according to the credibility fusion rules. Step S52: If the updated current path has a danger confidence level higher than a preset threshold... Or the unknown confidence level is higher than the preset threshold For grids containing certain elements, path replanning is triggered, with a threshold value of [value missing]. and threshold Preset positive real numbers; Step S53: Starting from the current position, proceed to step S4 to re-execute the global path search; Step S54: If no high-risk or high-unknown-confidence grid appears on the current path, continue moving along the current path; repeat steps S51 to S54 until the destination is reached.

7. A multi-source information mapping and path planning system for an autonomous unmanned underwater vehicle to implement the method of claim 1, characterized in that, The system includes: A multi-source sensing system, including sonar sensors and optical cameras, is used to acquire real-time data on the surrounding underwater environment; The reliability fusion module is used to generate real-time data reliability based on the detection data from the sonar sensor and the optical camera, and to fuse the current reliability with the real-time data reliability from the multi-source sensors step by step through the reliability fusion rules, thereby updating the safety reliability, danger reliability and unknown reliability of each grid. The comprehensive cost assessment module maps the current confidence level of each grid cell to a uniform risk cost value and combines the route and risk to define the comprehensive cost assessment function; The global path search module is used to generate a global reference path that takes into account both distance and risk based on the comprehensive cost evaluation function. The dynamic replanning module is used when the path becomes dangerous and the confidence level exceeds a preset threshold. Or the unknown confidence level is higher than the preset threshold When dealing with a grid, the global path search module is invoked again to replan the path, using the current position as the new starting point; threshold and threshold Preset positive real numbers; The map storage module is used to store the initial map and the updated raster reliability data.

Images