Intelligent detection multifunctional robot autonomous navigation method and system

Abstract

The application relates to the technical field of robot autonomous navigation, in particular to a smart detection multifunctional robot autonomous navigation method and system, which specifically comprises the following steps: acquiring a grid map of a robot movement scene, determining the static cost of each grid in the grid map by analyzing the distribution of static obstacles in each grid and its adjacent grids, determining the obstacle degree of each grid in combination with the similarity between the multi-modal information of all dynamic obstacles passing through the grid and the distance between the dynamic obstacles and the robot, determining the node cost value between the grids based on the distribution of obstacles in the feature region between two adjacent static obstacles in combination with the obstacle degree, and performing path planning on the robot in combination with a path planning algorithm. The movement cost evaluation between nodes is optimized, the control difficulty and energy consumption of actual movement of the robot are more accurately reflected, and the practicability and reliability of navigation are improved.

Description

Technical Field

[0001] This application relates to the field of robot autonomous navigation technology, specifically to an autonomous navigation method and system for a smart detection multifunctional robot. Background Technology

[0002] In the field of autonomous robot navigation, path planning is one of the core technologies for achieving intelligent mobility. Traditional A... The algorithm is widely used for optimal path search in static environments, estimating the minimum-cost path from the starting point to the target point through a heuristic function. However, in practical applications such as intelligent testing laboratories for engineering materials, the environment often contains both static and dynamic obstacles (such as moving equipment or personnel). Traditional algorithms... The algorithm relies solely on geometric distance or fixed time cost as the node's value, making it unsuitable for complex and dynamic environments. This leads to frequent path failures when encountering dynamic obstacles, requiring recalculation and causing robot response delays, movement stutters, or even task interruptions. Summary of the Invention

[0003] To address the aforementioned technical problems, the purpose of this application is to provide an autonomous navigation method and system for a smart detection multifunctional robot, the specific technical solution of which is as follows:

[0004] In a first aspect, embodiments of this application provide an autonomous navigation method for a smart detection multifunctional robot, the method comprising the following steps:

[0005] Acquire a static environmental grid map of the working area of ​​the intelligent inspection robot and multimodal information of each dynamic obstacle detected in history;

[0006] Based on the distribution of the location and number of static obstacles in each grid, as well as the number of edge corners of the static obstacles, the static cost of each grid is constructed; based on the distribution characteristics of the static cost of each grid's neighboring grids, the passability of each grid is constructed to optimize the static cost of each grid.

[0007] Based on the similarity between the multimodal information of all dynamic obstacles passing through each grid, and the distance between each dynamic obstacle and the robot, combined with the optimized static cost, the obstacle degree of each grid is determined.

[0008] Each grid is divided into regions. Based on the connectivity between each region and its neighboring regions, the characteristic regions between each grid and each of its neighboring grids are determined. Based on the distribution of static obstacles within the characteristic regions and the static cost of the characteristic regions, the movement cost between each grid and each of its neighboring grids is determined.

[0009] Based on the obstacle degree and the movement cost, the node cost between each grid and each of its neighboring grids is determined, and the robot performs path planning in conjunction with the path planning algorithm.

[0010] In one embodiment, the process of obtaining the static cost of each grid is as follows:

[0011] Obtain the number of static obstacles in each grid cell and the area percentage of each static obstacle in the grid cell; obtain the distance between each static obstacle in each grid cell and its nearest other static obstacle, and record it as the first distance;

[0012] The static cost of each grid is positively correlated with the number of static obstacles, the area ratio of all static obstacles, and the number of edge corner points, and negatively correlated with the first distance of all static obstacles.

[0013] In one embodiment, the process of obtaining the passability of each grid is as follows:

[0014] Obtain the minimum difference between the static cost normalization values ​​of any neighboring grid and all other neighboring grids for each grid; the accessibility of each grid is negatively correlated with the static cost normalization values ​​of all its neighboring grids and the minimum value.

[0015] In one embodiment, the expression for optimizing the static cost of each grid is:

[0016] In the formula, U represents the optimized static cost of the current grid; U represents the unoptimized static cost of the current grid; and Y represents the passability of the current grid.

[0017] In one embodiment, the process of obtaining the obstacle degree of each grid is as follows:

[0018] The multimodal vectors of each dynamic obstacle are determined by the multimodal information of each dynamic obstacle; among all dynamic obstacles passing through the current grid, the mean similarity between the multimodal vectors of any dynamic obstacle and all other dynamic obstacles is calculated, and the minimum value of the instantaneous position distance between each dynamic obstacle and the robot is obtained, which is denoted as the first closest distance.

[0019] Obtain the number of times the robot detected a dynamic obstacle passing through the current grid in all its historical detection tasks;

[0020] The dynamic influence of the current grid is determined based on the number of times, the mean, and the first nearest distance. The dynamic influence of the current grid is positively correlated with the number of times and the mean, and negatively correlated with the first nearest distance.

[0021] The product of the dynamic influence degree of each grid and the optimized static cost degree is used as the obstacle degree of each grid.

[0022] In one embodiment, the expression for the dynamic influence degree is:

[0023] In the formula, F is the dynamic influence degree of the current grid; N is the number of times; and M is the number of dynamic obstacles passing through the current grid. The mean value is the value of the m-th dynamic obstacle passing through the current grid. The first closest distance to the m-th dynamic obstacle in the current grid.

[0024] In one embodiment, the process of obtaining the feature region is as follows:

[0025] If any region of grid Q Any region of its neighboring grid W If connected, the area will be... With the region The region formed is used as the feature region between grid Q and its neighboring grid W.

[0026] In one embodiment, the process of obtaining the movement cost value is as follows:

[0027] For each grid cell and its neighboring grid cells, the distance between each static obstacle in the feature region and its nearest other static obstacle is recorded as the second distance of each static obstacle. The information entropy of the second distance of all static obstacles in the feature region is obtained. The feature region is divided into regions, and the proportion of the region occupied by obstacles in the feature region is obtained. The fusion value of the static cost of the feature region, the information entropy, and the proportion is used as the movement cost value between each grid cell and its neighboring grid cells.

[0028] In one embodiment, the process of obtaining the node cost value is as follows:

[0029] The value of the movement between each grid and each of its neighboring grids is multiplied by the obstacle degree of each neighboring grid, and the result is used as the node value.

[0030] Secondly, embodiments of this application also provide an intelligent detection multifunctional robot autonomous navigation system, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement the steps of any of the methods described above.

[0031] The embodiments of this application have at least the following beneficial effects:

[0032] This application introduces static cost and accessibility models to enable the planned path to proactively avoid static high-risk areas and select safer, smoother global paths. This reduces path failures and recalculations caused by local obstacles, significantly improving the efficiency and accuracy of path planning. Simultaneously, by analyzing historical dynamic obstacle data to construct a dynamic impact factor, the robot can predict and avoid areas where dynamic obstacles frequently appear, improving the continuity and smoothness of the navigation process, reducing the risk of task interruption, and enhancing adaptability and predictability to dynamic environments. Finally, by quantifying the spatial disorder of obstacle distribution, the assessment of movement costs between nodes is optimized, more accurately reflecting the control difficulty and energy consumption of the robot's actual movement. This ensures that the generated path is not only geometrically optimal but also better suited to the robot's motion characteristics, improving the practicality and reliability of navigation. Attached Figure Description

[0033] To more clearly illustrate the technical solutions and advantages in the embodiments of this application 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 some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0034] Figure 1 A flowchart illustrating the steps of an intelligent detection multifunctional robot autonomous navigation method provided in one embodiment of this application;

[0035] Figure 2 This is a schematic diagram of the static cost distribution of the grid.

[0036] Figure 3 This is a schematic diagram of the distribution of obstacles within the grid. Detailed Implementation

[0037] To further illustrate the technical means and effects adopted by this application to achieve the intended inventive purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of an intelligent detection multifunctional robot autonomous navigation method and system proposed in this application. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0038] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0039] The following description, in conjunction with the accompanying drawings, details the specific scheme of the autonomous navigation method and system for a smart detection multifunctional robot provided in this application.

[0040] Please see Figure 1 The diagram illustrates a flowchart of an intelligent detection multifunctional robot autonomous navigation method according to an embodiment of this application. The method includes the following steps:

[0041] Step S1: Obtain a static environmental grid map of the working area of ​​the intelligent detection robot and multimodal information of each dynamic obstacle detected in history.

[0042] First, an intelligent multi-functional robot equipped with a LiDAR sensor automatically cruises around the intelligent engineering materials testing laboratory, where there are no other moving objects, ensuring that the sensor can scan the environment from all directions. Simultaneously, a static environmental map (grid map) of the laboratory is constructed using the SLAM algorithm to estimate the robot's own pose. The process of constructing the static map is a well-known method and will not be elaborated here. The robot navigation system generates a large amount of pose information during operation, including acquiring the robot's pose information from historical tasks and using a global planner to obtain the A positional information for each robot operation. The algorithm-planned path and the dynamic obstacles encountered by the robot in each historical task are used to obtain the obstacle attributes, including position, speed, size, and category, by running a real-time obstacle tracking algorithm on the robot, such as laser clustering and Kalman filtering or multi-object tracking algorithm. The obstacle tracking algorithm used in this application is a multi-object tracking algorithm, but implementers may use other obstacle tracking algorithms, and this application does not impose specific restrictions.

[0043] Step S2: Based on the distribution of the location and number of static obstacles in each grid and the number of edge corners of the static obstacles, construct the static cost of each grid; construct the passability of each grid based on the distribution characteristics of the static cost of the neighboring grids of each grid, so as to optimize the static cost of each grid.

[0044] Based on the above steps, the static environment map of the intelligent testing laboratory for engineering materials and the operational data from historical experimental tasks of the intelligent testing multifunctional robot were acquired. Using A... When algorithms plan robot motion trajectories, traditional A Algorithms typically determine the node cost of a grid by measuring the distance between grid cells and the time consumed, thereby generating a static optimal path. However, they neglect the difficulty of traversing nodes and the changing conditions of dynamic obstacles. When dynamic obstacles (such as pedestrians) appear in the environment, the original path may immediately become invalid. Furthermore, recalculating a new path may result in prolonged pauses due to the computation of the new path, making it difficult to respond promptly to environmental changes and experimental task requirements. This can lead to sluggishness or even blockages.

[0045] In the experimental testing process of the intelligent testing laboratory for engineering materials, robots typically need to move from a fixed position to the area of ​​engineering materials to be tested. Using a robotic arm, they grasp the samples, perform a preliminary scan, and then move the scanned materials to different workstations in the testing process, thus connecting the entire testing workflow. Because the workstations for each test are fixed, the robot's movement process in each historical engineering material testing task in the laboratory is similar, always requiring it to pass through different fixed testing workstations. In the static environment map of the intelligent testing laboratory for engineering materials, this is represented by multiple target nodes that need to be reached sequentially.

[0046] For each grid cell of the static environment map of the intelligent testing laboratory for engineering materials, using A When the algorithm performs path planning, it creates a node object for each grid cell that needs to be explored; that is, each grid cell corresponds to one node. First, it obtains the node cost of each grid cell. Traditional A... Algorithms typically determine the node cost between grid cells by the path length between them, ignoring the difficulty of passage and the dynamic changes in obstacles. Taking any grid cell as the current grid cell, the static cost of the grid cell is first constructed using the static obstacle information within the cell, expressed as:

[0047] In the formula, Indicates the static cost of the current grid cell; A represents the number of static obstacles within the current grid cell. This represents the number of edge corner points of the i-th static obstacle within the current grid. In this embodiment, the edge corner points of static obstacles in the static environment map are obtained through Harris corner detection. This represents the ratio of the area of ​​the i-th static obstacle within the current grid to the area of ​​the current grid. This represents the distance between the i-th static obstacle in the current grid and its nearest other static obstacle, denoted as the first distance. Harris corner detection is a well-known method, and its specific process will not be elaborated further. It should be noted that implementers may also use other corner detection algorithms to obtain the edge corners of static obstacles; this application does not impose specific restrictions.

[0048] The more static obstacles and the more edge corners of the static obstacles there are in a grid, the more complex the obstacles are in the grid, and the more difficult it is for the robot to pass through the grid. The ratio of the area of ​​the static obstacle to the area of ​​the grid is used as a weighting factor for the number of edge corners. The larger the value, the greater the impact on passage. This represents the minimum distance between the i-th static obstacle and the other static obstacles within the grid. The smaller the value, the smaller the gap between the obstacles, the smaller the space for the robot to pass through, and the greater the static cost.

[0049] The static cost of each grid cell in the static environment map of the intelligent testing laboratory for engineering materials can be obtained by following the above steps, and the static cost of all grid cells can be normalized using the max-min normalization algorithm. The max-min normalization algorithm is a well-known technique, and its specific process will not be elaborated further.

[0050] It should be noted that this application provides only one normalization method for normalizing the static cost of the raster. There are many existing normalization methods, and implementers may also use other normalization methods to normalize the static cost of the raster. This application does not impose any specific restrictions.

[0051] Static cost estimation initially reflects the risk of the robot traversing the grid using static obstacles within it; however, it neglects the subsequent risks associated with grid movement, such as... Figure 2 As shown, let Figure 2 In the diagram, grid Q represents the current grid, and the numbers in each of the other grids represent the normalized static cost values ​​of each of the other grids. When using static cost values ​​for path planning, it is easy to encounter a situation where, after selecting a grid of 0.5, the static cost values ​​of the neighboring grids of the 0.5 grid are also relatively high. This phenomenon is relative to selecting a grid of 0.6. As a result, the robot plans a path that is either risky or costly.

[0052] Therefore, to avoid such phenomena, this application further constructs the traversability based on the static cost distribution of neighboring grates, thereby reducing the traversal difficulty of the planned path. The specific process is as follows: Taking the current grates as an example, this application obtains the neighboring grates of the current grates. It should be noted that, in this embodiment, neighboring grates refer to adjacent grates in the vertical and horizontal directions. The traversability of the current grates can be constructed based on the static cost distribution of the neighboring grates, expressed as:

[0053] In the formula, Y represents the current grid's throughput; X represents the number of neighboring grid cells of the current grid. This represents the static cost normalized value of the x-th neighboring grid of the current grid; This represents the minimum absolute value of the difference between the static cost normalization values ​​of the x-th neighboring grid and the other neighboring grids. The smaller the value, the more similar the static cost distribution of the neighboring grids of the current grid is. At the same time, the smaller the static cost value, the more traversable directions the robot has when passing through the area, i.e., the greater the traversability.

[0054] Furthermore, the static cost of the current raster is optimized based on its accessibility, as expressed by:

[0055] In the formula, U represents the optimized static cost of the current grid; U represents the unoptimized static cost of the current grid; and Y represents the traversability of the current grid. A higher traversability indicates a lower static cost of neighboring grids, meaning the current grid has more traversable directions. Conversely, a lower traversability indicates a higher static cost of neighboring grids, fewer traversable directions, and a greater likelihood of planning impassable paths.

[0056] Step S3: Based on the similarity between the multimodal information of all dynamic obstacles passing through each grid and the distance between each dynamic obstacle and the robot, combined with the optimized static cost, determine the obstacle degree of each grid.

[0057] Based on the above steps, the static cost and accessibility of each grid cell in the static environment map of the intelligent testing laboratory for engineering materials were obtained, and the static cost was optimized based on the accessibility. However, while directly using the optimized static cost as the node cost for path planning can effectively avoid static obstacles, it cannot avoid dynamic obstacles in the laboratory setting. Therefore, this application further considers the fixed characteristics of the engineering material testing process in the intelligent testing laboratory for engineering materials. It is known that the dynamic obstacles encountered by the robot during the material testing process are very likely caused by fixed steps in the engineering material testing process. Therefore, this application first analyzes the multimodal information of each dynamic obstacle detected by the robot in historical testing tasks, and forms a multimodal vector from the multimodal information of each dynamic obstacle. Based on the multimodal vector of each dynamic obstacle, the dynamic impact degree of each grid cell is constructed, and the expression is:

[0058] In the formula, F is the dynamic influence degree of the current grid; N represents the number of times the robot detected a dynamic obstacle passing through the current grid in all its historical detection tasks; M represents the number of dynamic obstacles passing through the current grid in these N historical detection tasks. This represents the mean similarity between the multimodal vector of the m-th dynamic obstacle passing through the current grid and the multimodal vectors of all other dynamic obstacles passing through the current grid; This represents the minimum distance between the m-th dynamic obstacle passing through the current grid and the robot at the same instant, denoted as the first closest distance.

[0059] It should be noted that, for the similarity between multimodal vectors, the embodiments of this application determine it by calculating the cosine similarity between multimodal vectors. There are many existing algorithms for calculating similarity, and implementers may also use other similarity algorithms to calculate the similarity between multimodal vectors. This application does not impose any specific restrictions.

[0060] The larger the value, the more similar the dynamic obstacle information passing through grid Q is, and the greater the possibility of regular influence, that is, the greater the dynamic influence. The smaller the value, the greater the dynamic influence (since dynamic obstacles do not collide with the robot, D will not be 0). The greater the dynamic influence, the more dynamic obstacles pass through the grid, and the greater the impact on the robot's movement. The smaller the dynamic influence, the fewer dynamic obstacles pass through the grid, and the smaller the impact on the robot's movement.

[0061] Based on the above steps, the dynamic influence degree of each grid can be obtained. Then, the dynamic influence degree F of each grid and the optimized static cost can be calculated. The product of these factors is used as the obstacle degree Z for each grid cell. A higher obstacle degree indicates a greater impact from static and dynamic obstacles when the robot passes through that grid cell, while a lower obstacle degree indicates a smaller impact from static and dynamic obstacles when the robot passes through that grid cell.

[0062] Step S4: Divide each grid into regions. Based on the connection between each region and the regions in its neighboring grids, determine the characteristic regions between each grid and each of its neighboring grids. Based on the distribution of static obstacles in the characteristic regions and the static cost of the characteristic regions, determine the movement cost between each grid and each of its neighboring grids.

[0063] The obstacle degree of each grid cell in the static environment map of the intelligent testing laboratory for engineering materials was obtained based on the above steps. The obstacle degree characterizes the degree to which each grid cell is affected by obstacles. However, during the actual movement of the robot, the obstacle degree may not accurately reflect the movement cost of the robot moving from one grid cell to another due to the different distribution of obstacles. For example, ... Figure 3 As shown, Figure 3 It contains two grids, namely grid Q and grid W. and For two different regions in grid Q, Given a region in raster W, if raster Q contains... There are many obstacles in the area. The area has relatively few obstacles, and the overall obstacle level of grid Q may be relatively high, but during actual movement, the robot tends to... Regional movement and destination The costs of regional relocation will inevitably differ. Figure 3 This is a schematic diagram of the distribution of obstacles within the grid.

[0064] Therefore, this application further constructs the movement cost between grids based on the obstacle distribution within the grid. Taking any grid Q and grid W as an example, grid W is the horizontal or vertical neighboring grid of grid Q. Since movement between grids is usually horizontal and vertical, the grid can be divided into four equal parts by its diagonal. Each divided grid region corresponds to a neighboring grid. (See reference...) Figure 3 That is, one of the four equal regions in grid Q. If the corresponding neighboring grid W is used, then the area can be obtained. With the region The distribution of static obstacles within the formed feature region (denoted as S) is analyzed, and the static cost of feature region S is calculated in the same way as the static cost of each grid. Furthermore, the distance between each static obstacle in feature region S and its nearest other static obstacle is obtained and denoted as the second distance of each static obstacle. Further, with the center point of feature region S as the origin, feature region S is divided into fan-shaped regions, with each fan-shaped region having an angle of 1°, thereby obtaining 360 fan-shaped intervals within feature region S. The proportion of fan-shaped intervals containing obstacles is obtained and used as the proportion of obstacles within feature region S.

[0065] Furthermore, taking grid Q as the current grid, the static cost of the feature region S between the current grid and its neighboring grids W, and the information entropy of the second distance of all static obstacles in the feature region S are obtained. The fused value of the static cost of the feature region S, the information entropy, and the proportion is used as the movement cost between each grid and each of its neighboring grids. It should be noted that the fusion can be an addition, multiplication, or other fusion method. Preferably, in this embodiment, the fusion process is to multiply the static cost of the feature region S, the information entropy, and the proportion to obtain the movement cost between the current grid Q and its neighboring grids W. In other embodiments of this application, the implementer may also use other fusion methods to obtain the movement cost between the current grid Q and its neighboring grids W.

[0066] The larger the value, the more chaotic the changes in distance between static obstacles are, and the greater the degree of chaos. The larger the value, the more likely the obstacles are to be distributed in all directions of the characteristic area, and the greater the degree of disorder in the obstacle distribution.

[0067] The more chaotic the obstacle distribution, the more turns the robot needs to make to navigate the area, the higher the control precision required, and the greater the cost of movement. Conversely, the less chaotic the obstacle distribution, the easier it is for the robot to navigate the area, and the lower the cost of movement.

[0068] Step S5: Based on the obstacle degree and the movement cost, determine the node cost between each grid and each of its neighboring grids, and combine the path planning algorithm to plan the path for the robot.

[0069] Based on the above steps, the movement cost between a grid and its neighboring grids can be obtained. Then, through A... When performing path planning, the algorithm can adaptively construct the node cost for the robot to reach its neighboring nodes from a node based on the movement cost between the node and its neighboring nodes (nodes are grids) and the obstacle degree. The expression is:

[0070] In the formula, DJ represents the node cost of reaching the neighboring node W from the node corresponding to grid Q, Z represents the obstacle degree of the neighboring grid node reached by the robot, and H represents the movement cost of reaching the neighboring node W from the node corresponding to grid Q. The obstacle degree, as a weighting factor, characterizes the impact of obstacles on the robot after reaching the neighboring node. The larger the obstacle degree value, the greater the difficulty of path planning after reaching the neighboring node, i.e., the greater the cost of subsequent path planning after reaching the neighboring node.

[0071] Based on the steps described above, the node cost values ​​from the starting node to each neighboring node are obtained. The algorithm starts from the starting node and dynamically maintains the nodes to be expanded using a priority queue (usually a min-heap structure). Each time, the node with the lowest cost value is proposed for expansion. For each neighboring node of the current node, its actual cost g(n) (i.e., the sum of the cumulative DJ values ​​from the starting point to that node) and heuristic cost h(n) (usually an estimate of the Euclidean or Manhattan distance to the target node) are calculated, and the node is evaluated using f(n) = g(n) + h(n) as the total cost function. If a neighboring node is not visited or a smaller cost value is found, its g(n) is updated and added to the priority queue. This process iterates until the target node is reached. At this point, the optimal path from the starting point to the target node is generated by tracing the parent node chain in reverse. Where A... The algorithm is publicly known. This application uses the above process to modify A. The node cost between grids in the algorithm is reset, and the specific process of obtaining the optimal path is well known and will not be described in detail in this application.

[0072] Based on the above steps, the optimal path between the starting point and the target node is planned. Since there are multiple target nodes in the intelligent testing laboratory for engineering materials, when the robot reaches the first target node, the first target node is used as the new starting point. The above steps are repeated to obtain the optimal path between the first target node and the second target node. This process is repeated to complete the path planning of the intelligent testing multifunctional robot.

[0073] Based on the same inventive concept as the above method, this application embodiment also provides an intelligent detection multifunctional robot autonomous navigation system, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of any one of the above-described intelligent detection multifunctional robot autonomous navigation methods.

[0074] In summary, this application provides an intelligent detection multifunctional robot autonomous navigation method. By introducing static cost and accessibility models, the planned path can proactively avoid static high-risk areas and select a safer, smoother global path, reducing path failures and recalculations caused by local obstacles, significantly improving the efficiency and accuracy of path planning. Simultaneously, by analyzing historical dynamic obstacle data to construct a dynamic influence degree, the robot can predict and avoid areas where dynamic obstacles frequently appear, improving the continuity and smoothness of the navigation process, reducing the risk of task interruption, and enhancing adaptability and predictability to dynamic environments. Finally, by quantifying the spatial disorder of obstacle distribution, the evaluation of movement costs between nodes is optimized, more accurately reflecting the control difficulty and energy consumption of the robot's actual movement. This ensures that the generated path is not only geometrically optimal but also more consistent with the robot's motion characteristics, improving the practicality and reliability of navigation.

[0075] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this application. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0076] The various embodiments in this application are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0077] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the principles of this application should be included within the protection scope of this application.

Claims

1. A method for autonomous navigation of a smart detection multifunctional robot, characterized in that, The method includes the following steps: Acquire a static environmental grid map of the working area of ​​the intelligent inspection robot and multimodal information of each dynamic obstacle detected in history; Based on the distribution of the location and number of static obstacles in each grid, as well as the number of edge corners of the static obstacles, the static cost of each grid is constructed; based on the distribution characteristics of the static cost of each grid's neighboring grids, the passability of each grid is constructed to optimize the static cost of each grid. Based on the similarity between the multimodal information of all dynamic obstacles passing through each grid, and the distance between each dynamic obstacle and the robot, combined with the optimized static cost, the obstacle degree of each grid is determined. Each grid is divided into regions. Based on the connectivity between each region and its neighboring regions, the characteristic regions between each grid and each of its neighboring grids are determined. Based on the distribution of static obstacles within the characteristic regions and the static cost of the characteristic regions, the movement cost between each grid and each of its neighboring grids is determined. Based on obstacle degree and movement cost, the node cost between each grid and each of its neighboring grids is determined, and combined with the path planning algorithm, the robot performs path planning. The process of obtaining the obstacle degree of each grid is as follows: First, determine the multimodal vector of each dynamic obstacle using its multimodal information. Second, among all dynamic obstacles passing through the current grid, calculate the mean similarity between the multimodal vectors of any dynamic obstacle and all other dynamic obstacles, and obtain the minimum distance between each dynamic obstacle and the robot at the same instant, denoted as the first closest distance. Third, obtain the number of times the robot detected a dynamic obstacle passing through the current grid in all its historical detection tasks. Fourth, determine the dynamic influence degree of the current grid based on the number of times, the mean, and the first closest distance. The dynamic influence degree of the current grid is positively correlated with the number of times and the mean, and negatively correlated with the first closest distance. Fifth, use the product of the dynamic influence degree of each grid and the optimized static cost degree as the obstacle degree of each grid. The process of obtaining the feature region is as follows: If any region of raster Q... Any region of its neighboring grid W If connected, the area will be... With the region The region formed is considered as the feature region between grid Q and its neighboring grid W; The process of obtaining the movement cost value is as follows: For each grid cell and each of its neighboring grid cells, the distance between each static obstacle in the feature region and the nearest other static obstacle is recorded as the second distance of each static obstacle, and the information entropy of the second distance of all static obstacles in the feature region is obtained; the feature region is divided into regions, and the proportion of the region occupied by obstacles in the feature region is obtained; the fusion value of the static cost value, information entropy and proportion of the feature region is used as the movement cost value between each grid cell and each of its neighboring grid cells.

2. The autonomous navigation method for a smart detection multifunctional robot as described in claim 1, characterized in that, The process of obtaining the static cost of each grid is as follows: Obtain the number of static obstacles in each grid cell and the area percentage of each static obstacle in the grid cell; obtain the distance between each static obstacle in each grid cell and its nearest other static obstacle, and record it as the first distance; The static cost of each grid is positively correlated with the number of static obstacles, the area ratio of all static obstacles, and the number of edge corner points, and negatively correlated with the first distance of all static obstacles.

3. The autonomous navigation method for a smart detection multifunctional robot as described in claim 1, characterized in that, The process of obtaining the passability of each grid is as follows: Obtain the minimum difference between the static cost normalization values ​​of any neighboring grid and all other neighboring grids for each grid; the accessibility of each grid is negatively correlated with the static cost normalization values ​​and the minimum value of all its neighboring grids.

4. The autonomous navigation method for a smart detection multifunctional robot as described in claim 1, characterized in that, The expression for optimizing the static cost of each grid cell is as follows: In the formula, U represents the optimized static cost of the current grid; U represents the unoptimized static cost of the current grid; and Y represents the passability of the current grid.

5. The autonomous navigation method for a smart detection multifunctional robot as described in claim 1, characterized in that, The expression for dynamic influence is: In the formula, F is the dynamic influence degree of the current grid; N is the number of times; and M is the number of dynamic obstacles passing through the current grid. This represents the average value of the m-th dynamic obstacle passing through the current grid. It represents the first closest distance to the m-th dynamic obstacle in the current grid.

6. The autonomous navigation method for a smart detection multifunctional robot as described in claim 1, characterized in that, The process of obtaining the node cost value is as follows: Multiply the movement cost between each grid and each of its neighboring grids by the obstacle degree of each neighboring grid, and the result is used as the node cost.

7. An intelligent detection multifunctional robot autonomous navigation system, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method as described in any one of claims 1-6.

Images