A control method of a mobile robot

Abstract

The application provides a mobile robot control method, and relates to the technical field of robot control. The mobile robot control method comprises the following steps: S1. acquiring a local grid map of a mobile robot currently constructed, and identifying key topological structures of a channel entrance, a corridor section and a boundary of an unknown area in the map; S2. constructing a non-uniform sampling probability distribution model facing a high information gain area; S3. generating random sampling points in a global sampling space, and constructing a global random tree; S4. supplementally sampling and verifying a path for a near-field area of the robot; S5. screening an effective boundary point set in the local grid map, and determining a target point in combination with a path cost and information gain; S6. controlling the mobile robot to move to the target point, synchronously updating the local grid map, and repeating the above steps until the whole area exploration is completed. Through the combination of topological guidance sampling, invalid exploration is reduced, the mapping is faster and more complete, and the path is more optimal and stable.

Description

Technical Field

[0001] This invention relates to the field of robot control technology, specifically a control method for a mobile robot. Background Technology

[0002] Autonomous exploration and environmental modeling technologies for mobile robots are the core foundation for enabling robots to navigate autonomously, plan paths, and execute tasks in unknown environments. In practical applications such as warehousing and logistics, industrial inspection, and household services, robots need to use their own sensors to complete environmental perception, boundary recognition, and path decision-making without prior maps, gradually building a complete and usable environmental grid map. Current mainstream autonomous exploration solutions mostly adopt path planning and exploration strategies based on Rapid Expanding Random Trees (RRTs). This type of method is widely used in real-time exploration systems for mobile robots due to its low computational cost, high adaptability, and lack of need for complex environmental modeling.

[0003] Among existing related technologies, the invention patent with authorization announcement number CN114153210B discloses a robot movement control method. This method generates a global random tree and a local random tree within a preset area, setting the node expansion step size of the global random tree to be negatively correlated with the distance from the node to the starting point, achieving rapid exploration of near areas and detailed exploration of far areas. Simultaneously, it determines the robot's target movement point by selecting boundary points and calculating boundary point gain values, driving the robot to complete environmental exploration and map updates. This scheme improves to some extent the problems of poor uniformity and low traversal efficiency in traditional random tree exploration, but it still has significant shortcomings in real-world complex scenarios.

[0004] 1. This method still adopts a global uniform random sampling mechanism. The distribution of sampling points is not guided by environmental structure. In scenarios where open areas account for a large proportion and include narrow passages, corners and boundary-type unknown areas, a large number of sampling points are concentrated in the explored open areas, which makes it difficult to effectively cover the boundary and passage positions with high information gain. This leads to the lag in the discovery of boundary points in key areas. The robot repeatedly grows random trees in low-value areas for a long time, resulting in high computational resource consumption.

[0005] 2. Random tree growth relies on random sampling points and lacks active recognition and utilization of map topology. In typical scenarios where open areas connect to narrow passages, the random tree may not be able to effectively extend to the passage entrance. The robot may repeatedly circle around in open areas, with highly overlapping exploration paths, significantly lengthening the overall mapping cycle.

[0006] 3. Boundary point selection and target point decision-making rely solely on information gain and path length, without incorporating map topological features for weighting. This makes it easy to select locally optimal boundary points, making it difficult to achieve globally optimal exploration. Ultimately, this results in insufficient map construction completeness, high redundancy in exploration paths, and overall mapping efficiency that fails to meet practical usage requirements.

[0007] In real indoor and outdoor environments, robots often face layouts where open spaces coexist with narrow passageways (as shown in the instruction manual). Figure 6 As shown in the figure, neither the traditional uniform sampling RRT scheme nor the method described in CN114153210B can actively adjust the sampling preference and exploration direction according to the environmental topology, making it difficult to balance exploration speed, path rationality and map completeness. Therefore, in order to address the shortcomings of the existing technology, a mobile robot control method with topology perception, strong sampling adaptability and the ability to avoid repeated exploration is proposed, which has important practical significance and engineering application value. Summary of the Invention

[0008] Technical problems to be solved

[0009] To address the shortcomings of existing technologies, this invention provides a control method for mobile robots, which solves the problems of lagging boundary discovery, repetitive exploration, and low mapping efficiency caused by uniform sampling in traditional RRT exploration algorithms in open and narrow passage scenarios, thereby improving the efficiency of robot autonomous exploration and map completion.

[0010] Technical solution

[0011] To achieve the above objectives, the present invention provides the following technical solution: a control method for a mobile robot, comprising the following steps:

[0012] S1. Obtain the local grid map that the mobile robot has built, and extract the topological skeleton of the local grid map to identify the key topological structures of the passage entrance, corridor section and unknown area boundary in the map;

[0013] S2. Based on the key topology, construct a non-uniform sampling probability distribution model for high information gain regions, suppress the sampling probability of explored and open regions, and enhance the sampling probability of regions corresponding to the key topology.

[0014] S3. Based on the non-uniform sampling probability distribution model, generate random sampling points in the global sampling space and construct a global random tree;

[0015] S4. Construct a local random tree based on the current position of the mobile robot, and perform supplementary sampling and path verification in the robot's near-field area;

[0016] S5. Filter the set of valid boundary points in the local raster map, and determine the target point by combining path cost and information gain;

[0017] S6. Control the mobile robot to move towards the target point, and update the local grid map synchronously. Repeat the above steps until the entire area is explored.

[0018] Preferably, the specific process of topological skeleton extraction in S1 is as follows: binarize the local grid map, extract the central axis skeleton of the passable area, identify the structure of the passage entrance and corridor section based on the connection relationship of the skeleton nodes, and mark the connected branches of the unknown area boundary.

[0019] Preferably, the specific method for constructing the non-uniform sampling probability distribution model in S2 is as follows: define a sampling weight function, and linearly reduce the sampling probability according to the distance for open areas that are more than a preset threshold away from the robot's initial position; linearly increase the sampling probability according to the distance for areas near the entrance of the channel, corridor section and the boundary of unknown areas.

[0020] Preferably, when constructing the global random tree in S3, a dynamic step size that is negatively correlated with the distance from the node to the initial position is used for node expansion. The closer the node is to the initial position, the larger the expansion step size; the farther the node is from the initial position, the smaller the expansion step size.

[0021] Preferably, the specific method for constructing the local random tree in S4 is as follows: taking the current position of the mobile robot as the root node, sampling and expanding within a preset radius of the robot's near field using a fixed step size; when the local random tree fails to find an effective unknown area after multiple consecutive expansions, the local random tree is cleared and reconstructed.

[0022] Preferably, the specific process of filtering the set of valid boundary points in S5 is as follows: mark the boundary points between passable areas and unknown areas in the local grid map, filter out invalid points that overlap with obstacles or are located outside the map boundary, perform deduplication processing on duplicate boundary points, and retain boundary points that are closer to the robot's current position.

[0023] Preferably, the specific method for determining the target point in S5 is as follows: calculate the information gain value of each effective boundary point. This value is positively correlated with the area of ​​the unknown region that can be explored by the boundary point and negatively correlated with the path length from the mobile robot to the boundary point. Select the boundary point with the largest information gain value as the target point.

[0024] Preferably, in step S6, when the mobile robot moves toward the target point, it simultaneously collects environmental data through onboard sensors and updates the traversable and unknown areas markers on the local grid map. When there are no unknown areas in the local grid map, the exploration process is terminated.

[0025] Beneficial effects

[0026] This invention provides a control method for a mobile robot. It has the following beneficial effects:

[0027] 1. This invention provides a control method for a mobile robot. By extracting the topological skeleton of a grid map and constructing a non-uniform sampling probability model, it changes the uniform random sampling method in the comparative patent. This method prioritizes the distribution of sampling points in high information gain areas such as channel entrances and unknown boundaries, significantly reducing invalid growth and repeated exploration in open areas, and improving the efficiency of random tree exploration and the speed of boundary discovery.

[0028] 2. This invention provides a control method for a mobile robot. Based on a dual-tree structure that combines global dynamic step size and local fixed step size, it incorporates a topological feature guidance and boundary point weighted screening mechanism to optimize the target point decision logic, making the robot's exploration path more reasonable, effectively avoiding local optima and detour problems, and maintaining high mapping integrity and system stability even in complex scenarios. Attached Figure Description

[0029] Figure 1 This is an overall flowchart of the mobile robot control method of the present invention;

[0030] Figure 2 This is a flowchart illustrating the topological skeleton extraction and non-uniform sampling model construction process in this invention.

[0031] Figure 3 This is a flowchart illustrating the construction of the global random tree and the local random tree in this invention.

[0032] Figure 4 This is a flowchart illustrating the boundary point selection and target point determination process in this invention.

[0033] Figure 5 This is a flowchart of the robot movement, map update, and exploration termination determination in this invention;

[0034] Figure 6 This is a schematic diagram of a robot exploring in open areas and narrow passages in existing technologies. The triangle represents the robot's initial position, and the narrow area on the right represents the edge of the map and an unknown area with high information gain. Detailed Implementation

[0035] 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.

[0036] like Figure 1-6 As shown in the figure, an embodiment of the present invention provides a control method for a mobile robot, the control method including the following steps:

[0037] S1. Obtain the local grid map that the mobile robot has built, and extract the topological skeleton of the local grid map to identify the key topological structures of the passage entrance, corridor section and unknown area boundary in the map;

[0038] S2. Based on the key topology, construct a non-uniform sampling probability distribution model for high information gain regions;

[0039] S3. Based on the non-uniform sampling probability distribution model, random sampling points are generated in the global sampling space to construct a global random tree;

[0040] S4. Construct a local random tree based on the current position of the mobile robot, and perform supplementary sampling and path verification in the robot's near-field area;

[0041] S5. Filter the set of valid boundary points in the local raster map, and determine the target point by combining path cost and information gain;

[0042] S6. Control the mobile robot to move towards the target point, and update the local grid map synchronously. Repeat the above steps until the entire area is explored.

[0043] The following detailed explanation of each step is provided in conjunction with the accompanying drawings:

[0044] In this embodiment, the resolution of the local raster map is set to 0.05 m / square, and the map size is 100 m × 100 m. First, the raster map is binarized, dividing the grid into passable areas (value 0) and obstacle areas (value 1). Then, a distance-transform-based skeleton extraction algorithm is used to extract the central axis skeleton of the passable areas. After skeleton extraction, the connection relationships between skeleton nodes are used to identify the structure of passage entrances and corridor segments, and the connected branches of unknown area boundaries are marked.

[0045] Based on the above topology, a non-uniform sampling probability distribution model is constructed. The sampling weight function is defined as follows:

[0046]

[0047] in, For the weighting coefficients of the explored regions, The weight coefficients for the topologically critical regions satisfy the following conditions: ; The suppression function for the explored region is defined as:

[0048]

[0049] in For point The distance to the robot's initial position. In this embodiment, the preset distance threshold is taken as... ; The topological critical region enhancement function is defined as:

[0050]

[0051] in For point Distance to the nearest topological critical node For the attenuation coefficient, in this embodiment, we take... , To enhance the radius of key areas, take .

[0052] In this embodiment, the following settings are provided: , This allows sampling points to be preferentially distributed near the entrance of the passage, the corridor section, and the boundary of unknown areas, thus suppressing invalid sampling in open areas.

[0053] After generating sampling points based on a non-uniform sampling probability distribution model, a global random tree is constructed. In this embodiment, the node expansion of the random tree adopts a dynamic step size that is negatively correlated with the distance from the node to the initial position. The step size calculation formula is as follows:

[0054]

[0055] in, To preset the maximum step size, take ; This represents the distance from the current node to the initial position. The step size decay coefficient is taken as in this embodiment. That is, the closer to the initial position, the larger the step size, allowing the robot to quickly traverse open areas; the farther from the initial position, the smaller the step size, ensuring exploration accuracy in complex areas.

[0056] A local random tree is constructed with the robot's current position as the root node. The local sampling radius is set to 3m, and the fixed step size is 0.3m. If the local random tree fails to find a valid unknown region after five consecutive expansions, the local random tree is cleared and reconstructed to avoid getting stuck in local ineffective exploration.

[0057] In this embodiment, the process of selecting valid boundary points is as follows:

[0058] Mark the boundary between passable and unknown areas in the raster map;

[0059] Filter out invalid points that overlap with obstacles or are located outside the map boundaries;

[0060] Repeated boundary points with a distance of less than 0.5m are deduplicated;

[0061] Retain boundary points within 10m of the robot's current position to reduce unnecessary calculations.

[0062] For the selected valid boundary points, calculate their information gain value:

[0063]

[0064] in, The area of ​​the unknown region that can be explored at the boundary point is calculated by counting the number of unknown grid cells within a 5m radius of the boundary point. The path length from the robot to this boundary point is obtained by planning the path using the A* algorithm. The boundary point with the largest information gain value is selected as the target point.

[0065] As the robot moves toward the target point, it collects environmental data at a frequency of 10Hz using an onboard LiDAR, updating the passable areas, unknown areas, and obstacle markers on the grid map in real time. When the percentage of unknown areas in the grid map is less than 1%, the map exploration is considered complete, random tree generation is terminated, and the exploration process ends.

[0066] This embodiment differs from the traditional uniform sampling RRT algorithm in that... Figure 6 The performance comparison data for the scenarios shown are shown in the table below:

[0067] First discovery time of boundary points s 128 32 Reduce by 75% Repeated exploration path length m 186 42 Reduced by 77.4% Total exploration time s 256 98 Reduced by 61.7% Map completion % 92.3 99.7 An increase of 7.4%

[0068] As shown in the table above, the method of the present invention, through non-uniform sampling guided by the topological skeleton, significantly reduces invalid sampling and repeated exploration in open areas, accelerates the discovery of key channel boundary points, and significantly improves mapping efficiency and map completion.

[0069] Figure 6 This diagram illustrates a robot's exploration in an open area with a narrow passage, as described in existing technologies. The triangle marks the robot's initial position, and the narrow area on the right represents the map edge and a high-information-gain unknown region. In this scenario, the traditional uniform sampling RRT algorithm is prone to problems such as delayed boundary detection and repeated exploration due to blind sampling. The method of this invention effectively solves this deficiency through topology-guided sampling.

[0070] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A control method for a mobile robot, characterized in that, Includes the following steps: S1. Obtain the local grid map that the mobile robot has built, and extract the topological skeleton of the local grid map to identify the key topological structures of the passage entrance, corridor section and unknown area boundary in the map; S2. Based on the key topology, construct a non-uniform sampling probability distribution model for high information gain regions, suppress the sampling probability of explored and open regions, and enhance the sampling probability of regions corresponding to the key topology. S3. Based on the non-uniform sampling probability distribution model, generate random sampling points in the global sampling space and construct a global random tree; S4. Construct a local random tree based on the current position of the mobile robot, and perform supplementary sampling and path verification in the robot's near-field area; S5. Filter the set of valid boundary points in the local raster map, and determine the target point by combining path cost and information gain; S6. Control the mobile robot to move towards the target point, and update the local grid map synchronously. Repeat the above steps until the entire area is explored.

2. The control method for a mobile robot according to claim 1, characterized in that, The specific process of topological skeleton extraction in S1 is as follows: binarize the local raster map, extract the central axis skeleton of the passable area, identify the structure of the passage entrance and corridor section based on the connection relationship of the skeleton nodes, and mark the connected branches of the unknown area boundary.

3. The control method for a mobile robot according to claim 1, characterized in that, The specific method for constructing the non-uniform sampling probability distribution model in S2 is as follows: define a sampling weight function, and linearly reduce the sampling probability according to the distance for open areas that are more than a preset threshold away from the robot's initial position; linearly increase the sampling probability according to the distance for areas near the entrance of the channel, corridor section and the boundary of unknown areas.

4. The control method for a mobile robot according to claim 1, characterized in that, When constructing the global random tree in S3, a dynamic step size that is negatively correlated with the distance from the node to the initial position is used to expand the node. The closer the node is to the initial position, the larger the expansion step size; the farther the node is from the initial position, the smaller the expansion step size.

5. The control method for a mobile robot according to claim 1, characterized in that, The specific method for constructing the local random tree in S4 is as follows: taking the current position of the mobile robot as the root node, sampling and expansion are performed within the preset radius of the robot's near field using a fixed step size. When the local random tree fails to find an effective unknown area after multiple consecutive expansions, the local random tree is cleared and reconstructed.

6. The control method for a mobile robot according to claim 1, characterized in that, The specific process of filtering the set of valid boundary points in S5 is as follows: mark the boundary points between passable areas and unknown areas in the local grid map, filter out invalid points that overlap with obstacles or are located outside the map boundary, deduplicate duplicate boundary points, and retain boundary points that are closer to the robot's current position.

7. The control method for a mobile robot according to claim 1, characterized in that, The specific method for determining the target point in S5 is as follows: calculate the information gain value of each effective boundary point. This value is positively correlated with the area of ​​the unknown region that can be explored by the boundary point and negatively correlated with the path length from the mobile robot to the boundary point. Select the boundary point with the largest information gain value as the target point.

8. The control method for a mobile robot according to claim 1, characterized in that, In step S6, when the mobile robot moves toward the target point, it simultaneously collects environmental data through onboard sensors and updates the markers for passable and unknown areas on the local grid map. When there are no unknown areas in the local grid map, the exploration process is terminated.

Images