A mobile robot motion planning method for unknown target collection

Abstract

The application provides a mobile robot motion planning method for unknown target collection, comprising the following steps: step S1, a mobile robot synchronously constructs a grid map during movement; step S2, a boundary region and a passable region are extracted; step S3, the mobile robot continuously explores the environment; step S4, the exploration degree of a local environment is judged according to a balance function; step S5, if the local environment is explored completely, the mobile robot collects targets in a task region; step S6, the mobile robot completes exploration and region coverage collection in a subregion; and step S7, when all unknown environments are explored completely and all targets in the task region are collected completely, the unknown target collection task is completed. The mobile robot motion planning method for unknown target collection has good performance, the mobile robot can quickly complete exploration of the unknown environment and collection of the targets in the task region, and can effectively meet the task demand.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of mobile robot motion planning, and particularly to a mobile robot motion planning method for unknown target collection. BACKGROUND

[0002] Unknown environment target collection is an important research direction in the field of robots, and is widely used in stain cleaning, search and rescue and other fields. Since stains and search targets are usually distributed in different sub-regions of the workspace and the robot does not have prior information about the environment, the robot needs to use on-board sensors to explore unknown regions in the workspace, identify the locations of all specific task regions, and traverse all targets in the collection region.

[0003] Unknown environment target collection task relies on robot autonomous exploration and full coverage path planning technology. The boundary-based method and the viewpoint sampling-based method are the main methods for solving the problem of robot autonomous exploration. Yamauch defines the intersection of known regions and unknown regions as a boundary, and the robot continuously moves to the boundary to explore the environment. Although this strategy can completely explore the unknown environment, it is easy to fall into the dilemma of local optimization. Wan et al. regard the boundary region as a candidate exploration target point region, and design a utility function to select the boundary point with the maximum function value in the candidate region as the exploration target point. Umari et al. combine the information gain contained in the node, and use global and local RRT trees to explore the environment. Lu et al. propose a sample-based frontier block detection method (SFBD) to reduce unnecessary sampling operations. After exploring all unknown regions, an offline coverage algorithm is called to cover all targets in the specific task region. However, the coverage order of each scattered task region is a traveling salesman problem (TSP), and when the number of task regions to be covered increases, the complexity of the problem increases exponentially. Therefore, this problem is NP-hard. In addition, separating the environment detection and task region coverage will increase the robot's repeated exploration actions, and consume additional time and energy.

[0004] Online full coverage path planning method is another method for solving unknown environment target collection task, which does not rely on prior information about the environment, and can complete the full coverage of the task region while exploring the unknown region,

[0005] The template-based method predefines the motion pattern mapped by the environment, and adopts the stored motion template to complete the coverage task when facing specific environmental features. BA * The plow-based method generates a coverage path, and uses A *The search algorithm plans an escape path to escape from the dead point state, fully covers the new uncovered free area, the Spiral-STC constructs a main line through the map, and the spiral path of full coverage is formed by moving along one side of the main line.

[0006] ∈ * The algorithm recursively decomposes to construct a hierarchical potential field map to guide the robot to generate back-and-forth coverage paths. The PPCPC is inspired by the predator and prey in nature, simulates the prey-predation behavior to guide the robot to plan a full coverage path,

[0007] The above online full coverage planning algorithm can complete the target collection task in a completely unknown or partially unknown environment, but the path will cover the entire free area rather than only the task area, resulting in a large amount of invalid redundant path.

[0008] In the face of unknown environment and task target information, how to effectively design a planning scheme that enables the mobile robot to quickly explore the environment and completely collect targets in multiple areas has become an important issue to be discussed and solved in this field. SUMMARY

[0009] Therefore, the purpose of the present application is to provide a mobile robot motion planning method for unknown target collection, which can quickly complete the exploration of unknown environment and the collection of targets in the task area, and can effectively meet the task requirements.

[0010] To achieve the above purpose, the present application adopts the following technical scheme: a mobile robot motion planning method for unknown target collection, comprising the following steps:

[0011] Step S1, the mobile robot constructs a grid map using the data obtained by the sensor carried thereon, and the grid has three states, namely passable free grid, obstacle grid and unexplored grid;

[0012] Step S2, the mobile robot extracts the boundary region and the passable region according to the constructed grid map;

[0013] Step S3, the mobile robot uses a planning method based on boundary guidance to continuously explore the environment;

[0014] Step S4, the mobile robot uses a balance framework to judge the exploration degree of the local environment according to a balance function;

[0015] Step S5, if the local environment is explored completely, the mobile robot uses a planning method based on coverage point reward function guidance to perform target collection task in the task area, and if the local environment is not explored completely, returns to step 2;

[0016] Step S6, after the robot completes the exploration of the sub-area and the area coverage collection, when there are unexplored areas or the target is not collected, return to step 2 to continue;

[0017] Step S7, when all unknown environments have been explored and all task area targets have been collected, the unknown target collection task is completed.

[0018] In a preferred embodiment: the specific method of extracting the boundary area in step S2 is that the mobile robot traverses the free passable grid in the grid map, and the free grid with at least one adjacent unexplored grid is marked as a boundary point and added to the boundary set.

[0019] In a preferred embodiment: the boundary-based planning method of step S3 specifically includes the following steps:

[0020] Step S31: sampling the exploration points of the observed environment in the boundary set with a distance δ and adding them to the candidate exploration point set;

[0021] Step S32: calculating the function value of each candidate point according to the information gain function and selecting the point with the maximum function value as the next exploration target point, the information gain function G(v i ) is as follows:

[0022]

[0023] Where S(W obs (v i )) and S(W unexp (v i )) are the number of obstacle boundary grid and unexplored grid within a certain radius r around v i , ω o represents the weight of the number of obstacle boundary units, K(v0,v i ) is the Manhattan distance from the current position v0 of the robot to the candidate view v i , S(M tar (v i )) is the number of task area grids within a certain radius around v i ;

[0024] Step S33: the mobile robot plans a path from the current unknown of the robot to the next exploration target point through A * algorithm;

[0025] Step S34: the robot receives the exploration path planned by the algorithm, and converts the path in the solution of the exploration path into the required linear velocity and angular velocity to control the robot to move to the target position.

[0026] In a preferred embodiment, the specific method of constructing the balance framework in step S4 is: the robot calculates the change rate of the number of boundary points and free grids in the map before and after exploration, constructs an environment exploration quantization function, if the function value is less than 0, it indicates that the local area is completely explored, if the function value is greater than 0, it indicates that the local area is not completely explored, and the function calculation formula is:

[0027]

[0028] Wherein, △(S F ) and △(S M ) are the difference values of the number of boundary points and the number of free area grids before and after exploration, L(σ e ) is the length of the exploration path, which is represented by the number of exploration path grids, ω k is the balance framework weight factor, when ω k is equal to 0, the algorithm degenerates to completely explore the map and then cover the task area.

[0029] In a preferred embodiment, the planning method based on the coverage point reward function guidance in step S5 specifically includes the following steps:

[0030] Step S51: the mobile robot removes the exploration path grid located in the task area and adjacent to the obstacle, and the remaining task area free grid is taken as the current to-be-covered area set

[0031] Step S52: use the function SetConductor(·) to select the grid closest to the robot position Manhattan distance in the set as the starting point P s of the coverage path, and the next exploration target point as the coverage guide point P g ;

[0032] Step S53: use the function Near(·) to check the adjacent grid of o k in , if the grid does not belong to the covered grid set or the obstacle set, add it to the to-be-covered adjacent grid set N(o k );

[0033] Step S54: if N(o k ) is not empty, traverse the to-be-covered adjacent grid set N(o k ), calculate the reward function value of each grid by using the reward function Reward(·), add o k to the covered grid set , remove o k from the current to-be-covered area set , and update ok The grid cell with the largest reward function value is selected as the next coverage point. If N(o k If ) is empty, the mobile robot uses function A * Planning from the robot's current position to The nearest grid point in Mid-Manhattan;

[0034] Step S55: If the current set of areas to be covered... If the result is empty, the planning process ends; otherwise, return to step S52 to continue.

[0035] In a preferred embodiment: the reward function Reward(·) in step S54 is designed, and the reward function is based on the distance evaluation value R. g (o j ), Directional evaluation value R s (o j Boundary evaluation value R b (o j The distance evaluation value is composed of the set of adjacent grid cells N(o) to be covered. k Evaluation point o j To cover guide point P g The distance is given by the formula:

[0036]

[0037] Wherein, D(o) j ) for o j With P g Euclidean distance, D min (o k ) for o k All adjacent grid cells j With P g The minimum distance, D max (o k ) for o k All adjacent grid cells j With P g The maximum distance;

[0038] The orientation evaluation value is the evaluation value of the change in coverage orientation of the mobile robot's current position, and the formula is:

[0039]

[0040] Among them, ∠(o k-1 o k o j ) represents the previous coverage point o k-1 Current coverage point o k and o j The radian value of the angle between them;

[0041] The boundary evaluation value is the number of adjacent closed set grids of a to-be-covered point, and the formula is:

[0042]

[0043] wherein, is the maximum value of the number of adjacent closed set grids of all o j , is the number of adjacent closed set grids of the current evaluation point o j ,

[0044] The reward function formula is:

[0045] R(o j )=R g (o j )+ω s (R s (o j ))+ω b (R b (o j )) (6)

[0046] wherein, ω s is a smoothness weight factor, and ω b is a boundary weight factor.

[0047] Compared with the prior art, the present application has the following beneficial effects:

[0048] (1) The present application proposes a balance framework for switching between environment exploration and target coverage collection actions, designs a balance function in combination with the number of feature points in the environment, and combines the target coverage collection task in the local environment exploration process, thereby improving the completion efficiency of the overall task.

[0049] (2) The present application proposes a path planning method for autonomously exploring an environment, designs an information gain function by fusing corner obstacle information, selects a boundary point with the maximum function value as an exploration target point, optimizes exploration path selection, and guides the mobile robot to preferentially explore the local environment to reduce repeated exploration.

[0050] (3) The present application proposes a regional full coverage path planning method, constructs a coverage point reward value model by fusing exploration trajectories, gradually selects the grid with the maximum reward value to form a coverage path, ensures the comprehensive coverage of the task area by the robot, and reduces the repeated coverage rate of the path. BRIEF DESCRIPTION OF DRAWINGS

[0051] Figure Figure 1 is a flowchart of the present application;

[0052] Figure Figure 2 is a schematic diagram of a mobile robot performing an unknown target collection task according to the present application;

[0053] FigureFigure 3 is a schematic diagram of a region to be covered by a mobile robot of the present application;

[0054] attached Figure 4 is a schematic diagram of a mobile robot covering a task region of the present application;

[0055] attached Figure 5 is a schematic diagram of a simulation result of a static scene of the present application;

[0056] attached Figure 6 is a schematic diagram of a simulation result of a static scene of the present application;

[0057] attached Figure 7 is a schematic diagram of a simulation result of a static scene of the present application;

[0058] attached Figure 8 is a schematic diagram of a static real experiment of the present application. DETAILED DESCRIPTION

[0059] The present application will be further described below in conjunction with the accompanying drawings and embodiments.

[0060] It should be noted that the following detailed description is illustrative only, and is intended to provide further description in order to provide a further example of embodiments according to the present application. 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 belongs.

[0061] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of example embodiments according to the present application; as used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise, it will be further understood that the terms "comprises" and / or "comprising," when used in this specification, specify the presence of stated features, steps, operations, devices, components and / or combinations thereof.

[0062] As shown in Figures 1-8 , a motion planning method for a mobile robot facing unknown target collection includes the following steps:

[0063] Step S1, the mobile robot constructs a grid map using data obtained by a sensor carried by the mobile robot;

[0064] Step S2, the mobile robot extracts a boundary region and a passable region according to the constructed grid map;

[0065] Step S3, the mobile robot uses a planning method based on boundary guidance to continuously explore the environment, as shown in Figure 2 (a);

[0066] Step S4, the mobile robot uses a balance framework to judge the exploration degree of the local environment according to a balance function;

[0067] Step S5: If the local environment has been fully explored, the mobile robot uses a planning method guided by the coverage point reward function to perform target collection tasks in the task area, as shown in 2(c). If the local environment has not been fully explored, return to step 2.

[0068] Step S6: After the mobile robot has completed the exploration of the sub-region and the collection of area coverage, if there are unexplored areas or the target has not been collected, it returns to step 2 to continue.

[0069] Step S7: When all unknown environments have been explored and all targets in the mission area have been collected, the unknown target collection mission is complete.

[0070] The specific method for extracting the boundary region in step S2 is as follows: the mobile robot traverses the free and accessible grid cells in the grid map, and at least one adjacent unexplored grid cell is marked as a boundary point and added to the boundary set.

[0071] The boundary-guided planning method in step S3 specifically includes the following steps:

[0072] Step S31: Add the exploration points of the observation environment generated by sampling at interval δ in the boundary set to the candidate exploration point set;

[0073] Step S32: Calculate the function value of each candidate point based on the information gain function and select the point with the largest function value as the next target point for exploration. The formula for the information gain function is:

[0074]

[0075] Among them, S(W obs (v i )) and S(W unexp (v i )) is v i The number of obstacle boundary point grids and unexplored grids within a certain radius r, ω o The weights representing the number of obstacle boundary cells, L(v0,v i () represents the distance from the robot's current position v0 to the candidate viewpoint v. i Manhattan distance, S(M tar (v i )) is v i The number of grid cells in the task area within a certain radius.

[0076] Step S33: The mobile robot passes through A * The algorithm plans a path from the robot's current unknown location to the next exploration target point, such as... Figure 2 As shown in (b);

[0077] Step S34: The robot receives the exploration path planned by the algorithm, and converts the path in the exploration path solution into the required linear velocity and angular velocity, and controls the robot to move to the target position.

[0078] The specific method for constructing the balanced framework in step S4 is as follows:

[0079] The mobile robot calculates the rate of change in the number of boundary points and free grid cells on the map before and after a single exploration, constructing an environmental exploration quantification function. If the function value is less than 0, it indicates that the local area has been completely explored; if the function value is greater than 0, it indicates that the local area has not been completely explored. The function calculation formula is as follows:

[0080]

[0081] Among them, △(S) F ) and △(S) M L(σ) represents the difference between the number of boundary points and the number of free region grid cells before and after a single exploration. e ω represents the length of the exploration path, expressed as the number of grid cells in the exploration path. k As a balancing frame weighting factor, when ω k When the value is 0, the algorithm degenerates into fully exploring the map before covering the task area.

[0082] The planning method based on the coverage point reward function in step S5 specifically includes the following steps:

[0083] Step S51: The mobile robot removes the exploration path grid cells located in the task area and adjacent to obstacles, and the remaining free grid cells in the task area are used as the current set of areas to be covered. like Figure 3 As shown;

[0084] Step S52: Select using the function SetConductor(·) The nearest grid cell in the Manhattan distance to the robot's mid-range position is used as the starting point P of the coverage path. s The next exploration target point is used as the coverage guide point P. g ;

[0085] Step S53: Use the Near(·) function to check o k lie in Adjacent cells in the set, if the cell does not belong to the set of covered cells If it is a set of obstacles, then add it to the set of adjacent grid cells N(o) to be covered. k );

[0086] Step S54: If N(o k If the value is not empty, iterate through the set of adjacent grid cells N(o) to be covered. k), the reward function value of each grid is calculated by using a reward function Reward(·), and o k is added to the covered grid set, and o k is removed from the current to-be-covered region set, and o is updated k The grid with the maximum reward function value is taken as the next covering point, as shown in Figure 4 , if N(o k ) is empty, the mobile robot plans a path from the current position of the robot to the grid point closest to P * in Manhattan distance;

[0087] Step S55: if the current to-be-covered region set is empty, the planning is ended, otherwise, return to step S52 to continue;

[0088] The reward function Reward(·) in step S54 is designed, and the reward function is composed of a distance evaluation value R g (o j ), a direction evaluation value R s (o j ), and a boundary evaluation value R b (o j ). The distance evaluation value is the distance from the evaluation point o k in the to-be-covered neighboring grid set N(o j ) to the covering guide point P g , and the formula is:

[0089]

[0090] wherein, D(o j ) is the Euclidean distance between o j and P g , D min (o k ) is the minimum value of the distances between all neighboring grids o k of o j and P g , and D max (o k ) is the maximum value of the distances between all neighboring grids o k of o j and P g ;

[0091] The direction evaluation value is the evaluation value of changing the covering direction of the mobile robot at the current position, and the formula is:

[0092]

[0093] wherein, ∠(o k-1 o k ​​o j ) is the radian value of the angle between the last coverage point o k-1 , the current coverage point o k and o j ;

[0094] The boundary evaluation value is the number of adjacent closed set grids of the to-be-covered point, and the formula is:

[0095]

[0096] wherein, is the maximum value of the number of adjacent closed set grids in all o j , is the number of adjacent closed set grids of the current evaluation point o j ;

[0097] The reward function formula is:

[0098] R(o j ) = R g (o j ) + ω s (R s (o j )) + ω b (R b (o j )) (6)

[0099] wherein, ω s is a smoothing weight factor, and ω b is a boundary weight factor.

[0100] Embodiment:

[0101] In this embodiment, the complete pseudo code of the algorithm is adopted, and the specific embodiments are as follows:

[0102]

[0103]

[0104]

[0105] wherein, algorithm 1 is the main part of the algorithm, algorithm 2 is the exploration path planning part, and algorithm 3 is the coverage path planning part.

[0106] The embodiments of the present application are described in detail by specific experiments, and the present application proposes a mobile robot motion planning algorithm for unknown target collection tasks, which mainly verifies the effectiveness of the mobile robot in executing the target collection task in the unknown environment through simulation experiments. The specific experimental settings are as follows:

[0107] Simulation experiment:

[0108] The simulation experiment is performed in MATLAB R2018b software.

[0109] (1) Ground Figure 1

[0110] The simulation map size is X = 50 m, Y = 50 m, all obstacles are static obstacles, and the detection range of the mobile robot is 4 m x 4 m. The starting point coordinates are (4, 4). As shown in the simulation map of the ground, the white area in the map is the mobile robot passable area, the black area is the mobile robot impassable area, the cross represents the target to be collected, and the black solid line is the mobile robot trajectory. Figure 5 Figure 1

[0111] (2) Ground Figure 2

[0112] The simulation map size is X = 50 m, Y = 50 m, all obstacles are static obstacles, and the detection range of the mobile robot is 4 m x 4 m. The starting point coordinates are (4, 4). As shown in the simulation map of the ground, the white area in the map is the mobile robot passable area, the black area is the mobile robot impassable area, the cross represents the target to be collected, and the black solid line is the mobile robot trajectory. Figure 6 Figure 2

[0113] (3) Ground Figure 3

[0114] The simulation map size is X = 50 m, Y = 50 m, all obstacles are static obstacles, and the detection range of the mobile robot is 4 m x 4 m. The starting point coordinates are (4, 4). As shown in the simulation map of the ground, the white area in the map is the mobile robot passable area, the black area is the mobile robot impassable area, the cross represents the target to be collected, and the black solid line is the mobile robot trajectory. Figure 7 Figure 3

[0115] Real experiment:

[0116] The static real experiment environment is Ubuntu16.04, ROS Kinetic, and the ice mobile robot using a depth camera and a laser radar, the radar detection range is 0.6 m x 0.6 m, and the Intel Core i7-6700HQ, the memory is 16 GB of Lenovo notebook computer. The scene map size is X = Y = 2.4 m, all obstacles are known static obstacles, and the black and white grid area is the task area, assisted by laser radar and Tag positioning. As shown in the simulation map of the ground, the white area in the map is the mobile robot passable area, the black area is the mobile robot impassable area, the cross represents the target to be collected, and the black solid line is the mobile robot trajectory. Figure 8 ​​​​​​As shown, it is a static real experiment process, the upper side figure is a real shot of the experiment process, and the lower side figure is an Rviz visual interface when the algorithm proposed in the application plans a path during the experiment process, and the white solid line represents a mobile robot trajectory.

[0117] The above merely describes preferred embodiments of the application, and for those skilled in the art, different forms of motion planning algorithms can be designed according to the teaching of the application without creative labor, and any equivalent changes, modifications, replacements and variations made within the scope of the application should fall within the scope of the application.

Claims

1. A motion planning method for a mobile robot collecting towards an unknown target, characterized in that: The method comprises the following steps: Step S1, the mobile robot constructs a grid map using data obtained by a sensor carried by the mobile robot, the grid has three states, namely a passable free grid, an obstacle grid and an unexplored grid; Step S2, the mobile robot extracts a boundary region and a passable region according to the constructed grid map; Step S3, the mobile robot uses a planning method based on boundary guidance to continuously explore the environment; Step S4, the mobile robot uses a balance framework to judge the exploration degree of a local environment according to a balance function; Step S5, if the local environment is completely explored, the mobile robot uses a planning method based on a coverage point reward function to guide a target collection task in a task region, and if the local environment is not completely explored, the method returns to step 2; Step S6, after the mobile robot completes the exploration and coverage collection of a sub-region, if there is an unexplored region or the target is not completely collected, the method returns to step 2 for continuous exploration; Step S7, when all unknown environments are completely explored and all task region targets are completely collected, the unknown target collection task is completed.

2. The motion planning method for mobile robots facing unknown target collection according to claim 1, characterized in that: The specific method for extracting the boundary region in step S2 is that the mobile robot traverses the free passable grid in the grid map, marks a free grid having at least one adjacent unexplored grid as a boundary point, and adds the boundary point to a boundary set.

3. The motion planning method for mobile robots facing unknown target collection according to claim 1, characterized in that: The planning method based on boundary guidance in step S3 comprises the following steps: Step S31: select a boundary set in the boundary set with a distance The sampling generates the exploration points of the observed environment and adds them to the candidate exploration point set. Step S32: Calculate the function value of each candidate point according to the information gain function and select the point with the maximum function value as the next exploration target point, the information gain function The formula is: （1） wherein, with for a certain radius the number of obstacle boundary cell and unexplored cell within a certain radius, a weight representing the number of obstacle boundary cell, is the current position of the robot the Manhattan distance from the candidate viewpoint to the current position of the robot, is the number of task area cell within a certain radius; Step S33: The mobile robot plans a path from the current unknown to the next exploration target point by an algorithm. Step S34, the robot receives an exploration path planned by an algorithm, converts the path in the solution of the exploration path into required linear and angular velocities, and controls the robot to move to a target position.

4. The motion planning method for mobile robots facing unknown target collection according to claim 1, characterized in that: The specific method for constructing the balance framework in step S4 is that the mobile robot calculates the number change rate of boundary points and free grids in the map before and after one exploration, constructs an environment exploration quantification function, and if the function value is less than 0, it indicates that the local region is completely explored, and if the function value is greater than 0, it indicates that the local region is not completely explored, and the function calculation formula is: （2） wherein, and is the difference between the number of boundary points before and after one exploration, is the exploration path length, represented by the number of exploration path grids, is the balancing frame weight factor, when is equal to 0, the algorithm degenerates to a full exploration of the map followed by a coverage of the task area.

5. The motion planning method for mobile robots facing unknown target collection according to claim 1, characterized in that: The planning method based on the coverage point reward function in step S5 comprises the following steps: Step S51: the mobile robot culls the exploration path grids located in the task area and adjacent to the obstacles, and the remaining task area free grids are taken as the current to-be-covered region set ; Step S52: using function Select The Manhattan distance of the nearest grid to the robot position is selected as the starting point of the coverage path , the next exploration target point is selected as the coverage guide point ; Step S53: Using functions examine lie in Adjacent cells in the set, if the cell does not belong to the set of covered cells If it is a set of obstacles, then add it to the set of adjacent grid cells to be covered. ; Step S54: If If not empty, iterate through the set of adjacent grid cells to be covered. Using reward function Calculate the reward function value for each grid cell, and then... Add to the covered raster set ,Will Set up in the current area to be covered Remove from, update The grid cell with the largest reward function value is selected as the next coverage point. If empty, the mobile robot uses the function Planning from the robot's current position to The nearest grid point in Mid-Manhattan; Step S55: If the current set of areas to be covered... If the result is empty, the planning process ends; otherwise, return to step S52 to continue.

6. The motion planning method for a mobile robot facing unknown target collection according to claim 5, characterized in that: The reward function in step S54 The design incorporates a reward function based on distance evaluation values. Directional evaluation value Boundary assessment value Composition, distance evaluation value is the set of adjacent grid cells to be covered. Mid-level assessment points To cover guide point The distance is given by the formula: （3） in, for and European distance, for All adjacent grids and Minimum distance, for All adjacent grids and The maximum distance; The direction evaluation value is an evaluation value of changing the coverage direction of the mobile robot at the current position, and the formula is: （4） wherein, the last coverage point , the current coverage point and the radian value of the angle between them; The boundary evaluation value is the number of adjacent closed set grids of the to-be-covered point, and the formula is: （5） wherein, for all the maximum number of adjacent closed set grids, the current evaluation point adjacent closed set grid number; The reward function formula is: （6） wherein is a smoothness weight factor, is a boundary weight factor.

Images