A ground mobile robot path planning method and system

Abstract

The application discloses a ground mobile robot path planning method and system, and relates to the technical field of intelligent robot autonomous control; the method comprises the following steps: a grid map model of a robot moving environment is constructed; a kinematic model of the robot is established based on the grid map model, and a moving rule is set; a plurality of paths from a starting point to an ending point are initialized to generate an initial path set; a group merging strategy is used to iteratively merge and optimize the paths in the initial path set, and a current global optimal path is selected; the current global optimal path is used as a guide to update and derive paths based on an improved pigeon optimization map and a compass operator to obtain a first-stage optimization set; landmark operators are iteratively optimized to dynamically calculate landmark centers and update path positions based on the value of the iteration to obtain a second-stage optimization set; an optimal path is selected and connected and smoothed to generate an optimal complete path; and the application can efficiently and stably plan a global optimal path in a complex environment.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent robot autonomous control, in particular to a ground mobile robot path planning method and system. BACKGROUND

[0002] In recent years, the rapid development of robot technology has brought significant progress in many fields such as home service, search and rescue, daily entertainment, etc. However, in complex task environments, the path planning of ground mobile robots is still a challenging problem.

[0003] The existing path planning methods for ground mobile robots are mainly based on genetic algorithm, particle swarm optimization algorithm, ant colony optimization algorithm and other technical paths. These methods have exposed many problems in practical application, such as genetic algorithm being too sensitive to parameter setting, improper parameter selection leading to slow convergence speed or falling into local optimal solution, and search efficiency significantly decreasing when dealing with large-scale complex environment, and the optimization result being greatly affected by the quality of the initial population. Particle swarm optimization algorithm is also prone to fall into local optimal solution in complex environment or large-scale path search, its performance is strongly dependent on parameter setting, improper parameter selection will reduce the convergence speed and search accuracy, and when dealing with high-dimensional path planning problems, particle diversity is easily lost quickly, leading to insufficient coverage of search space. Although the ant colony optimization algorithm simulates the behavior of swarm intelligence, it has slow convergence speed, and needs multiple iterations to find a better path in the initial stage due to low pheromone concentration, and when some path pheromone accumulates too quickly, it will attract a large number of individuals to gather, making it difficult for the algorithm to explore other potential better paths. The common core problem of these traditional methods is that it is difficult to accurately find the global optimal path, and it is easily disturbed and falls into local optimum, which cannot meet the demand of efficient path planning of robots in complex dynamic environment.

[0004] Pigeon swarm algorithm has the characteristics of simple principle, few adjustment parameters and easy implementation. Compared with other algorithms, it has obvious advantages such as relatively simple calculation and relatively strong robustness. However, it still has problems of path oscillation and unstable convergence speed when dealing with complex terrain, the reason is that the algorithm is prone to premature convergence when solving, which may fall into local optimum, leading to decreased global search ability, and in the map and compass operator, the motion weight value decays nonlinearly with iteration, which may cause the search range to decrease too fast, affecting the global search ability, so the weight decay strategy needs to be optimized. By introducing an adaptive weight adjustment mechanism, the search step and direction inertia are dynamically adjusted according to the complexity of the terrain.

[0005] In addition, when dealing with high-dimensional or large-scale environments, the search efficiency of existing algorithms decreases significantly, the computational resource consumption is large, and the optimization result is greatly affected by the initial conditions. Once the initial population quality is poor, the algorithm may not be able to find a feasible path within a reasonable time, or even completely fail. These problems seriously limit the practical application effect of mobile robots in dynamic and complex environments, and therefore a more efficient and stable path planning method is needed to improve the autonomous performance and task execution ability of the robot. SUMMARY

[0006] To solve the problems in the prior art, the present application provides a ground mobile robot path planning method and system, which can efficiently and stably plan a globally optimal path in a complex environment, significantly improving the autonomous navigation ability and task execution efficiency of the robot.

[0007] To achieve the above-mentioned purpose, the present application provides a ground mobile robot path planning method, comprising:

[0008] Constructing a grid map model of the robot's moving environment and identifying obstacle regions and passable regions;

[0009] Based on the grid map model, establishing a kinematic model of the robot and setting a movement rule, initializing to generate a plurality of paths from the starting point to the ending point, forming an initial path set;

[0010] Using a population merging strategy to iteratively merge and optimize the paths in the initial path set, and selecting a current globally optimal path;

[0011] Using the current globally optimal path as a guide, based on an improved pigeon optimization map and compass operator, updating and optimizing a new path set derived or reinitialized from the current globally optimal path, to obtain a first-stage optimized path set;

[0012] Based on a pigeon optimization landmark operator, iteratively optimizing the first-stage optimized path set, dynamically calculating a landmark center path according to the path's generation value, and updating the positions of the paths in the first-stage optimized path set, to obtain a second-stage optimized path set;

[0013] Selecting an optimal path from the second-stage optimized path set, and connecting and smoothing the path according to the movement rule, to generate an optimal complete path.

[0014] Optionally, using a population merging strategy to iteratively merge and optimize the paths in the initial path set, and selecting a current globally optimal path, comprising:

[0015] taking the initial path set as a to-be-processed set, and iteratively performing the following steps until one path remains in the to-be-processed set as a current global optimal path:

[0016] randomly grouping the paths in the to-be-processed set two by two;

[0017] dividing the two paths in each group into a better path and a to-be-optimized path; the generation value of the better path is less than or equal to that of the to-be-optimized path;

[0018] based on a preset position updating rule, learning and optimizing the to-be-optimized path guided by the better path to obtain an optimized group path;

[0019] according to a predefined elimination strategy, screening and reserving the two optimized paths in each group;

[0020] forming a new to-be-processed set by using the reserved paths.

[0021] Optionally, the position updating rule is:

[0022] when the better path is not equal to the to-be-optimized path and a learning end condition is not met, updating the position of the to-be-optimized path by using the following formula:

[0023]

[0024] when the better path is equal to the to-be-optimized path and the learning end condition is not met, updating the position of the to-be-optimized path by using the following formula:

[0025]

[0026] wherein, is a random reverse factor, which is any constant in [-1, 1]; is the number of learning times; is a floor function; denotes the to-be-optimized path determined after the i-th learning, denotes the new path after learning; denotes the to-be-optimized path determined after the i-th learning, denotes the new path after learning; denotes the learning factor under the i-th optimization times, denotes the learning factor under the i-th optimization times; wherein, , denotes the better path, denotes the to-be-optimized path, and N is a set learning end condition.

[0027] ​​Optionally, a new to-be-processed set is formed by using the reserved paths, comprising:

[0028] combining the reserved paths in each group to form an intermediate set;

[0029] judging whether the number of paths in the intermediate set is an odd number greater than 1;

[0030] if yes, discarding the path with the maximum value from the intermediate set to form a new to-be-processed set;

[0031] if no, directly taking the intermediate set as the new to-be-processed set.

[0032] Optionally, the expression of the improved pigeon swarm optimization map and compass operator is:

[0033]

[0034]

[0035]

[0036]

[0037] wherein, is the map and compass operator, is the iteration number, represents a better path in the th iteration process; represents the position of a pigeon in the th iteration process, represents the position of a pigeon in the th iteration process; represents the speed of a pigeon in the th iteration process, represents the speed of a pigeon in the th iteration process; is a random number in the interval [0, x], and x is the position; is the motion weight, is the adaptive weight adjustment factor, is the maximum iteration number of the map and compass operator; is a first constant ， is a second constant, is a third constant, is a fourth constant.

[0038] Optionally, the connecting and smoothing according to the movement rule comprises:

[0039] Based on the movement rule, it is judged whether any adjacent path points in the optimal path can be directly reached.

[0040] If not, one or more intermediate points are inserted until all adjacent path points can be directly reached to form a continuous path.

[0041] If the intermediate point is located in the obstacle region, the intermediate point is adjusted to the nearest feasible grid according to the movement rule.

[0042] Optionally, the method further comprises:

[0043] The optimal complete path is visually displayed in the grid map model, and a convergence curve of the corresponding generation value with the iteration number is drawn.

[0044] The application also provides a ground mobile robot path planning system, comprising:

[0045] An environment modeling unit is configured to construct a grid map model of a robot moving environment and identify obstacle regions and passable regions.

[0046] A path initialization unit is configured to establish a kinematic model of the robot based on the grid map model and set a movement rule, initialize to generate a plurality of paths from a starting point to an ending point, and form an initial path set.

[0047] A group merging optimization unit is configured to use a group merging strategy to iteratively merge and optimize the paths in the initial path set, and select a current global optimal path.

[0048] A map and compass optimization unit is configured to use the current global optimal path as a guide, update and optimize a new path set derived or reinitialized from the current global optimal path based on an improved pigeon optimization map and compass operator, and obtain a first-stage optimized path set.

[0049] A landmark optimization unit is configured to iteratively optimize the first-stage optimized path set based on a pigeon optimization landmark operator, dynamically calculate a landmark center path according to the generation value of the path, and update the positions of the paths in the first-stage optimized path set, and obtain a second-stage optimized path set.

[0050] A path processing unit is configured to select an optimal path from the second-stage optimized path set, and perform connecting and smoothing according to the movement rule to generate an optimal complete path.

[0051] Optionally, in the aspect of iteratively merging and optimizing the paths in the initial path set by adopting the group merging strategy, and screening a current global optimal path, the group merging and optimizing unit is specifically used for:

[0052] Taking the initial path set as a to-be-processed set, and iteratively performing the following steps until one path remains in the to-be-processed set as a current global optimal path:

[0053] Randomly grouping the paths in the to-be-processed set two by two;

[0054] Dividing the two paths in each group into a more optimal path and a to-be-optimized path; the generation value of the more optimal path is less than or equal to that of the to-be-optimized path;

[0055] Based on a preset position updating rule, learning and optimizing the to-be-optimized path by taking the more optimal path as a guide to obtain an optimized grouped path;

[0056] According to a predefined elimination strategy, screening and reserving the two optimized paths in each group;

[0057] Using the reserved paths to form a new to-be-processed set.

[0058] According to the specific embodiments provided by the present application, the following technical effects are provided:

[0059] The ground mobile robot path planning method provided by the present application iteratively merges and optimizes the initial path set by adopting the group merging strategy, effectively avoids the problem that the traditional algorithm is prone to falling into local optimization due to directly selecting an initial optimal solution, enhances the global exploration capability of the algorithm in a complex environment, significantly improves the possibility of finding a global optimal path, and, compared with the traditional pigeon group algorithm, reconstructs the compass operator, so that the performance of convergence speed and convergence accuracy is improved. The method takes the current global optimal path as a guide to update the path set derived or reinitialized, and this mechanism effectively coordinates the global exploration and local development process, greatly improves the search efficiency and convergence speed, and reduces the sensitivity to algorithm parameter setting. Further, the landmark operator based on pigeon group optimization is iteratively optimized, the landmark center is dynamically calculated according to the path generation value, and the path position is updated, this strategy enables the algorithm to adaptively focus on the promising search area, enhances the stability and robustness in processing in a complex or high-dimensional environment, and reduces the excessive dependence on the quality of the initial population. Finally, by selecting an optimal path and performing connection and smoothing processing, a continuous and smooth optimal complete path that meets the kinematic constraints of the robot and can safely avoid all obstacles is generated, which significantly improves the reliability, safety and execution efficiency of robot navigation. BRIEF DESCRIPTION OF DRAWINGS

[0060] The above and other objects, features and advantages of the present application will become more apparent from the following detailed description when taken in conjunction with the accompanying drawings in which like reference characters refer to like parts throughout the figures, and wherein:

[0061] Figure 1 Method flowchart of the ground mobile robot path planning method shown in the embodiment of the present application;

[0062] Figure 2 Pigeon swarm optimization method model schematic diagram shown in the embodiment of the present application;

[0063] Figure 3 Ground mobile robot path planning flowchart based on group merging pigeon swarm intelligent optimization shown in the embodiment of the present application;

[0064] Figure 4 Robot path trajectory schematic diagram shown in the embodiment of the present application;

[0065] Figure 5 Optimization effect convergence curve schematic diagram shown in the embodiment of the present application;

[0066] Figure 6 Module structure schematic diagram of the ground mobile robot path planning system shown in the embodiment of the present application. DETAILED DESCRIPTION

[0067] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0068] Please refer to Figure 1 , Figure 1 Method flowchart of the ground mobile robot path planning method.

[0069] The ground mobile robot path planning method comprises:

[0070] S101: Construct a grid map model of the robot moving environment, and identify obstacle regions and passable regions.

[0071] Please refer to Figure 2 , the pigeon swarm optimization method is a new bionic group intelligent optimization method inspired by the mechanism of relying on geomagnetic field, sun and landmarks for navigation in the homing process of pigeon swarm. When the pigeon swarm flies together to find the pigeon nest, each pigeon in the group adjusts the flight direction and flight speed of the next step by referring to the individual closest to the pigeon nest position in the pigeon swarm.

[0072] The implementation flowchart of the path planning method of the ground mobile robot based on colony merging pigeon swarm intelligence optimization is as shown in the figure. Figure 3 In application, in order to describe the moving process and state of the mobile robot, the grid method is adopted to model the map environment. In the process of constructing the grid map model of the robot moving environment, the actual task environment is abstractly expressed by adopting the grid method, and the continuous space is discretized into uniform grid units, each of which represents a region in the environment. The number of rows and columns of the grid map can be determined according to the size and resolution of the actual scene, for example, a square grid structure is set to facilitate computer processing and robot positioning. In the grid map, the obstacle region and the passable region are distinguished by different marks, and the obstacle can be represented by black or dark grid, and the passable region is represented by white or light grid. Each grid has a unique number, and the numbering rule starts from the lower left corner grid, and increases from left to right and from bottom to top until the upper right corner grid is the maximum number.

[0073] In order to facilitate the position calculation and path representation in the path planning process, the row number and column number of each grid can be calculated from its number, and the specific relationship is determined by the integer and remainder operations. This modeling method can clearly and structurally represent the distribution of obstacles and the connectivity of passable regions in the environment, and provide accurate environmental information basis for subsequent path search and optimization. Through the grid map model, the robot can identify the walkable region and the obstacle position, avoid collision, and on this basis, initialize and optimize the path, thereby improving the feasibility and efficiency of path planning.

[0074] For example, a typical activity scene of the robot is built as follows: [0 0 0 0 0 0 0 0 0 0 2 2 2 0 0 0 0 0 0 0; 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0; 0 0 1 0 0 0 1 1 1 0 0 0 0 0 0 1 1 0 0 0; 0 0 0 0 0 0 1 1 1 0 0 0 0 0 0 1 1 0 0 0; 0 0 0 0 0 0 1 1 1 0 0 0 0 0 0 0 0 0 0 0; 0 1 1 1 0 0 0 0 0 2 0 0 0 0 0 0 0 1 1 0; 0 1 1 1 0 0 1 1 1 2 0 0 0 0 0 0 0 1 0 0; 0 0 0 0 0 0 1 1 1 0 1 1 0 2 2 1 0 0; 0 1 1 1 1 0 0 0 0 0 1 1 0 2 2 0 0 0; 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 2 0 0 0; 0 0 0 0 0 0 0 1 1 0 1 1 1 1 0 0 0 0 0 0; 0 0 0 0 0 0 0 1 1 0 0 0 0 0 0 0 0 0 0 0; 0 0 0 0 0 0 0 0 0 0 0 1 1 1 0 1 1 1 1 0; 0 0 0 2 2 2 0 0 0 0 1 0 0 0 0 0 0 0 0 0; 0 0 1 1 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 0; 0 0 1 1 0 0 1 1 1 0 1 0 0 0 0 0 0 0 0 0; 0 0 0 0 0 0 1 1 1 0 0 0 0 0 0 0 0 1 1 0; 0 0 0 0 2 2 2 2 0 0 0 0 0 0 1 0 0 1 1 0; 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0; 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0];

[0095] The grid is set as shown in Figure 4 0 is white, a passable area; 1 is black, an obstacle area; and 2 is gray, a complex area. The grid has a total of 20 rows and 20 columns, numbered 0-399. The starting point is numbered 0, and the end point is numbered 399.

[0096]

[0097] The length of each square is 1, and the relationship between the number of rows, columns, and the number of the grid is calculated according to the above formula (1). In the formula, represents the number of the grid, ∈0,1,2,3,… is the total number of columns of the grid map, is the number of the number of columns of the grid, the number of rows of the grid numbered denotes a floor operation; denotes a modulo operation.

[0098] S102: Based on the grid map model, a kinematic model of the robot is established and a movement rule is set, and a plurality of paths from the starting point to the ending point are generated to form an initial path set.

[0099] wherein the movement rule is that the robot can only move to one of the eight adjacent grids of the current grid each time, and the movement cost is calculated according to the Euclidean distance; and if the movement path shares only a boundary point or an edge with the obstacle grid, it is not considered as a collision, and the path is passable.

[0100] Based on the constructed grid map model, a kinematic model of the robot is established to describe its movement ability and constraints. The movement rule of the robot is set to move in eight adjacent directions each time, including four positive directions (up, down, left, right) and four diagonal directions; wherein moving in a positive direction by one unit distance, the distance moved in a diagonal direction is times the unit distance. This rule simulates the continuous motion of the robot in the actual environment and ensures the rationality and realizability of the path. In order to facilitate subsequent optimization processing, the path planning is not directly searching for a continuous path, but using an abstract representation called "simple path". The simple path is composed of a series of path points, each path point is located on a different row of the map, and the column coordinates are randomly generated within a certain range, thereby forming a polyline from the starting point to the ending point.

[0101] The initialization process aims to generate an initial path set containing a plurality of such simple paths to provide a rich optimization starting point. Each path generates its path points in a random way to ensure population diversity. Then, the fitness value of each path is calculated, which is composed of three parts: path length, obstacle penalty and complex region penalty. The path length is obtained by accumulating the Euclidean distance between adjacent path points; if the path point is located on the obstacle grid, a larger penalty value is applied to significantly increase the total fitness value, thereby guiding the optimization process to avoid obstacle regions; if the path passes through a complex region, the additional consumption of time and energy of the complex region is considered.

[0102] For example, the movement rule of the robot is as follows:

[0103] 1) The robot can only move to the adjacent grid each time, i.e. 4 positive directions and 4 diagonal directions. The positive direction moves a distance of 1, and the diagonal direction moves a distance of . ​

[0104] 2) When the robot passes the edge of an obstacle or two diagonal obstacles, it is considered as a normal pass.

[0105] 3) Simple path: The robot chooses one point in each row of the grid map except the first and last row, and these points form a path with the start and end points of the robot's task. This is convenient for optimization, and in the following steps, the simple path will be generated into a complete path according to the constraints in 1).

[0106] Cost function Set as follows:

[0107]

[0108] Calculate the row number and column number of each point in the robot's simple path, and calculate the length of the polyline formed by the simple path according to formula (2). This length is the first cost function.

[0109] When the robot simple path point is an obstacle, the second cost function increases by 1000.

[0110] When the robot path passes through a complex area, it will significantly increase the time used by the robot to pass through and will lose more energy, and the third cost function is set as:

[0111]

[0112] In the formula, is the time consumption coefficient, is the energy loss coefficient, L is the length of the complex path passed. In Figure 4 the case, = , = 2.

[0113] Total cost function:

[0114]

[0115] Initialize the robot simple path

[0116]

[0117] In the formula, is the total number of columns of the grid map; is a random integer from 1 to c; is the order of the path point in the simple path, which also represents the row number of the path point, 2, 3, … r-1; r is the total number of rows; is the first the number of the point in the grid, the number of the start point, the number of the end point.

[0118] a total of simple paths are generated as a population. Taking a scenario as an example, the number of the population NP = 200, the simple paths are initialized according to formula (5), and the path generation value is calculated according to formulas (1), (2), (3), and (4). Figure 4

[0119] S103: The paths in the initial path set are iteratively merged and optimized by using the population merging strategy, and a current global optimal path is selected.

[0120] In an application, the overall quality of the path can be gradually improved and guided to converge to a high-quality solution through the cooperative competition and iterative refinement mechanism within the path population. Specifically, the initial path set is taken as the to-be-processed set for the first iteration. In each iteration, a systematic merging operation is performed on the to-be-processed set, including randomly grouping the paths in the set, guiding the paths in the group to learn from each other and optimize, and implementing selection and elimination according to the path quality. After each round of operation, all the retained paths are collected to form a new path set. The size of the new set is significantly reduced, and the overall path quality is improved, and the new set is taken as the to-be-processed set for the next iteration.

[0121] The above process is repeatedly executed, and each iteration reduces the size of the path set and continuously improves the overall quality of the path. The iteration process continues until only one path remains in the to-be-processed set, at which time the iteration terminates. The final retained path is the current global optimal path selected by the population merging strategy. The path serves as the guide center for the subsequent pigeon optimization stage and provides a high-quality starting point for subsequent path refinement optimization.

[0122] The population merging strategy effectively avoids the defect that the traditional optimization method is easy to fall into local optimum through multiple iterations of merging and optimization, enhances the global exploration ability of the algorithm in complex environments, and lays a solid foundation for generating a high-quality global path.

[0123] In one embodiment, the above-mentioned iteration merging and optimization of the paths in the initial path set by using the population merging strategy to select a current global optimal path include:

[0124] The initial path set is taken as the to-be-processed set, and the following steps are iteratively executed until only one path remains in the to-be-processed set as the current global optimal path:

[0125] The paths in the to-be-processed set are randomly grouped two by two;

[0126] ​divide the two paths in each group into a better path and a path to be optimized, the generation value of the better path being less than or equal to that of the path to be optimized;

[0127] based on the preset position updating rule, learning and optimizing the path to be optimized with the better path as a guide to obtain optimized paths in the group;

[0128] according to a predefined elimination strategy, screening and reserving the two optimized paths in each group;

[0129] forming a new to-be-processed set by using the reserved paths.

[0130] In the application, firstly, an initial set containing multiple random paths is taken as a to-be-processed set in the first iteration, and an optimization cycle is started. All paths contained in the current to-be-processed set are randomly paired and grouped. For each group, the generation values are compared according to the pre-calculated generation values, and the two paths in the group are divided into a better path and a path to be optimized; wherein the generation value of the better path is strictly less than or equal to that of the path to be optimized. This division ensures that there is a clear learning guide direction in each group. After the division is completed, the optimization learning stage is entered, and based on the preset position updating rule, the node positions of the path to be optimized are updated and adjusted with the spatial position information of the better path as a guide, so as to obtain the two optimized paths in the group.

[0131] Subsequently, according to a predefined elimination strategy, the two optimized paths in each group are screened and reserved. The core principle of the elimination strategy is to reserve the path with a smaller generation value in each group; if the generation values of the two optimized paths are the same, one of them is randomly selected and reserved. After the screening of all groups is completed, the paths reserved by each group are collected and combined to form a new path set, and the size of the set is significantly reduced and taken as a to-be-processed set in the next iteration.

[0132] The grouping, learning, screening and set updating process is executed in a loop. After each iteration is completed, the number of paths in the to-be-processed set is roughly halved, and the average quality of the remaining paths in the set is steadily improved through continuous learning and optimization and strict screening. This iteration mechanism based on group competition and cooperation effectively simulates the natural law of survival of the fittest, and promotes the continuous refinement of the path set. Finally, when the iteration process meets the termination condition, i.e., only one path remains in the to-be-processed set, the path is determined as the current global optimal path. This group merging strategy significantly enhances the global exploration ability of the algorithm in the early stage, effectively avoids the problem of falling into a local optimal solution due to premature convergence, provides a high-quality initial guide path for the subsequent pigeon optimization stage, and lays a solid foundation for generating a global optimal path.

[0133] The position updating rule is:

[0134] When the better path is not equal to the path to be optimized, then the position of the path to be optimized is updated using the following formula:

[0135]

[0136] When the better path is equal to the path to be optimized, then the position of the path to be optimized is updated using the following formula:

[0137]

[0138] wherein, is a random reverse factor, being any constant in [-1, 1]; is the number of learning times; is the set learning end condition; represents the better path; represents the path to be optimized; is the floor function; represents the path to be optimized determined after the (n-1)th learning, represents the new path after learning; (n) represents the learning factor under the nth optimization times.

[0139] Illustratively, the above-mentioned new set to be processed formed by the reserved paths includes:

[0140] The reserved paths in each group are merged to form an intermediate set;

[0141] It is determined whether the number of paths in the intermediate set is an odd number greater than 1;

[0142] If yes, the path with the largest value is discarded from the intermediate set to form a new set to be processed;

[0143] If no, the intermediate set is directly taken as the new set to be processed.

[0144] In the path optimization process, forming a new set to be processed is a key link to ensure final stable convergence. This process starts after the path screening of all groups is completed, and first the paths reserved in each group are collected to form an intermediate set. This set contains all path schemes considered to have potential in the current iteration round.

[0145] Since the size of the initial path set can be odd, or the number fluctuates during the execution of the elimination strategy, the number of paths in the intermediate set is often uncertain. To ensure the stability of the algorithm convergence, the number of paths in the intermediate set can be accurately determined. When it is detected that the number of paths in the set is an odd number greater than 1, the optimization mechanism is started, and the path with the largest generation value is actively discarded from the set. This operation is based on strict path quality evaluation, and by removing the current worst path, the overall quality level of the new set is effectively improved, while ensuring the reasonable reduction of the set size.

[0146] If the number of paths in the intermediate set is even, it indicates that the current size meets the expectations, and it can be directly used as the new set to be processed. When there is only a single path left in the set, it indicates that the algorithm is close to the end of convergence, and the path will be output as the final result.

[0147] By actively discarding the worst path, not only the convergence process is accelerated, but also the quality of the starting point of subsequent iterations is significantly improved. This mechanism complements the grouping and screening steps, and together forms a complete optimization cycle, providing a reliable guarantee for the efficient operation of the path planning algorithm in complex environments, and significantly enhancing the practicality and robustness of the method.

[0148] The following formula is used when updating the set size:

[0149]

[0150] wherein, represents the updated total number of current paths, represents the total number of paths before updating, is the floor function; if is odd, the path with the largest generation value is discarded from the newly formed path set.

[0151] Taking Figure 4 as an example, R takes 0.2, and the initialization speed V is a random number in the interval [0, 1]. Combined with the generation value function, formulas (6), (7), and (8) are used to optimize the initialized simple path until the optimal path is left.

[0152] S104: Based on the improved pigeon optimization map and compass operator, the new path set derived from the current global optimal path or reinitialized is updated and optimized to obtain the first-stage optimized path set.

[0153] In the application, the current global optimal path obtained by the population merging strategy is taken as the core guide, and a newly generated path set is updated and optimized based on the improved pigeon optimization map and compass operator, so that a first-stage optimized path set is obtained. The generation of the new path set includes two approaches: one is to derive the current global optimal path by introducing random disturbance to develop the adjacent area; the other is to generate a batch of paths by the similar method in the initialization stage to maintain population diversity and explore potential new areas.

[0154] In this optimization stage, the map and compass operator dominates the path updating process. The operator simulates the mechanism of pigeons using the geomagnetic field and the position of the sun for navigation. Each path is regarded as a pigeon, and the current position and speed of the pigeon are updated in the iteration process according to specific rules. The core of the update rule is that the speed of each path is attenuated by an exponential factor related to the parameters of the map and compass operator, and a random term is superimposed, which is the product of the vector difference between the current global optimal path and its current position and a random number. Subsequently, the new position of the path is determined by adding the updated speed to the original position.

[0155] The parameters of the operator are not fixed but adopt a dynamic attenuation strategy, such as linear or nonlinear reduction with the increase of iteration number. This design makes the algorithm have strong global exploration ability in the early stage of optimization, which can widely search the solution space; while in the later stage of optimization, it focuses on local fine development and deep mining around the current optimal solution. Through multiple iterations of the above updating operation, each path in the new path set is constantly adjusting itself, learning and evolving under the guidance of the current global optimal path, and recording the optimal value of the entire set after each iteration. Finally, after a predetermined number of iterations, a first-stage optimized path set with significantly improved path quality is obtained.

[0156] Pigeon in pigeon group In the first iteration process, the position and speed are represented as and , respectively, where , is the maximum number of iterations of the map and compass operator. After the first iteration, the speed and position updating rules of the pigeon are shown in equations (9) and (10), and the optimal value of each iteration is recorded. To solve the problem of fast convergence of the algorithm, which leads to falling into local optimum and cannot continue to optimize, an improved pigeon optimization algorithm is proposed, which integrates an adaptive weight adjustment mechanism:

[0157]

[0158] wherein, is a map and compass operator, is the number of iterations, represents the better path in the th iteration process; is a random number in the interval [0, x], x is the position; is a motion weight, is an adaptive weight adjustment factor, is a first constant ， is a second constant, is a third constant, is a fourth constant; After joining, the motion weight slowly decays in a certain range with the increase of the number of iterations, The addition of the parameter makes the algorithm maintain a high exploration ability at the beginning of the search, and gradually reduce the search range as the iteration deepens, improving the stability of the optimization algorithm.

[0159] Taking the Figure 4 scene as an example, , =0.4, =0.5, =2.4, =-1, = 0.8, By using such a dynamic optimization operator, combined with and formulas (9), (10), (11), (12), the original initialization simple path is optimized. When the number of iterations > , the next step is entered.

[0160] S105: Based on the pigeon optimization landmark operator, the first stage optimized path set is iteratively optimized, the landmark center path is dynamically calculated according to the path generation value, and the position of each path in the first stage optimized path set is updated, and the second stage optimized path set is obtained.

[0161] In the application, on the basis of obtaining the first stage optimized path set, the second stage iterative optimization process based on the pigeon optimization landmark operator is entered. This stage simulates the group intelligence behavior of pigeons relying on familiar landmarks to navigate when approaching the destination, aiming to further develop the path set in a deeper level, and further improve the path quality.

[0162] The core of this optimization process is to dynamically calculate and approach a virtual center path called "landmark center". In each iteration, first, the size of the path set is reduced according to the pre-defined rules, for example, halving the population number, eliminating those individuals far from the destination or unfamiliar with the landmark, so as to concentrate the computing resources on the more promising search area. Then, according to the generation value of all paths in the current path set, the landmark center path is dynamically calculated. The calculation is not a simple geometric center, but a weighted center, where the weight of each path to its contribution is inversely proportional to its generation value. This means that the path with smaller generation value and higher quality has greater influence in the composition of the landmark center, thereby guiding the whole population to move in the direction of better solution.

[0163] After the landmark center path is determined, each path in the set updates its own position with it as the learning target. Through multiple iterations of the above population reduction, landmark center calculation, individual path position update and optimal value recording process, the path set is continuously refined and optimized. Finally, after a preset number of iterations, a second-stage optimized path set is obtained through the landmark operator.

[0164] Specifically, in the second-stage landmark operator dominated process, the number of pigeons is reduced by half after each iteration, and pigeons far from the destination and unfamiliar with the landmark are discarded. In the process of gradually approaching the destination, the landmark center position of the pigeon group in the As the first iteration of the pigeon group landmark center position , and assuming that each pigeon can fly directly to the destination. The position update rules of the pigeons in the t iteration are as formula (13), (14) and (15), and the optimal generation value of each iteration is recorded.

[0165]

[0166] Where, is the number of pigeons after iteration, is the number of pigeons before iteration, [ ] is the integer function; the weight coefficient is the position of the pigeon in the iteration, is a random number in the interval [0, x].

[0167] Continue to take Figure 4 the scene as an example, according to formula (13) to halve the population number , and according to formula (14), (15) to continue optimizing the path optimized in the fourth step, until the optimal path is generated or the iteration number > .

[0168] = (0, 27, 50, 72, 92, 111, 130, 149, 168, 187, 206, 226, 246, 267, 288, 309, 330, 351, 373, 399).

[0169] S106: Select the optimal path from the second-stage optimized path set, and perform connection and smoothing processing according to the movement rule to generate an optimal complete path.

[0170] In which, the connection and smoothing processing according to the movement rule includes:

[0171] Based on the movement rule, it is determined whether any adjacent path points in the optimal path can be directly reached;

[0172] If not, one or more intermediate points are inserted until all adjacent path points can be directly reached to form a continuous path;

[0173] If the intermediate point is located in the obstacle region, the intermediate point is adjusted to the nearest feasible grid according to the movement rule.

[0174] In the application, from the second-stage path set optimized by the landmark operator, the path with the smallest value is selected as the optimal solution to be processed. The path has a shorter geometric length and can effectively avoid obstacles under the current abstract representation, but its form is still a simple path composed of a series of discrete path points, which may not fully meet the movement rules in the robot kinematics model, for example, there may be non-adjacent path points between the path points, which cannot be directly used for robot navigation.

[0175] Therefore, the path needs to be connected and smoothed according to the pre-set movement rule to generate an optimal complete path that can be directly executed by the robot. The processing process first determines the position relationship between adjacent path points in the simple path on the grid map. If the two points are not adjacent, one or more intermediate path points are automatically inserted by calculating the arithmetic average of the row and column numbers of the two points, gradually filling the missing part, ensuring that every point on the final path is within the adjacent range allowed by the movement rule. The movement rule, thereby realizing the continuity of the path.

[0176] In the process of inserting intermediate points, the newly inserted point may be located in the obstacle region. At this time, the row coordinate or column coordinate of the intermediate point can be fine-tuned to offset it to the nearest passable grid, thereby ensuring the feasibility of the generated path. This adjustment process changes the overall direction of the original path as little as possible under the premise of avoiding obstacles to maintain its optimality.

[0177] After the above connection and smoothing processing, the originally possible jumping, discrete simple path is converted into a continuous, smooth and strictly following the movement rules of the optimal complete path. The path not only ensures that the robot can move from the starting point to the end point safely and without obstacles, but also usually has the advantages of short path length, reasonable turning, etc. due to its construction based on the optimized high-quality solution, significantly improving the efficiency and motion fluency of the robot in task execution. Finally, the processed optimal complete path is output as the final result of the path planning process.

[0178] In one embodiment, the above method further comprises:

[0179] The optimal complete path is visualized in the grid map model, and a convergence curve of the corresponding generation value with the number of iterations is drawn.

[0180] Referring to Figure 4 , after the completion of the path planning process, in order to further verify the algorithm performance and show the planning results, the final generated optimal complete path can be visualized in the constructed grid map model. This visualization process converts abstract grid data into an intuitive graphical interface, where the row number and column number of the map correspond to the vertical and horizontal coordinates of the visualization interface, respectively. The obstacle area and the passable area are clearly distinguished by different colors or textures of the grid, and the optimal complete path is displayed in a prominent color, such as a red broken line, superimposed on the grid map, thus intuitively showing the complete travel trajectory of the robot from the starting point to the end point safely avoiding all obstacles.

[0181] In addition, as Figure 5 indicated, in order to quantitatively analyze the convergence performance and search efficiency of the optimization algorithm, a convergence curve of the corresponding generation value with the number of iterations can also be drawn. The curve takes the number of iterations as the horizontal axis and the generation value of the global optimal path recorded at the end of each iteration as the vertical axis. By observing the trend of the curve, the quality improvement speed of the solution, whether it is trapped in a local optimum, and the final convergence of the algorithm in the optimization process can be clearly evaluated. This convergence curve provides important data support and intuitive graphical basis for the adjustment of algorithm parameters and the verification of optimization effect.

[0182] Combining the visualized path with the convergence curve for analysis can comprehensively evaluate the overall effect of this path planning. It not only proves the feasibility of the optimal complete path in the given environment, but also confirms the effectiveness and stability of the group merging pigeon intelligence optimization method used in solving this problem through the convergence curve, providing an important reference for the display and subsequent research of algorithm effect.

[0183] Corresponding to the foregoing application function implementation method embodiment, the present application also provides a ground mobile robot path planning system and corresponding embodiments.

[0184] Referring to Figure 6 , Figure 6 Fig. 1 is a schematic diagram of a module structure of a path planning system for a ground mobile robot.

[0185] The path planning system for the ground mobile robot comprises:

[0186] An environment modeling unit 61 is configured to construct a grid map model of a moving environment of the robot and identify an obstacle region and a passable region.

[0187] A path initialization unit 62 is configured to establish a kinematic model of the robot and set a moving rule based on the grid map model, initialize a plurality of paths from a start point to an end point, and form an initial path set.

[0188] A colony merging optimization unit 63 is configured to iteratively merge and optimize the paths in the initial path set by using a colony merging strategy, and filter out a current global optimal path.

[0189] A map and compass optimization unit 64 is configured to update and optimize a new path set derived or reinitialized from the current global optimal path based on an improved pigeon optimization map and compass operator guided by the current global optimal path, and obtain a first-stage optimized path set.

[0190] A landmark optimization unit 65 is configured to iteratively optimize the first-stage optimized path set based on a pigeon optimization landmark operator, dynamically calculate a landmark center path according to a generation value of the path, and update positions of the paths in the first-stage optimized path set, and obtain a second-stage optimized path set.

[0191] A path processing unit 66 is configured to select an optimal path from the second-stage optimized path set, and perform connection and smoothing processing according to the moving rule, and generate an optimal complete path.

[0192] In an embodiment, in the aspect of iteratively merging and optimizing the paths in the initial path set by using the colony merging strategy, and filtering out the current global optimal path, the colony merging optimization unit 63 is specifically configured to:

[0193] Take the initial path set as a to-be-processed set, and iteratively execute the following steps until there is only one path left in the to-be-processed set as the current global optimal path:

[0194] Randomly group the paths in the to-be-processed set two by two;

[0195] Divide the two paths in each group into a better path and a to-be-optimized path; the generation value of the better path is less than or equal to that of the to-be-optimized path.

[0196] Based on the preset position updating rule, the to-be-optimized path is learned and optimized with the more optimal path as a guide to obtain an optimized grouping path;

[0197] According to a predefined elimination strategy, the two optimized paths in each group are screened and reserved;

[0198] The reserved paths are used to form a new to-be-processed set.

[0199] In one embodiment, in the aspect of forming the new to-be-processed set by using the reserved paths, the population merging optimization unit 63 is specifically configured to:

[0200] merge the reserved paths in each group to form an intermediate set;

[0201] determine whether the number of paths in the intermediate set is an odd number greater than 1;

[0202] if yes, discard the path with the maximum value from the intermediate set to form the new to-be-processed set;

[0203] if no, directly take the intermediate set as the new to-be-processed set.

[0204] In one embodiment, in the aspect of performing connection and smoothing processing according to the movement rule, the path processing unit 66 is specifically configured to:

[0205] determine, based on the movement rule, whether any adjacent path points in the optimal path can be directly reached;

[0206] if no, insert one or more intermediate points until all adjacent path points can be directly reached to form a continuous path;

[0207] if the intermediate point is located in an obstacle region, adjust the intermediate point to the nearest feasible grid according to the movement rule.

[0208] In one embodiment, the system further includes:

[0209] a visualization display unit configured to visually display the optimal complete path in the grid map model and draw a convergence curve of the corresponding generation value with respect to the iteration number.

[0210] As to the system in the above embodiments, the specific manner in which each unit module performs operations has been described in detail in the embodiments of the method, and will not be described in detail here.

[0211] Having described various embodiments of the application, it is to be understood that the above description is meant to be illustrative only, and that many modifications and variations of the embodiments described herein are possible. It is therefore to be understood that within the scope of the appended claims and their equivalents, modifications and variations will occur to those skilled in the art upon reading this description. The selection of the terms to be used in the description is intended to best explain the principles of the embodiments, the practical application of the embodiments, or the improvement over technology in the art, as well as to enable other of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. A path planning method for a ground mobile robot, characterized in that, include: Construct a grid map model of the robot's mobile environment and mark obstacle areas and passable areas; Based on the grid map model, a kinematic model of the robot is established and movement rules are set. Multiple paths from the starting point to the ending point are generated to form an initial path set. The movement rules stipulate that the robot can only move to one of the eight adjacent grids of the current grid each time it moves, and the movement cost is calculated based on Euclidean distance. Furthermore, if the movement path and the obstacle grid only share boundary points or edges, it is not considered a collision and the path is passable. A group merging strategy is used to iteratively merge and optimize the paths in the initial path set, and a current globally optimal path is selected. Guided by the current global optimal path, and based on the improved pigeon flock optimization map and compass operator, the set of new paths derived from or re-initialized by the current global optimal path is updated and optimized to obtain the first-stage optimized path set. The landmark operator based on pigeon flock optimization iteratively optimizes the first-stage optimized path set, dynamically calculates the landmark center path based on the path cost, and updates the position of each path in the first-stage optimized path set to obtain the second-stage optimized path set. The optimal path is selected from the second-stage optimized path set, and connected and smoothed according to the movement rules to generate the optimal complete path; A group merging strategy is used to iteratively merge and optimize the paths in the initial path set, selecting a currently globally optimal path, including: The initial set of paths is used as the set to be processed, and the following steps are performed iteratively until only one path remains in the set to be processed as the current globally optimal path: Randomly group the paths in the set to be processed into pairs; In each group, the two paths are divided into a better path and a path to be optimized; the cost of the better path is less than or equal to that of the path to be optimized. Based on the preset location update rules, the path to be optimized is learned and optimized using the better path as a guide to obtain the optimized grouped path; Based on a predefined elimination strategy, the two optimized paths in each group are selected and retained. A new set of processes is formed using the retained paths.

2. The path planning method for a ground mobile robot according to claim 1, characterized in that, The location update rule is as follows: When the better path is not equal to the path to be optimized and the learning termination condition is not met, the position of the path to be optimized is updated using the following formula: ； When the better path is equal to the path to be optimized and the learning termination condition is not met, the position of the path to be optimized is updated using the following formula: ； In the formula, It is a random inverse factor, which is an arbitrary constant in [-1, 1]. Number of times for learning; To round down; Indicates the first The path to be optimized identified after this learning process. express Through the new path learned; Indicates the first Learning factor at each optimization iteration Indicates the first The learning factor under the number of optimization iterations; where... , Indicates a better path. This represents the path to be optimized, and N is the set learning termination condition.

3. The path planning method for a ground mobile robot according to claim 1, characterized in that, A new set of processes is formed using the retained paths, including: The paths retained in each group are merged to form an intermediate set; Determine whether the number of paths in the intermediate set is an odd number greater than 1; If so, the path with the highest cost is discarded from the intermediate set to form a new set to be processed; If not, then the intermediate set will be directly used as the new set to be processed.

4. The path planning method for a ground mobile robot according to claim 1, characterized in that, According to the movement rules, connection and smoothing processes are performed, including: Based on the movement rules, determine whether any adjacent path points in the optimal path are directly reachable; If a path cannot be reached directly, insert one or more intermediate points until all adjacent path points can be reached directly to form a continuous path. If the intermediate point is located in the obstacle area, then the intermediate point is adjusted to the nearest feasible grid according to the movement rules.

5. The path planning method for a ground mobile robot according to claim 1, characterized in that, The method further includes: The optimal complete path is visualized in the grid map model, and the corresponding convergence curve of cost change with the number of iterations is plotted.

6. The path planning method for a ground mobile robot according to claim 1, characterized in that, The expressions for the improved pigeon flock optimization map and compass operator are as follows: ； In the formula, For map and compass operators, For the number of iterations, Indicates the first A better path in the next iteration; Represents pigeon In the Position in the next iteration. Represents pigeon In the Position in the next iteration; Represents pigeon In the Speed ​​during each iteration Represents pigeon In the The speed during the next iteration; The value is a random number within the interval [0, x], where x is the position. For motion weights, For adaptive weight adjustment factor, This represents the maximum number of iterations for the map and compass operators. The first constant ， It is the second constant. It is the third constant. It is the fourth constant.

7. A path planning system for a ground mobile robot, characterized in that, include: The environment modeling unit is used to construct a grid map model of the robot's mobile environment and to identify obstacle areas and passable areas; The path initialization unit is used to establish the robot's kinematic model and set the movement rules based on the grid map model, and initialize and generate multiple paths from the starting point to the ending point to form an initial path set; the movement rules are that the robot can only go to one of the eight adjacent grids of the current grid each time it moves, and the movement cost is calculated according to Euclidean distance; and if the movement path and the obstacle grid only share boundary points or edges, it is not considered a collision, and the path is passable; The group merging optimization unit is used to iteratively merge and optimize the paths in the initial path set using a group merging strategy, and select a current globally optimal path. The map and compass optimization unit is used to update and optimize the new path set derived from or reinitialized from the current global optimal path, based on the improved pigeon flock optimization map and compass operator, guided by the current global optimal path, to obtain the first-stage optimized path set. The landmark optimization unit is used to iteratively optimize the first-stage optimization path set by the landmark operator based on pigeon flock optimization, dynamically calculate the landmark center path according to the cost value of the path, and update the position of each path in the first-stage optimization path set to obtain the second-stage optimization path set. The path processing unit is used to select the optimal path from the second-stage optimized path set, and perform connection and smoothing processing according to the movement rules to generate the optimal complete path; In employing a group merging strategy to iteratively merge and optimize paths in the initial path set, and selecting a currently globally optimal path, the group merging optimization unit is specifically used for: The initial set of paths is used as the set to be processed, and the following steps are performed iteratively until only one path remains in the set to be processed as the current globally optimal path: Randomly group the paths in the set to be processed into pairs; In each group, the two paths are divided into a better path and a path to be optimized; the cost of the better path is less than or equal to that of the path to be optimized. Based on the preset location update rules, the path to be optimized is learned and optimized using the better path as a guide to obtain the optimized grouped path; Based on a predefined elimination strategy, the two optimized paths in each group are selected and retained. A new set of processes is formed using the retained paths.

Images