Substation Multi-Inspection Point Path Planning Method Based on Dynamic Learning Factor

Abstract

The present invention discloses a substation multi-inspection point path planning method based on a dynamic learning factor. Aiming at the situation that substation inspection robots lack global planning for multiple task points, the particle swarm optimization algorithm is selected for task planning. Aiming at the problems that the particle swarm optimization algorithm is prone to falling into local optimum and has a slow algorithm convergence speed, the inertia weight and learning factor of the particle swarm optimization algorithm are improved, an adaptive inertia weight and a dynamic learning factor are designed, and the crossover and mutation operations in the genetic algorithm idea are integrated. Aiming at the problem that the improved particle swarm optimization algorithm only considers the straight-line distance between task points and cannot avoid obstacles, the Lazy Theta* algorithm is integrated with the IPSO algorithm. Through experiments, it is verified that the present invention can avoid the situation of the robot colliding with obstacles while ensuring a relatively short total path length. This method is applied to the global planning of substation inspection robots, which can greatly improve the operation efficiency of the robots, make up for the deficiencies in multi-task point inspection planning, and avoid collisions with static obstacles.

Description

Technical Field

[0001] The present invention belongs to the technical field of robot path planning, and particularly relates to a path planning method for multiple inspection points in a substation based on a dynamic learning factor. Background Art

[0002] The inspection methods of substations are divided into manual inspection and machine inspection. There are many problems with traditional manual inspection, and it has become difficult to achieve the substation operation and maintenance work with continuously increasing real-time requirements and increasing complexity. With the continuous development of technology and the increasing improvement of inspection standards, intelligent inspection robots are expected to replace humans and gradually become the main force for undertaking the daily inspection work of substations. Path planning technology is the basis for efficient inspection of robots. When performing path planning, a reasonable path needs to be planned for the robot to traverse the inspection points and avoid obstacles in the environment.

[0003] Currently, the methods for path planning of substation inspection robots include: invention patent (application number: CN202211495883.9, title: A Method for Path Planning of Substation Inspection Robots), which uses a direction heuristic path planning with multiple nodes in parallel and performs path smoothing optimization, improving the efficiency of path planning for substation inspection robots; invention patent (application number: CN202010693555.4, title: A Method for Path Planning of Substation Inspection Robots), which solves the path planning problem of a robot carrying a thermal imager to autonomously detect the temperature of high-voltage lines under road network constraints and pose constraints by constructing a road network constraint model for the movement of the robot and aiming at minimizing the time to complete the inspection; invention patent (application number: CN201710153238.1, title: A Substation Inspection Robot Path Planning System) uses a path planning system based on reinforcement learning to complete special inspection tasks for key specified equipment under conditions such as special weather, avoiding the track maintenance work of path planning methods such as magnetic tracks. For the path planning problem of inspection robots, researchers have proposed various solutions, including different methods based on graph theory, random sampling, and artificial intelligence. The graph theory path planning algorithm is favored in these solutions due to its ability to ensure the optimality of the path, adapt to complex environments, strong interpretability, and distributed characteristics, and has been widely used in the field of path planning for inspection robots. However, due to the large number of inspection task points in substations, it is difficult to plan the inspection order of multiple task points only relying on the graph theory path planning algorithm, and thus it is impossible to achieve more efficient inspections. Literature research found that the particle swarm optimization (PSO) algorithm has been widely used in solving nonlinear optimization problems, providing a new solution to the problem of weak path planning ability of traditional inspection robots. However, the traditional particle swarm algorithm has problems such as being prone to falling into local optima and having a slow algorithm convergence speed. How to reasonably improve the particle swarm algorithm and improve the path search efficiency requires further research. In addition, after solving the problem of the inspection order of multiple task points, how to plan the path between two task points and avoid obstacles at the same time also needs to be studied. Summary of the Invention

[0004] The object of the present invention is to provide a method for path planning of multiple inspection points in a substation with a dynamic learning factor, which is applicable to the inspection operation of multiple task points of a substation inspection robot and takes into account the efficiency and safety of the inspection, aiming at the problems existing in the above-mentioned prior art.

[0005] To achieve the object of the present invention, the technical solution is as follows: On the one hand, a method for path planning of multiple inspection points in a substation based on a dynamic learning factor is provided, and the method includes the following steps:

[0006] Step 1, construct an environmental grid map and input the xy coordinates of multiple task inspection points in the substation and the Euclidean distance between each coordinate.

[0007] Step 2, use the improved PSO algorithm, i.e., the IPSO algorithm, for multi-inspection point task planning. The improved PSO algorithm can balance the global search ability and local improvement ability of the PSO algorithm, improve the population diversity, and expand the search range of the algorithm.

[0008] Step 3, fuse the Lazy Theta* algorithm. Based on the sequential planning of the substation multi-task point inspection tasks using the IPSO algorithm to output a list of task points, realize the path planning between two task points.

[0009] Further, the IPSO algorithm in Step 2 specifically includes:

[0010] Adopt a non-linear dynamic inertia weight coefficient to balance the global search ability and local improvement ability of the PSO algorithm;

[0011] Dynamically adjust the learning factors in the PSO algorithm. Construct the learning factor c1 as a monotonically decreasing function, and construct the learning factor c2 as a monotonically increasing function;

[0012] Introduce the crossover and mutation operations of the genetic algorithm into the PSO algorithm.

[0013] Further, the constant inertia weight factor ω in the PSO algorithm is replaced by a non-linear dynamic inertia weight coefficient, and the formula is as follows:

[0014]

[0015] In the formula, f represents the real-time objective function value of the particle, f avg and f min respectively represent the average value of all current particles and the minimum objective value, ω max 、ω min are fixed values, which are the maximum and minimum values of the set constant inertia weight factor respectively.

[0016] Further, for the non-linear dynamic inertia weight coefficient, when the particle objective values are dispersed, reduce the inertia weight; when the particle objective values are consistent, increase the inertia weight.

[0017] Further, the formulas for the learning factors c1 and c2 are:

[0018]

[0019]

[0020] Among them, k is a constant coefficient used to set the maximum values of c1 and c2; t is the current iteration number, and T amx is the maximum iteration number of the particle swarm.

[0021] Further, k takes values from 1 to 5.

[0022] Further, introducing the crossover and mutation operations of the genetic algorithm into the PSO algorithm specifically includes:

[0023] Through an embedded hybrid method, embed the crossover and mutation operations in the genetic algorithm into the IPSO algorithm;

[0024] (1) The steps of the crossover operation are as follows:

[0025] Randomly select two crossover points to determine the crossover interval;

[0026] Extract the segments within the crossover interval;

[0027] Delete the gene segments within the crossover interval from the first parent;

[0028] Check and remove duplicate elements within the crossover interval;

[0029] Insert the segment after removing duplicate elements into the corresponding position of the first parent to form the first offspring;

[0030] (2) The steps of the mutation operation are as follows:

[0031] Randomly generate two different indexes, which correspond to the positions in the chromosome excluding the start and end points;

[0032] Perform a transposition operation on these two indexes, that is, exchange the gene values at two positions in the chromosome;

[0033] Randomly generate two different indexes again for a secondary transposition operation to exchange the gene values at two other positions in the chromosome.

[0034] Further, the specific process of step 3 includes:

[0035] Initialize the parameters of the IPSO algorithm and the Lazy Theta* algorithm, and load the task point coordinates and distances;

[0036] Use the IPSO algorithm to construct a multi-task point inspection solution, sequentially load the multi-task point inspection task list, set the start and end positions of the robot, and use the Lazy Theta* algorithm for path planning;

[0037] By calculating the path cost, determine whether the error accuracy or the number of iterations is reached. If so, obtain the optimal path. Otherwise, use the hybrid algorithm combining IPSO and Lazy Theta* to continue optimizing the existing path until the optimal path is found. Finally, output the optimal path and draw the route.

[0038] On the other hand, a substation multi-inspection point path planning system with a dynamic learning factor is provided, and the system includes:

[0039] A first module for constructing an environmental grid map and inputting the xy coordinates of multi-task inspection points in the substation and the Euclidean distances between each coordinate.

[0040] A second module for using an improved PSO algorithm, namely the IPSO algorithm, to perform multi-inspection point task planning. The improved PSO algorithm can balance the global search ability and local improvement ability of the PSO algorithm, improve population diversity, and expand the search range of the algorithm.

[0041] A third module for fusing the Lazy Theta* algorithm. Based on the sequential planning of the multi-task point inspection tasks in the substation using the IPSO algorithm to output a list of task points, the path planning between two task points is realized.

[0042] Compared with the prior art, the remarkable advantages of the present invention are:

[0043] (1) The present invention provides a substation multi-inspection point path planning method based on a dynamic learning factor, which analyzes the global path planning problem of substation inspection robots by dividing it into two cases: single-task point inspection and multi-task point inspection.

[0044] (2) Aiming at the problem that the particle swarm algorithm is prone to fall into local optimum and the algorithm convergence speed is slow when performing multi-task point path planning, by improving the inertia weight and learning factor of the particle swarm algorithm and fusing the genetic algorithm idea, the performance of the algorithm for multi-task point path planning is improved, jumping out of the local optimum, and the algorithm convergence speed is also significantly accelerated.

[0045] (3) Aiming at the problem that the improved particle swarm algorithm only considers the straight-line distance between task points and cannot avoid obstacles, by fusing the Lazy Theta* algorithm with the IPSO algorithm, while ensuring that the total path length is shorter, the situation of the robot colliding with obstacles is avoided.

[0046] The present invention will be further described in detail below with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 It is a flowchart of the fusion of the IPSO-Lazy Theta* algorithm based on a dynamic learning factor in an embodiment.

[0048] Figure 2 It is a schematic diagram of the particle swarm algorithm (bird foraging behavior) in an embodiment.

[0049] Figure 3 It is a flowchart of the IPSO algorithm in an embodiment.

[0050] Figure 4 Comparison chart of path planning effects between the Lazy Theta* algorithm and other algorithms selected in an embodiment, where Figure 4 (a) in it is a schematic diagram of path planning by the Lazy Theta* algorithm, Figure 4 (b) in it is a schematic diagram of path planning by the A* algorithm, Figure 4 (c) in it is a schematic diagram of path planning by the greedy search algorithm, Figure 4 (d) in it is a schematic diagram of path planning by the Jump A* algorithm, Figure 4 (e) in it is a schematic diagram of path planning by the ant colony algorithm, Figure 4 (f) in it is a schematic diagram of path planning by the fast search random number algorithm

[0051] Figure 5 Instance diagram of multi-task point distribution in an embodiment, where Figure 5 (a) in it is a schematic diagram of the distribution of 30 task points, Figure 5 (b) in it is a schematic diagram of the distribution of 50 task points, Figure 5 (c) in it is a schematic diagram of the distribution of 75 task points.

[0052] Figure 6 Comparison chart of multi-task point path planning before and after improving the algorithm in an embodiment. Among them Figure 6 (a) in it is a schematic diagram of path planning for 30 task points before algorithm improvement, Figure 6 (b) in it is a schematic diagram of path planning for 30 task points after algorithm improvement, Figure 6 (c) in it is a schematic diagram of path planning for 50 task points before algorithm improvement, Figure 6 (d) in it is a schematic diagram of path planning for 50 task points after algorithm improvement, Figure 6 (e) in it is a schematic diagram of path planning for 75 task points before algorithm improvement, Figure 6 (f) in it is a schematic diagram of path planning for 75 task points after algorithm improvement.

[0053] Figure 7 Comparison chart of path iteration before and after improving the particle swarm optimization algorithm in an embodiment, where Figure 7 (a) in it is a schematic diagram of planning iteration for 30 task points before and after improving the algorithm, Figure 7 (b) in it is a schematic diagram of planning iteration for 50 task points before and after improving the algorithm, Figure 7 (c) in it is a schematic diagram of planning iteration for 75 task points before and after improving the algorithm.

[0054] Figure 8 Schematic diagram after adding static obstacles to the multi-task point instance in an embodiment.

[0055] Figure 9 Schematic diagram of path planning of the improved fusion algorithm under the background of static obstacles in one embodiment.

[0056] Figure 10 Schematic diagram of path planning of the unimproved fusion algorithm under the background of static obstacles in one embodiment.

[0057] Figure 11 Schematic diagram of grid search of the fusion algorithm under the background of obstacles in one embodiment. Specific implementation manner

[0058] In order to make the objectives, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0059] It should be noted that if there are directional indications (such as up, down, left, right, front, back...) involved in the embodiments of the present invention, the directional indications are only used to explain the relative positional relationship and movement conditions between components in a specific posture (as shown in the drawings). If the specific posture changes, the directional indications will also change accordingly.

[0060] In addition, if there are descriptions such as "first" and "second" involved in the embodiments of the present invention, the descriptions of "first", "second", etc. are only for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include at least one of the features. In addition, the technical solutions between various embodiments can be combined with each other, but it must be based on the fact that those skilled in the art can implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such a combination of technical solutions does not exist and is not within the protection scope required by the present invention.

[0061] In one embodiment, in combination with Figure 1 , a substation multi-inspection point path planning method based on a dynamic learning factor is provided, including the following steps:

[0062] Step 1, construct an environmental grid map and input the xy coordinates of the multi-task inspection points of the substation and the Euclidean distances between the respective coordinates;

[0063] Step 2, use the improved PSO algorithm, that is, the IPSO algorithm, to perform multi-inspection point task planning. The improved PSO algorithm can balance the global search ability and local improvement ability of the PSO algorithm, improve the population diversity and expand the search range of the algorithm;

[0064] Step 3: Integrate the Lazy Theta* algorithm. Based on the sequential planning of the inspection tasks at multiple task points in the substation by the IPSO algorithm to output a list of task points, implement the path planning between every two task points.

[0065] Furthermore, in one embodiment, the IPSO algorithm described in step 2 specifically includes:

[0066] (1) Adopt a non-linear dynamic inertia weight coefficient to balance the global search ability and local improvement ability of the PSO algorithm. Specifically:

[0067] The mathematical description of the PSO algorithm is as follows: The algorithm first establishes a D-dimensional space and creates n particles in the space. These particles form a complete particle swarm, which is represented as: X = [X1, X2, …, X n , where the i-th particle is represented as X i = [x i1 , x i2 , …, x iD T , and the velocity of the i-th particle is represented as V i = [V i1 , V i2 , …, V iD T . Substituting the particles into the problem to be solved can obtain the fitness value, which can represent the quality of the particle position. During the update process, the current fitness value of the particle is changed in real time. The best of the particle itself is the individual extreme value P i = [P i1 , P i2 , …, P iD T . The best among all particles is the global extreme value P g = [P g1 , P g2 , …, P gD T . The update formula is as follows:

[0068]

[0069]

[0070] In the formula, ω represents the inertia weight; the value range of d is: d = 1, 2, …, D, where D is the dimension of the space; the value range of i is i = 1, 2, …, n; k represents the current iteration number of the algorithm; c1 and c2 are constants greater than zero, called learning factors, and the selection of learning factors will have different relatively excellent values according to different optimization problems and actual engineering problems; r1 and r2 are random numbers distributed between [0, 1]. ​​​​

[0071] When the value of the inertia weight factor ω is large, it is beneficial to the search in the global range. When the value of ω is small, it is beneficial to accelerate the convergence speed of the algorithm and the optimization of local extrema.

[0072] In order to balance the global search ability and local improvement ability of the PSO algorithm, the present invention adopts a non-linear dynamic inertia weight coefficient formula, and its expression is:

[0073]

[0074] In the formula, f represents the real-time objective function value of the particle, f avg and f min respectively represent the average value of all current particles and the minimum objective value.

[0075] It can be seen from the above formula that the inertia weight will change with the change of the particle objective function value. When the particle objective values are dispersed, the inertia weight is reduced; when the particle objective values are consistent, the inertia weight is increased.

[0076] (2) Dynamically adjust the learning factors in the PSO algorithm, construct the learning factor c1 as a monotonically decreasing function, and construct the learning factor c2 as a monotonically increasing function, so that the population can quickly search for the optimal value in a short time at the initial stage of particle swarm evolution, and can quickly converge to the optimal solution at the later stage of evolution. Specifically:

[0077] In the standard particle swarm algorithm, the learning factors c1 and c2 generally take fixed values. c1 represents the particle's thinking about itself, that is, the part where the particle learns from itself; c2 represents the particle's sociality, that is, the characteristic that the particle learns from the global optimal particle.

[0078] During the process of the particle swarm algorithm, at the beginning of the algorithm, it should search widely in space to increase particle diversity, and in the later stage, it should pay attention to the convergence of the algorithm. This is reflected in the algorithm that the weights of the learning factors c1 and c2 should change with the progress of the algorithm instead of being fixed.

[0079] Therefore, in order to prevent the particles from quickly gathering around the local optimal solution at the initial stage of evolution and enable the particles to search in a large range in the global domain, let c1 take a larger value and c2 take a smaller value. In the later stage of the search, in order to facilitate the particles to quickly and accurately converge to the global optimal solution and improve the convergence speed and accuracy of the algorithm, set c1 to take a smaller value and c2 to take a larger value.

[0080] Therefore, the present invention constructs c1 as a monotonically decreasing function and constructs c2 as a monotonically increasing function, and their expressions are as follows:

[0081]

[0082]

[0083] where k is a constant coefficient used to set the maximum values of c1 and c2, t is the current iteration number, and T max is the maximum iteration number of the particle swarm.

[0084] Through the above formula, the values of the learning factors can be dynamically adjusted, enabling the population to quickly search for the optimal value in a short time at the initial stage of evolution and quickly converge to the optimal solution at the later stage of evolution.

[0085] (3) Through an embedded hybrid method, the crossover and mutation operations in the genetic algorithm are embedded into the IPSO algorithm to improve the population diversity and expand the search range of the algorithm, thereby obtaining the global optimal solution. Specifically:

[0086] The steps of the crossover operation are as follows:

[0087] Randomly select two crossover points to determine the crossover interval;

[0088] Extract the segments within the crossover interval;

[0089] Delete the gene segments within the crossover interval from the first parent;

[0090] Check and remove the duplicate elements within the crossover interval;

[0091] Insert the segment after removing the duplicate elements into the corresponding position of the first parent to form the first offspring;

[0092] The steps of the mutation operation are as follows:

[0093] Randomly generate two different indices, which correspond to the positions in the chromosome (task point list) excluding the start and end points;

[0094] Perform a transposition operation on these two indices, that is, exchange the gene values at the two positions in the chromosome;

[0095] Randomly generate two different indices again and perform a secondary transposition operation to exchange the gene values at another two positions in the chromosome.

[0096] Furthermore, in one of the embodiments, in step 3, by fusing the improved PSO algorithm with the Lazy Theta* algorithm, the obstacle problem on the patrol path of the patrol robot is solved. The specific fusion algorithm process is as follows:

[0097] First, initialize the parameters of the IPSO algorithm and the Lazy Theta* algorithm, and load the task point coordinates and distances;

[0098] Construct a multi-task point inspection solution using the IPSO algorithm, sequentially load the multi-task point inspection task list, set the start and end positions of the robot, and use the Lazy Theta* algorithm for path planning;

[0099] By calculating the path cost, determine whether the error accuracy or the number of iterations is reached. If so, obtain the optimal path. Otherwise, continue to optimize the existing path using the hybrid algorithm combining IPSO and Lazy Theta* until the optimal path is found. Finally, output the optimal path and draw the route.

[0100] In one embodiment, a substation multi-inspection point path planning system based on a dynamic learning factor is provided. The system includes:

[0101] The first module is used to construct an environmental grid map and input the xy coordinates of the substation multi-task inspection points and the Euclidean distances between their respective coordinates;

[0102] The second module is used to perform multi-inspection point task planning using the improved PSO algorithm, i.e., the IPSO algorithm. The improved PSO algorithm can balance the global search ability and local improvement ability of the PSO algorithm, improve population diversity, and expand the search range of the algorithm;

[0103] The third module is used to integrate the Lazy Theta* algorithm and implement path planning between two task points on the basis of sequentially planning and outputting a task point list for the substation multi-task point inspection task based on the IPSO algorithm.

[0104] For the specific limitations of the substation multi-inspection point path planning system based on the dynamic learning factor, reference can be made to the limitations of the substation multi-inspection point path planning method based on the dynamic learning factor in the above text, which will not be elaborated here. Each module in the above-mentioned substation multi-inspection point path planning system based on the dynamic learning factor can be implemented in whole or in part through software, hardware, and their combinations. The above-mentioned modules can be embedded in the processor of the computer device in hardware form or independent of it, or stored in the memory of the computer device in software form, so that the processor can call and execute the operations corresponding to the above-mentioned modules.

[0105] Next, the present invention will be described in detail with reference to the accompanying drawings.

[0106] As Figure 1 shown, a substation multi-inspection point path planning method based on a dynamic learning factor is provided, including the following steps:

[0107] Step 1: Input the coordinates of the task points to be planned and the Euclidean distances between every two task points, so as to mark the positions of the task points on the grid map. The coordinate unit is m. Each inspection point is marked in the input order, with the starting point marked as 0, and the other target points are incremented by 1 in sequence. The schematic diagram of the task point marking is as shown in Figure 5 .

[0108] Step 2: Use the IPSO algorithm to construct a multi-task point inspection solution. The particle swarm algorithm has the advantages of rapid search, easy implementation, simple parameter setting, etc., and is suitable for solving multi-task point planning problems. However, it has problems such as being prone to falling into local optima and poor algorithm convergence speed. The schematic diagram of its algorithm is as shown in Figure 2 . In view of these shortcomings, the present invention has made improvements. First, an adaptive inertia weight is constructed.

[0109] The inertia weight factor ω affects the performance of the PSO algorithm. When the value of ω is large, it is beneficial to the search in the global range. When the value of ω is small, it is beneficial to accelerating the convergence speed of the algorithm and the optimization of local extrema. In order to balance the global search ability and local improvement ability of the PSO algorithm, the present invention adopts a non-linear dynamic inertia weight coefficient formula, and its expression is:

[0110]

[0111] In the formula, f represents the real-time objective function value of the particle, f avg and f min respectively represent the average value of all current particles and the minimum objective value. It can be seen from the above formula that the inertia weight will change with the change of the particle objective function value. When the particle objective values are scattered, that is, f ≤ f avg , the inertia weight is reduced; when the particle objective values are consistent, that is, f > f avg , the inertia weight is increased. ω max is set to 1, and ω min is set to 0.3.

[0112] Step 3: Optimize the PSO algorithm by dynamically adjusting the learning factors in the PSO algorithm. Construct the learning factor c1 as a monotonically decreasing function and construct c2 as a monotonically increasing function, so that the population can quickly search for the optimal value in a short time in the initial stage of the particle swarm evolution, and can quickly converge to the optimal solution in the later stage of evolution.

[0113] In the standard particle swarm algorithm, the learning factors c1 and c2 generally take fixed values. c1 represents the particle's thinking about itself, that is, the part where the particle learns from itself; c2 represents the sociality of the particle, that is, the characteristic that the particle learns from the globally optimal particle.

[0114] During the process of the particle swarm algorithm, at the beginning of the algorithm, it should search widely in the space to increase particle diversity. In the later stage, it should focus on the convergence of the algorithm. In the algorithm, the learning factors c1 and c2 should have changing weights as the algorithm progresses instead of being fixed.

[0115] Therefore, to prevent particles from quickly aggregating around the local optimal solution at the initial stage of evolution and enable particles to search widely in the global domain, let c1 take a larger value and c2 take a smaller value. In the later stage of the search, to facilitate the particles to converge quickly and accurately to the global optimal solution and improve the convergence speed and accuracy of the algorithm, set c1 to take a smaller value and c2 to take a larger value. Therefore, in this description, c1 is constructed as a monotonically decreasing function, and c2 is constructed as a monotonically increasing function. Their expressions are as follows:

[0116]

[0117]

[0118] k is an adjustment coefficient used to set the maximum values of c1 and c2. In the present invention, the preferred value is 2. t is the current iteration number, and T max is the maximum number of iterations of the particle swarm. Through the above formula, the values of the learning factors can be dynamically adjusted, enabling the population to quickly search for the optimal value in a short time at the initial stage of evolution and quickly converge to the optimal solution at the later stage of evolution.

[0119] Step 4, as Figure 3 shown, embed the crossover and mutation operations in the genetic algorithm into the improved particle swarm algorithm to improve population diversity and expand the search range of the algorithm so as to obtain the global optimal solution. Specifically:

[0120] The steps of the crossover operation are as follows:

[0121] (1) Randomly select two crossover points to determine the crossover interval.

[0122] (2) Extract the segments in the crossover interval and extract the parts of the other parent within the crossover interval for subsequent insertion.

[0123] (3) Delete the gene segments within the crossover interval from the first parent.

[0124] (4) Check and remove duplicate elements within the crossover interval.

[0125] (5) Insert the segments after removing duplicate elements into the corresponding positions of the first parent to form the first offspring.

[0126] The path of the offspring after crossover is saved in the specified variable a. This process aims to generate offspring with better adaptability by fusing the information of the two parents.

[0127] The steps of the mutation operation are as follows:

[0128] (1) Randomly generate two different indices, which correspond to the positions in the chromosome (task point list) excluding the start and end points. (RandIndex = randperm(length(route)-2)+1)

[0129] (2) Perform a transposition operation on these two indices, that is, exchange the gene values at the two positions in the chromosome. (route(RandIndex(2:-1:1)) = route(RandIndex(1:2)))

[0130] (3) Randomly generate two different indices again, perform a secondary transposition operation, and exchange the gene values at another two positions in the chromosome. (route(RandIndex(4:-1:3)) = route(RandIndex(3:4)))

[0131] The mutation operation can improve the diversity of solutions and is conducive to the algorithm jumping out of the local optimal solution.

[0132] Through steps 1, 2, 3, and 4, finally output the optimal solution for multi-task point inspection and store it as a task list in the form of a table. The table records the marked serial numbers of the inspection points.

[0133] Step 5, integrate the Lazy Theta* algorithm. On the basis of sequentially planning the task point list for the substation multi-task point inspection task based on IPSO, realize the path planning between two task points. Specifically:

[0134] Such as Figure 4 , compare the results of single-task point path planning between the Lazy Theta* algorithm and other algorithms. The map is a grid map. When building the map, the obstacle area is valued at 2 and represented by a black square, the passable area is set to 1 and represented by a white square, the blue square represents the starting point, the red square represents the target point, and the actual planned route is represented by a red dotted line. Compared with these mainstream algorithms, the path planned by the Lazy Theta* algorithm adopted by the present invention is the shortest, and the path is relatively smooth with fewer path turning points, which is very suitable for the working requirements of the substation inspection robot.

[0135] Then sequentially load the multi-task point inspection task list, set the start and end positions of the robot, use the Lazy Theta* algorithm for path planning, calculate the path cost, and judge whether the error accuracy or the number of iterations is reached. If so, find the best path, otherwise use the hybrid algorithm to continue to optimize the existing path until the best path is found. Finally, output the best path and draw the route. The route drawing result is asFigure 9 。

[0136] Implementation verification:

[0137] The above-mentioned substation multi-inspection point path planning method based on dynamic learning factor was placed under the following conditions for simulation experiments. (1) Simulation platform: MATLAB 2020b; (2) Hardware environment: The processor is AMD Ryzen 7 5800H with Radeon Graphics, and the memory size is 16 GB; (3) Example grid map size: 100 * 100; (4) Number of inspection points: 30, 50, 75.

[0138] The specific experiments are as follows:

[0139] Referring to the schematic diagram of the multi-task point distribution as shown in Figure 5 , the improved particle swarm optimization algorithm in the present invention was used to perform task planning for 30, 50, and 75 numbers of task points. The experimental results are shown in Table 1.

[0140] Table 1 IPSO Global Task Planning List

[0141]

[0142]

[0143] The route drawing results before and after the improvement of the PSO algorithm are as shown in Figure 6 . It can be seen from the figure that the path lengths planned by the improved particle swarm optimization algorithm are shortened by 22.1%, 23.3%, and 31.2% respectively, and the path planning effect is significantly improved.

[0144] Figure 7 For the path iteration comparison diagram of the particle swarm optimization algorithm before and after improvement for different numbers of task points, it can be found that the improved particle swarm optimization algorithm jumps out of the local optimum, and the algorithm convergence speed is also significantly accelerated.

[0145] Taking 30 task points as an example, the schematic diagram after adding static obstacles is as shown in Figure 8 . Figure 9 This is the schematic diagram of the path planning of the improved fusion algorithm under the background of static obstacles in the embodiment of the present invention. Figure 10 This is the schematic diagram of the path planning of the unimproved fusion algorithm under the background of static obstacles in the embodiment of the present invention. Figure 11 This is the schematic diagram of the grid search of the fusion algorithm under the background of obstacles in the embodiment of the present invention. It can be seen from the figure that the substation multi-inspection point path planning based on dynamic learning factor in the present invention can not only achieve shorter path planning under multiple task points, but also effectively avoid static obstacles. This will greatly improve the inspection efficiency of substation inspection robots and ensure their safe operation at the same time, and has good application prospects.

[0146] When this method is applied to the global planning of substation inspection robots, it can greatly improve the operation efficiency of the robots, make up for the deficiencies in the inspection planning of multiple task points, and avoid collisions with static obstacles.

[0147] Although the specific implementation manners of the present invention have been described above, those skilled in the art should understand that these specific implementation manners are only illustrative. Without departing from the principles and essence of the present invention, those skilled in the art can make various omissions, substitutions, and changes to the details of the above methods and systems. For example, combining the above method steps so as to perform substantially the same function in a substantially the same manner to achieve substantially the same result falls within the scope of the present invention. Therefore, the scope of the present invention is only defined by the appended claims.

Claims

1. A substation multi-inspection point path planning method based on dynamic learning factors, characterized in that: The method comprises the following steps: Step 1: construct an environmental grid map and input the xy coordinates of the substation multi-task inspection points and the Euclidean distances between their coordinates; Step 2: Using the improved PSO algorithm, namely the IPSO algorithm, to plan the multi-inspection point tasks. The improved PSO algorithm can balance the global search capability and local improvement capability of the PSO algorithm, improve the population diversity and expand the search scope of the algorithm. Step 3: Integrate the Lazy Theta* algorithm to sequentially plan the inspection tasks of multiple task points of the substation based on the IPSO algorithm and output the task point list to achieve path planning between two task points. The specific process includes: Initialize the IPSO algorithm and Lazy Theta* algorithm parameters, and load the task point coordinates and distances; Use the IPSO algorithm to build a multi-task point inspection solution, load the multi-task point inspection task list in sequence, set the robot's start and end positions, and use the Lazy Theta* algorithm for path planning; By calculating the path cost, it is determined whether the error accuracy or the number of iterations has been reached. If so, the optimal path is obtained. Otherwise, a hybrid algorithm combining IPSO and Lazy Theta* is used to continue optimizing the existing path until the optimal path is found. Finally, the optimal path is output and the route is drawn.

2. The substation multi-inspection point path planning method based on dynamic learning factor according to claim 1 is characterized in that: The IPSO algorithm described in step 2 specifically includes: A nonlinear dynamic inertia weight coefficient is used to balance the global search capability and local improvement capability of the PSO algorithm; Dynamically adjust the learning factor in the PSO algorithm, construct the learning factor c1 as a monotonically decreasing function, and construct the learning factor c2 as a monotonically increasing function; The crossover and mutation operations of genetic algorithm are introduced into the PSO algorithm.

3. The substation multi-inspection point path planning method based on dynamic learning factor according to claim 2 is characterized in that: The constant inertia weight factor ω in the PSO algorithm is replaced by a nonlinear dynamic inertia weight coefficient, as shown in the following formula: In the formula, f represents the real-time objective function value of the particle, f avg and f min Respectively represent the average value and minimum target value of all particles at present, ω max ,ω min are fixed values, which are the maximum and minimum values ​​of the set constant inertia weight factor respectively.

4. The substation multi-inspection point path planning method based on dynamic learning factor according to claim 2 is characterized in that: The nonlinear dynamic inertia weight coefficient reduces the inertia weight when the particle target values ​​are dispersed, and increases the inertia weight when the particle target values ​​are consistent.

5. The substation multi-inspection point path planning method based on dynamic learning factor according to claim 2 is characterized in that: The formulas of the learning factors c1 and c2 are: Among them, k is a constant coefficient used to set the maximum value of c1 and c2; t is the current number of iterations, T max is the maximum number of iterations of the particle swarm.

6. The substation multi-inspection point path planning method based on dynamic learning factor according to claim 5 is characterized in that: The value of k ranges from 1 to 5.

7. The substation multi-inspection point path planning method based on dynamic learning factor according to claim 2 is characterized in that: The crossover and mutation operations of the genetic algorithm are introduced into the PSO algorithm, specifically including: Through the embedded hybrid method, the crossover and mutation operations in the genetic algorithm are embedded into the IPSO algorithm; (1) The steps of crossover operation are as follows: Randomly select two intersection points and determine the intersection interval; In the intersection interval, extract the fragments therein; Delete the gene fragment within the crossover interval from the first parent; Check and remove duplicate elements in the intersection interval; Insert the fragment after removing duplicate elements into the corresponding position of the first parent generation to form the first child generation; (2) The steps of mutation operation are as follows: Randomly generate two different indexes, which correspond to the positions of the chromosome excluding the first and last starting points; Perform a transposition operation on these two indexes, that is, exchange the gene values ​​of the two positions in the chromosome; Two different indexes are randomly generated again, and a second transposition operation is performed to exchange the gene values ​​at the other two positions in the chromosome.

8. A substation multi-inspection point path planning system based on dynamic learning factors based on the method according to any one of claims 1 to 7, characterized in that: The system comprises: The first module is used to construct an environmental grid map and input the xy coordinates of the substation multi-task inspection points and the Euclidean distances between their coordinates; The second module is used to perform multi-inspection point task planning using an improved PSO algorithm, namely the IPSO algorithm. The improved PSO algorithm can balance the global search capability and local improvement capability of the PSO algorithm, improve population diversity and expand the search scope of the algorithm; The third module is used to integrate the Lazy Theta* algorithm. Based on the sequential planning and output of the task point list for the substation multi-task point inspection tasks based on the IPSO algorithm, the path planning between each task point is realized.

9. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 7 is implemented.

Images