Charging robot path planning method

Abstract

The invention relates to a charging robot path planning method. The method comprises the following steps: calculating an initial path from an initial grid to a target grid of a charging robot in a grid map by using an A * algorithm; dividing a search space determined by the initial path into sub search spaces; initializing in each sub-search space and expanding a population through a pseudo-reverse learning strategy; the obtained current population is randomly divided into two sub-populations, and the two sub-populations are respectively marked as a first sub-population and a second sub-population; for the first sub-population, iterative optimization is carried out through a first strategy according to the fitness, and for the second sub-population, iterative optimization is carried out through a second strategy according to the fitness; the first 50% of individuals in the first sub-population and the second sub-population form a third population; performing crossover operation on the No.3 population to obtain a current population; calculating the fitness of the current population individuals; and iteration is carried out until the number of iterations reaches the maximum number of iterations, or the optimal values of the two sub-populations tend to be the same, and an optimal path is obtained.

Description

Technical Field

[0001] The present invention relates to the technical field of robot path planning, and in particular to a charging robot path planning method. Background Art

[0002] With the continuous development of the new energy vehicle market, the intelligence and automation of charging equipment have also received increasing attention. As a new type of charging equipment, shared charging robots have the advantages of flexible deployment and automated charging, providing a new solution for charging new energy vehicles. However, in a complex environment, how to enable charging robots to complete charging tasks efficiently and accurately has become an urgent problem to be solved. As an important part of charging robot technology, the study of path planning and its optimization method is of great significance.

[0003] Existing path planning methods, such as the A* algorithm, perform path planning on a grid map. Although they can effectively solve the path planning problem from the starting point to the target point, the cost used in the optimized cost function is only based on the distance, which has limitations. As a heuristic optimization algorithm, the whale optimization algorithm simulates the migration, foraging and social behavior of whales in the ocean to find the optimal solution to the problem. The whale optimization algorithm has good global search capabilities and local development capabilities, and is suitable for solving complex optimization problems. Although the whale optimization algorithm has good performance in solving path planning problems, it still has disadvantages such as slow convergence speed, low solution accuracy, and easy to fall into local optimal solutions. Summary of the invention

[0004] In order to solve the above technical problem or at least partially solve the above technical problem, the present invention provides a charging robot path planning method.

[0005] In a first aspect, the present invention provides a charging robot path planning method, comprising:

[0006] The A* algorithm is used to calculate the initial path from the starting grid to the target grid of the charging robot in the grid map;

[0007] Plan the initial path based on the path optimization algorithm to obtain the optimal path, including:

[0008] The search space determined by the initial path is divided into sub-search spaces, and the path optimization is further performed in the sub-search spaces;

[0009] Initialize and expand the population through a pseudo-reverse learning strategy;

[0010] Loop execution:

[0011] The current population is randomly divided into two sub-populations, marked as sub-population 1 and sub-population 2 respectively;

[0012] For sub-population No. 1, the first strategy is used to iterate and optimize according to the fitness; for sub-population No. 2, the second strategy is used to iterate and optimize according to the fitness;

[0013] The first 50% of individuals in subpopulation 1 and subpopulation 2 form population 3;

[0014] Perform a crossover operation on population 3 to obtain the current population;

[0015] Calculate the fitness of individuals in the current population;

[0016] Iterate until the number of iterations reaches the maximum number of iterations T max , or the optimal values ​​of the two sub-populations converge within the allowable error range, the local optimal path of the corresponding sub-search space is obtained based on the optimization result, and the global optimal path is obtained through the local optimal path.

[0017] Furthermore, the grid method is used to model the environment to obtain a grid map. The height of the environment map is h and the width is w. The map is evenly divided into H×W small grids of equal length and width. xy Represents each small grid, and the entire grid map G is represented by:

[0018] G=S xy ,x∈Z∩[1,M],y∈Z∩[1,N];

[0019] Assign values ​​to each grid based on whether there is an obstacle:

[0020]

[0021] And determine the starting grid and target grid of the charging robot in the grid map.

[0022] Furthermore, the A* algorithm is used to calculate the initial path from the starting grid to the target grid of the charging robot in the grid map, including:

[0023] Create an open set for storing the points to be checked and a closed set for storing the checked points;

[0024] Add the starting grid to the open set, and calculate the total cost between the starting grid and the target grid through the cost function; the expression of the cost function is as follows:

[0025] f(n)=g(n)+h(n);

[0026] Where: f(n) represents the total cost value of the grid n to be estimated, g(n) represents the shortest path cost value from the grid n to be estimated to the starting grid, and h(n) represents the estimated cost value from the grid n to be estimated to the target grid;

[0027] Loop execution:

[0028] Take the grid with the smallest total cost in the open set as the current grid;

[0029] Check whether the current grid is the target grid;

[0030] If yes, the loop ends, and the path is reconstructed according to the parent-child relationship between grids to obtain the initial path;

[0031] If not, move the current grid from the open set to the closed set; obtain the neighbor grid of the current grid, traverse each neighbor grid, if the neighbor grid of the current grid is in the closed set, skip it, if the neighbor grid of the current grid is not in the open set, add it to the open set, and calculate the g(n) value from the starting grid to the neighbor grid, where the path used to calculate the g(n) value is: the path from the starting grid to the neighbor grid via the current grid. If the new g(n) value of the neighbor grid is smaller than the previously recorded value, update the g(n) value and the total cost value, and record the current grid as the parent grid of the neighbor grid to facilitate reconstruction of the path based on the parent-child relationship.

[0032] Furthermore, g(n) and h(n) are calculated using Euclidean distance or Manhattan distance.

[0033] Furthermore, the initialization and expansion of the population through the pseudo-reverse learning strategy includes:

[0034] Initialize a population with a population size of N;

[0035] The initial position of each individual in the population is generated in the sub-search space as follows:

[0036] Determine the supremum of the entire spatial dimension of the subsearch space and infimum in, represents the maximum value of each spatial dimension from 1 to d, Represents the minimum value of each spatial dimension from 1 to d, where d is the number of spatial dimensions;

[0037] Randomly select non-repeating intervals from each spatial dimension, and randomly select a point in the interval of each spatial dimension as the initial position of the individual. For the kth individual in the population, its initial position is expressed as:

[0038]

[0039] After obtaining the initial position of each individual, the relative initial position of the individual is obtained through the pseudo-reverse learning strategy to expand the population:

[0040] for Its relative initial position is expressed as

[0041] in, for A random number in the interval, j=1,2...d.

[0042] Furthermore, the method of randomly and non-repetitively selecting intervals from each spatial dimension is: create d groups of integer arrays that increase by 1 from 1 to N and randomly shuffle the arrangement order to obtain a dimension interval array, the d groups of dimension interval arrays subdivide each spatial dimension of the sub-search space into N different intervals, and shuffle the interval order, and take the kth element of all d groups of dimension interval arrays as the interval of the spatial dimension of the kth individual in the population.

[0043] Furthermore, the first strategy includes:

[0044] The convergence factor a is updated nonlinearly according to the iteration rounds as follows:

[0045]

[0046] Among them, a i is the initial value of the convergence factor a, a f is the final value of the convergence factor a, α is a nonlinear adjustment coefficient greater than zero, t is the current number of iterations, and T is the maximum number of iterations;

[0047] Update the control coefficient vector A based on the convergence factor a and calculate the coefficient vector C:

[0048] C = 2·r1,

[0049] A=2a·r2-a, where r1 and r2 are random numbers in the range [0,1];

[0050] The strategy for updating individual positions is selected according to the value of the random probability Pr and the control coefficient vector A:

[0051] When Pr<0.5 and |A|<1, the position P of the current best individual in subpopulation 1 is determined according to the individual's fitness. * , the remaining individuals update their positions based on the position of the current optimal individual, and the update formula is as follows:

[0052] P(t+1)=P * (t)-A·(|C·P * (t)|-P(t));

[0053] Among them, P * (t) is the optimal individual position of the current iteration round t; P(t) is the position of the remaining single individuals, and P(t+1) is the individual position of the t+1th iteration round;

[0054] When Pr<0.5, |A|≥1, randomly select an individual position from the current subpopulation 1 as a reference to update the positions of the remaining individuals. The update formula is as follows:

[0055] P(t+1)=P r (t)-A·(|C·P r (t)|-P(t));

[0056] Among them, P r (t) is the randomly selected individual position of the current iteration round t.

[0057] When Pr≥0.5, the Lévy flight strategy is used to update the position, and the position update formula is:

[0058]

[0059] in, is the point product; L is the Lévy flight step length that conforms to the Lévy distribution;

[0060] The Lévy flight step length that conforms to the Lévy distribution is expressed as:

[0061]

[0062] Where: u and v are d-dimensional vectors, each element of u and v is a random number in the range [0,1] that follows a standard normal distribution; λ = 1.5; the calculation formula for σ is:

[0063]

[0064] Where: Γ() is the gamma function, that is, Γ(x) = (x-1)!.

[0065] Furthermore, the second strategy includes:

[0066] Randomly generate five random numbers κ1, κ2, κ3, κ4, κ5, and pre-set parameter Th;

[0067] When κ1<κ2, κ3<κ4, the first global exploration is achieved by updating individual positions in the following way:

[0068]

[0069] in, is a random number that follows a normal distribution, is a random number in the interval [0,1];

[0070] When κ1<κ2, κ3≥κ4, the second global exploration is achieved by updating individual positions in the following way:

[0071]

[0072] Among them: U1 is a binary vector, P r1 (t), P r2 (t) are the positions of two randomly selected individuals in the current iteration round t, is a random number in the interval [0, 1];

[0073] When κ1 ≥ κ2 and κ5 < Th, the first type of local exploration is achieved by updating the individual positions in the following way:

[0074]

[0075] Among them, P r3 (t), P r4 (t), P r5 (t) are the positions of two randomly selected individuals in the current iteration round t, is a random number generated in the interval [0, 1], δ is a parameter used to control the search direction determined by the random number, and the calculation formula is:

[0076]

[0077] ψ(t) is the defense factor, and its calculation formula is:

[0078]

[0079] S(t) is the diffusion factor. For the i-th individual, its diffusion factor calculation formula is:

[0080]

[0081] Among them, fit(P i (t)) is the fitness of the i-th individual at the position P i (t) in the t-th iteration, and ε is a minimum value to prevent the divisor from being 0;

[0082] When κ1 ≥ κ2 and κ5 ≥ Th, the second type of local exploration is achieved by updating the individual positions in the following way:

[0083]

[0084] Among them: b is the velocity convergence factor, and are random numbers within the interval [0, 1], U2 is a binary vector,

[0085]

[0086] Among them: is a random vector generated between 0 and 1.

[0087] Furthermore, the individual fitness of the current population is calculated:

[0088] Among them, f is the total distance cost of the individual position, using the Euclidean distance, Δθ is the heading angle change of the current position compared to the previous position, and ρ1, ρ2 are the corresponding weights.

[0089] In a second aspect, the present invention provides a charging robot path planning device, comprising: at least one processing unit, the processing unit and a storage unit are interconnected through a bus unit, the storage unit stores a computer program, and when the computer program is executed by the processing unit, the charging robot path planning method is implemented.

[0090] In a third aspect, the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the charging robot path planning method is implemented.

[0091] The above technical solution provided by the embodiment of the present invention has the following advantages compared with the prior art:

[0092] The present application uses the A* algorithm to calculate the initial path from the starting grid to the target grid of the charging robot in the grid map; plans the initial path based on the path optimization algorithm to obtain the optimal path, including: dividing the search space determined by the initial path into sub-search spaces, and further optimizing the path in the sub-search spaces; by dividing the search space into sub-search spaces, optimizing in the sub-search spaces, reducing the amount of calculation for each optimization; during optimization, initializing and expanding the population through a pseudo-reverse learning strategy; making the positions in each spatial dimension more uniform and covering the entire sub-search space more uniformly, thereby greatly reducing the probability of the algorithm falling into a local optimum. And this application is executed cyclically: the current population obtained is randomly divided into two sub-populations, marked as sub-population 1 and sub-population 2 respectively; for sub-population 1, iteratively optimize through the first strategy according to the fitness, and for sub-population 2, iteratively optimize through the second strategy according to the fitness; the first 50% of individuals in sub-populations 1 and 2 form population 3; crossover operation is performed on population 3 to obtain the current population; the fitness of individuals in the current population is calculated; iterate until the number of iterations reaches the maximum number of iterations, or the optimal values ​​of the two sub-populations converge within the allowable error range, and the local optimal path of the corresponding sub-search space is obtained based on the optimization result, and the global optimal path is obtained through the local optimal path. This application optimizes the sub-search space through two optimization strategies, and the first 50% of individuals in sub-populations 1 and 2 form population 3. Further looping is performed, and population 3 is formed through two strategies to fully search globally to avoid falling into the local optimum.

[0093] In addition, in the present application, the first strategy combines proximity to the current optimal and random positions, and introduces a good search capability based on optimization. The first strategy uses a nonlinear convergence factor, and the convergence factor has a high value in the early stage to ensure the global search capability of the algorithm; the speed of reduction in the middle of the iteration is fast, so that the algorithm quickly transforms from a global search mode to a local search; in the later stage of the iteration, the value is kept low and not reduced to 0, which can ensure the local search capability of the algorithm in the later stage. Based on the Lévy flight strategy, the update position is more diverse, and it can quickly jump out of the local optimum, thereby improving the overall optimization efficiency of the population.

[0094] In this application, the second strategy combines two global and two local exploration methods, which can better balance global exploration and local utilization during the search process, which helps to improve the search efficiency and solution quality of the algorithm. BRIEF DESCRIPTION OF THE DRAWINGS

[0095] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0096] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0097] Figure 1 A flow chart of a charging robot path planning method provided by an embodiment of the present invention;

[0098] Figure 2 A flowchart of the A* algorithm provided by an embodiment of the present invention;

[0099] Figure 3 A flow chart of a path optimization algorithm provided by an embodiment of the present invention;

[0100] Figure 4 A flow chart of a first strategy provided by an embodiment of the present invention;

[0101] Figure 5 A flow chart of a second strategy provided by an embodiment of the present invention;

[0102] Figure 6 A schematic diagram of a device for implementing a charging robot path planning method provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0103] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0104] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.

[0105] Example 1

[0106] like Figure 1 As shown, the technology of the present invention implements a charging robot path planning method, the steps comprising:

[0107] The grid method is used to model the environment to obtain a grid map, and the starting grid and target grid of the charging robot in the grid map are determined.

[0108] Assume that the height of the environment map is h and the width is w. Divide the map into H×W small grids of equal length and width. xy Represents each small grid, and the entire grid map G is represented by:

[0109] G=S xy ,x∈Z∩[1,M],y∈Z∩[1,N];

[0110] Assign values ​​to each grid based on whether there is an obstacle:

[0111]

[0112] In the grid map, the A* algorithm is used to calculate the initial path from the starting grid to the target grid of the charging robot. Figure 2 As shown, the process of the A* algorithm calculating the initial path from the starting grid to the target grid of the charging robot includes:

[0113] Create an open set for storing the points to be checked and a closed set for storing the checked points;

[0114] The starting grid is added to the open set, and the total cost between the starting grid and the target grid is calculated by the cost function; the expression of the evaluation function is as follows:

[0115] f(n)=g(n)+h(n);

[0116] Where: f(n) represents the total cost value of the grid n to be estimated, g(n) represents the shortest path cost value from the grid n to be estimated to the starting grid, using Euclidean distance or Manhattan distance, and h(n) represents the estimated cost value from the grid n to be estimated to the target grid, using Euclidean distance or Manhattan distance. For the starting grid, g(n) = 0, f(n) = h(n).

[0117] Loop execution:

[0118] Take the grid with the smallest total cost in the open set as the current grid;

[0119] Check whether the current grid is the target grid;

[0120] If yes, the loop ends, and the path is reconstructed according to the parent-child relationship between grids to obtain the initial path;

[0121] If not, move the current grid from the open set to the closed set; get the neighbor grid of the current grid, traverse each neighbor grid, if the neighbor grid of the current grid is in the closed set, skip it, if the neighbor grid of the current grid is not in the open set, add it to the open set, calculate the g(n) value from the starting grid to the neighbor grid, where the path used to calculate the g(n) value is: the path from the starting grid to the neighbor grid through the current grid, if the new g(n) value of the neighbor grid is smaller than the previously recorded value, update the g(n) value and the total cost. Record the current grid as the parent grid of the neighbor grid to facilitate the reconstruction of the path based on the parent-child relationship.

[0122] Plan the initial path based on the path optimization algorithm to obtain the optimal path, such as Figure 3 As shown, the specific process is as follows:

[0123] The search space determined by the initial path is divided into sub-search spaces, and the path optimization is further performed in the sub-search space. If the initial path is long, the scope of the search space determined by the initial path is large, and the subsequent optimization calculation is large. This application divides the search space into several sub-search spaces to reduce the calculation by narrowing the optimization range.

[0124] Initialize and expand the population through a pseudo-reverse learning strategy:

[0125] Initialize a population with a population size of N;

[0126] The initial position of each individual in the population is generated in the sub-search space as follows:

[0127] Determine the supremum of the entire spatial dimension of the subsearch space and infimum in, represents the maximum value of each spatial dimension from 1 to d, Represents the minimum value of each spatial dimension from 1 to d, where d is the number of spatial dimensions.

[0128] Create d groups of integer arrays that increase by 1 from 1 to N and randomly shuffle the order to obtain dimension interval arrays. The d groups of dimension interval arrays subdivide each spatial dimension of the sub-search space into N different intervals and shuffle the interval order. Take the kth element of all d groups of dimension interval arrays as the interval of the spatial dimension of the kth individual in the population, and realize random and non-repetitive selection of intervals from each spatial dimension; and randomly select a point in the interval of each spatial dimension as the initial position of the individual. For the kth individual in the population, its initial position is expressed as:

[0129]

[0130] This makes the positions of points in each spatial dimension more uniform and covers the entire sub-search space more evenly, greatly reducing the probability of the algorithm falling into a local optimum.

[0131] After obtaining the initial position of each individual, the relative initial position of the individual is obtained through the pseudo-reverse learning strategy to expand the population:

[0132] for Its relative initial position is expressed as

[0133] in, for A random number in the interval, j=1,2...d.

[0134] The cycle executes the following process:

[0135] The current population is randomly divided into two sub-populations, marked as sub-population 1 and sub-population 2 respectively;

[0136] For sub-population 1, the first strategy is used to iterate and optimize the sub-population, that is, Figure 4 As shown, the loop executes the following process to update the position:

[0137] The convergence factor a is updated nonlinearly according to the iteration rounds as follows:

[0138]

[0139] Among them, a i is the initial value of the convergence factor a, a f is the final value of the convergence factor a, α is a nonlinear adjustment coefficient greater than zero, t is the current number of iterations, and T is the maximum number of iterations.

[0140] Update the control coefficient vector A based on the convergence factor a and calculate the coefficient vector C:

[0141] C = 2·r1,

[0142] A=2a·r2-a, where r1 and r2 are random numbers in the range of [0,1].

[0143] The control coefficient vector A can realize the conversion between global search and local optimization iteration, and the control coefficient vector A is determined by the convergence factor a. The larger the convergence factor a is, the larger the control coefficient vector A is, which improves the global search ability of the algorithm. The smaller the convergence factor a is, the smaller the control coefficient vector A is, and the stronger the local search ability of the system is. In order to improve the optimization ability of the algorithm, the global and local search capabilities of the algorithm are dynamically compensated during the iteration process, and the above nonlinear convergence factor a update method is adopted.

[0144] The strategy for updating individual positions is selected according to the value of the random probability Pr and the control coefficient vector A:

[0145] When Pr<0.5 and |A|<1, the position P of the current best individual in subpopulation 1 is determined according to the individual's fitness. * , the remaining individuals update their positions based on the position of the current optimal individual, and the update formula is as follows:

[0146] P(t+1)=P * (t)-A·(|C·P * (t)|-P(t));

[0147] Where: P * (t) is the optimal individual position of the current iteration round t; P(t) is the remaining individual positions, and P(t+1) is the individual position of the t+1th iteration round;

[0148] When Pr<0.5, |A|≥1, randomly select an individual position from the current subpopulation 1 as a reference to update the positions of the remaining individuals. The update formula is as follows:

[0149] P(t+1)=P r (t)-A·(|C·P r (t)|-P(t));

[0150] Where: P r (t) is the randomly selected individual position of the current iteration round t.

[0151] When Pr≥0.5, the Lévy flight strategy is used to update the position, making the position more diverse, being able to quickly jump out of the local optimum, and improving the optimization efficiency of the population. The position update formula is:

[0152]

[0153] in, is the point product; L is the Lévy flight step length that conforms to the Lévy distribution.

[0154] The mathematical model of Lévy distribution is:

[0155]

[0156] Where: γ is the scale parameter, μ is the shift parameter; the Lévy flight step length that conforms to the Lévy distribution is expressed as:

[0157]

[0158] Where: u and v are d-dimensional vectors, each element of u and v is a random number in the range [0,1] that follows a standard normal distribution; λ = 1.5; the calculation formula for σ is:

[0159]

[0160] Where: Γ() is the gamma function, that is, Γ(x) = (x-1)!.

[0161] At each iteration, the optimal solution of subpopulation 1 is output, and the top 50% of subpopulation 1 is selected as part of population 3.

[0162] For subpopulation 2, the second strategy is used to iterate and optimize, that is, Figure 5 As shown, the loop executes the following process to update the position:

[0163] Randomly generate five random numbers κ1, κ2, κ3, κ4, κ5;

[0164] When κ1<κ2, κ3<κ4, the first global exploration is achieved by updating individual positions in the following way:

[0165]

[0166] in, is a random number that follows a normal distribution, is a random number in the interval [0,1].

[0167] When κ1<κ2, κ3≥κ4, the second global exploration is achieved by updating individual positions in the following way:

[0168]

[0169] Among them: U1 is a binary vector, P r1 (t), P r2 (t) are the positions of two randomly selected individuals at the current iteration round t, is a random number in the interval [0, 1].

[0170] When κ1 ≥ κ2 and κ5 < Th, the first type of local exploration is achieved by updating the individual positions in the following way:

[0171]

[0172] Among them, P r3 (t), P r4 (t), P r5 (t) are the positions of two randomly selected individuals at the current iteration round t, is a random number generated in the interval [0, 1], and δ is a parameter used to control the search direction determined by the random number, and its calculation formula is:

[0173]

[0174] ψ(t) is the defense factor, and its calculation formula is:

[0175]

[0176] S(t) is the diffusion factor. For the i-th individual, its diffusion factor calculation formula is:

[0177]

[0178] Among them, fit(P i (t)) is the fitness of the position P i (t) of the i-th individual at the t-th iteration, and ε is a minimum value to prevent the divisor from being 0.

[0179] When κ1 ≥ κ2 and κ5 ≥ Th, where Th is a preset parameter., the second type of local exploration is achieved by updating the individual positions in the following way:

[0180]

[0181] Among them: b is the velocity convergence factor, and are random numbers in the interval [0, 1], U2 is a binary vector,

[0182]

[0183] Among them: is a random vector generated between 0 and 1.

[0184] At each iteration, the optimal value of subpopulation 2 is output, and the top 50% of individuals in subpopulation 2 are selected as part of population 3.

[0185] Perform a crossover operation on population 3 to obtain the current population;

[0186] Calculate the individual fitness of the current population:

[0187] Among them, f is the total distance cost of the individual position, using the Euclidean distance, Δθ is the heading angle change of the current position compared to the previous position, and ρ1, ρ2 are the corresponding weights.

[0188] Continue to iterate until the number of iterations reaches the maximum number of iterations T, or the optimal values ​​of the two sub-populations converge within the allowable error range. Based on the optimization result, the local optimal path of the corresponding sub-search space is obtained, and the global optimal path is obtained through the local optimal path.

[0189] Example 2

[0190] See also Figure 6 As shown, an embodiment of the present invention provides a robot path planning device, including: the processing unit and the storage unit are interconnected through a bus unit, and the storage unit can be used to store software programs, computer executable programs and modules, such as the software programs, computer executable programs and modules corresponding to a charging robot path planning method in an embodiment of the present invention. The processing unit implements the above-mentioned charging robot path planning method by running the software programs, computer executable programs and modules stored in the storage unit, including:

[0191] The A* algorithm is used to calculate the initial path from the starting grid to the target grid of the charging robot in the grid map;

[0192] Plan the initial path based on the path optimization algorithm to obtain the optimal path, including:

[0193] The search space determined by the initial path is divided into sub-search spaces, and the path optimization is further performed in the sub-search spaces;

[0194] Initialize and expand the population through a pseudo-reverse learning strategy;

[0195] Loop execution:

[0196] The current population is randomly divided into two sub-populations, marked as sub-population 1 and sub-population 2 respectively;

[0197] For sub-population No. 1, the first strategy is used to iterate and optimize according to the fitness; for sub-population No. 2, the second strategy is used to iterate and optimize according to the fitness;

[0198] The first 50% of individuals in subpopulation 1 and subpopulation 2 form population 3;

[0199] Perform a crossover operation on population 3 to obtain the current population;

[0200] Calculate the fitness of individuals in the current population;

[0201] Iterate until the number of iterations reaches the maximum number of iterations, or the optimal values ​​of the two sub-populations converge within the allowable error range. Based on the optimization result, the local optimal path of the corresponding sub-search space is obtained, and the global optimal path is obtained through the local optimal path.

[0202] Of course, the computer program stored in the storage unit of the electric vehicle charging robot arm compliant control device provided in an embodiment of the present invention is not limited to the method operations described above, and can also execute related operations in the charging robot path planning method provided in any embodiment of the present invention.

[0203] Example 3

[0204] An embodiment of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program. When the computer program is executed, the charging robot path planning method is implemented, including:

[0205] The A* algorithm is used to calculate the initial path from the starting grid to the target grid of the charging robot in the grid map;

[0206] Plan the initial path based on the path optimization algorithm to obtain the optimal path, including:

[0207] The search space determined by the initial path is divided into sub-search spaces, and the path optimization is further performed in the sub-search spaces;

[0208] Initialize and expand the population through a pseudo-reverse learning strategy;

[0209] Loop execution:

[0210] The current population is randomly divided into two sub-populations, marked as sub-population 1 and sub-population 2 respectively;

[0211] For sub-population No. 1, the first strategy is used to iterate and optimize according to the fitness; for sub-population No. 2, the second strategy is used to iterate and optimize according to the fitness;

[0212] The first 50% of individuals in subpopulation 1 and subpopulation 2 form population 3;

[0213] Perform a crossover operation on population 3 to obtain the current population;

[0214] Calculate the fitness of individuals in the current population;

[0215] Iterate until the number of iterations reaches the maximum number of iterations, or the optimal values ​​of the two sub-populations converge within the allowable error range. Based on the optimization result, the local optimal path of the corresponding sub-search space is obtained, and the global optimal path is obtained through the local optimal path.

[0216] A computer-readable storage medium provided in an embodiment of the present invention stores a computer program which is not limited to the method operations described above, but can also execute related operations in a charging robot path planning method provided in any embodiment of the present invention.

[0217] In the embodiments provided by the present invention, it should be understood that the disclosed structures and methods can be implemented in other ways. For example, the structural embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, structures or units, which can be electrical, mechanical or other forms.

[0218] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0219] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0220] The foregoing is merely a specific embodiment of the present invention, which enables those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features claimed herein.

Claims

1. A charging robot path planning method, characterized in that: include: The A* algorithm is used to calculate the initial path from the starting grid to the target grid of the charging robot in the grid map; Plan the initial path based on the path optimization algorithm to obtain the optimal path, including: The search space determined by the initial path is divided into sub-search spaces, and the path optimization is further performed in the sub-search spaces; Initialize and expand the population through a pseudo-reverse learning strategy; Loop execution: The current population is randomly divided into two sub-populations, marked as sub-population 1 and sub-population 2 respectively; For sub-population No. 1, the first strategy is used to iterate and optimize according to the fitness; for sub-population No. 2, the second strategy is used to iterate and optimize according to the fitness; The first 50% of individuals in subpopulation 1 and subpopulation 2 form population 3; Perform a crossover operation on population 3 to obtain the current population; Calculate the fitness of individuals in the current population; Iterate until the number of iterations reaches the maximum number of iterations, or the optimal values ​​of the two sub-populations converge within the allowable error range. Based on the optimization result, the local optimal path of the corresponding sub-search space is obtained, and the global optimal path is obtained through the local optimal path.

2. The charging robot path planning method according to claim 1, characterized in that: The grid method is used to model the environment and obtain a grid map. The height of the environment map is h and the width is w. The map is evenly divided into H×W small grids of equal length and width. xy Represents each small grid, and the entire grid map G is represented by: G=S xy ,x∈Z∩[1,M],y∈Z∩[1,N]; Assign values ​​to each grid based on whether there is an obstacle: And determine the starting grid and target grid of the charging robot in the grid map.

3. The charging robot path planning method according to claim 1, characterized in that: The A* algorithm is used to calculate the initial path from the starting grid to the target grid of the charging robot in the grid map, including: Create an open set for storing the points to be checked and a closed set for storing the checked points; Add the starting grid to the open set, and calculate the total cost between the starting grid and the target grid through the cost function; the expression of the cost function is as follows: f(n)=g(n)+h(n); Where: f(n) represents the total cost value of the grid n to be estimated, g(n) represents the shortest path cost value from the grid n to be estimated to the starting grid, and h(n) represents the estimated cost value from the grid n to be estimated to the target grid; Loop execution: Take the grid with the smallest total cost in the open set as the current grid; Check whether the current grid is the target grid; If yes, the loop ends, and the path is reconstructed according to the parent-child relationship between grids to obtain the initial path; If not, move the current grid from the open set to the closed set; obtain the neighbor grid of the current grid, traverse each neighbor grid, if the neighbor grid of the current grid is in the closed set, skip it, if the neighbor grid of the current grid is not in the open set, add it to the open set, and calculate the g(n) value from the starting grid to the neighbor grid, where the path used to calculate the g(n) value is: the path from the starting grid to the neighbor grid via the current grid. If the new g(n) value of the neighbor grid is smaller than the previously recorded value, update the g(n) value and the total cost value, and record the current grid as the parent grid of the neighbor grid to facilitate reconstruction of the path based on the parent-child relationship.

4. The charging robot path planning method according to claim 1, characterized in that: The initialization and expansion of the population through the pseudo-reverse learning strategy include: Initialize a population with a population size of N; The initial position of each individual in the population is generated in the sub-search space as follows: Determine the supremum of the entire spatial dimension of the subsearch space and infimum in, represents the maximum value of each spatial dimension from 1 to d, Represents the minimum value of each spatial dimension from 1 to d, where d is the number of spatial dimensions; Randomly and non-repeatedly select intervals from each spatial dimension, and randomly select a point within the intervals of each spatial dimension as the initial position of an individual. For the k-th individual in the population, its initial position is expressed as: After obtaining the initial position of each individual, obtain the relative initial position of the individual through the pseudo-opposition-based learning strategy to expand the population: for Its relative initial position is expressed as in, for A random number in the interval, j=1,2...d.

5. The charging robot path planning method according to claim 4, characterized in that: The method of randomly and non-repeatedly selecting intervals from each spatial dimension is as follows: Create d groups of integer arrays that increase by 1 from 1 to N and randomly shuffle the order to obtain the dimension interval arrays. The d groups of dimension interval arrays divide each spatial dimension of the sub-search space into N different intervals and shuffle the interval order. Take the k-th element of all d groups of dimension interval arrays as the interval of the spatial dimension of the k-th individual in the population.

6. The charging robot path planning method according to claim 1, characterized in that: The first strategy includes: Non-linearly update the convergence factor a according to the iteration round in the following way: Among them, a i is the initial value of the convergence factor a, a f is the final value of the convergence factor a, α is a nonlinear adjustment coefficient greater than zero, t is the current number of iterations, and T is the maximum number of iterations; Update the control coefficient vector A based on the convergence factor a, and calculate the coefficient vector C: C＝2·r1， A = 2a·r2 - a, where r1 and r2 are random numbers in the range of [0, 1]; Select the strategy for updating the individual position according to the random probability Pr and the value of the control coefficient vector A: When Pr<0.5 and |A|<1, the position P of the current best individual in subpopulation 1 is determined according to the individual's fitness. * , the remaining individuals update their positions based on the position of the current optimal individual, and the update formula is as follows: P(t+1)=P * (t)-A·(|C·P * (t)|-P(t)); Among them, P * (t) is the optimal individual position of the current iteration round t; P(t) is the position of the remaining single individuals, and P(t+1) is the individual position of the t+1th iteration round; When Pr < 0.5 and |A| ≥ 1, randomly select an individual position from the current sub-population No. 1 as a reference to update the positions of the remaining individuals. The update formula is as follows: P(t+1)=P r (t)-A·(|C·P r (t)|-P(t)); Among them, P r (t) is the randomly selected individual position of the current iteration round t. When Pr ≥ 0.5, use the Lévy flight strategy to update the position. The position update formula is: in, is the point product; L is the Lévy flight step length that conforms to the Lévy distribution; The Lévy flight step size that conforms to the Lévy distribution is expressed as: where: u and v are d-dimensional vectors, and each element of u and v is a random number subject to the standard normal distribution in the range of [0, 1]; λ = 1.

5. The calculation formula of σ is: where: Γ( ) is the gamma function, that is, Γ(x) = (x - 1)!.

7. The charging robot path planning method according to claim 1, characterized in that: The second strategy includes: Randomly generate five random numbers κ1, κ2, κ3, κ4, κ5, and preset the parameter Th; When κ1 < κ2 and κ3 < κ4, update the individual position in the following way to achieve the first type of global exploration: in, is a random number that follows a normal distribution, is a random number in the interval [0,1]; When κ1 < κ2 and κ3 ≥ κ4, update the individual position in the following way to achieve the second type of global exploration: Among them: U1 is a two-element vector, P r1 (t),P r2 (t) are two randomly selected individual positions in the current iteration round t, is a random number in the interval [0,1]; When κ1 ≥ κ2 and κ5 < Th, update the individual position in the following way to achieve the first type of local exploration: Among them, P r3 (t),P r4 (t),P r5 (t) are two randomly selected individual positions in the current iteration round t, is a random number generated in the interval [0,1], and δ is a parameter determined by the random number to control the search direction. The calculation formula is: ψ(t) is the defense factor, and its calculation formula is: S(t) is the diffusion factor. For the i-th individual, its diffusion factor calculation formula is: Among them, fit(P i (t)) is the position P of the i-th individual at the t-th iteration i The fitness of (t), ε, is to prevent the minimum value of division by 0; When κ1 ≥ κ2 and κ5 ≥ Th, update the individual position in the following way to achieve the second type of local exploration: Where: b is the velocity convergence factor, and is a random number in the interval [0,1], U2 is a binary vector, in: is a random vector generated between 0 and 1.

8. The charging robot path planning method according to claim 1, characterized in that: Calculate the individual fitness of the current population: Among them, f is the total distance cost of the individual position, using the Euclidean distance, Δθ is the heading angle change of the current position compared to the previous position, and ρ1, ρ2 are the corresponding weights.

9. A charging robot path planning device, characterized in that: Including: At least one processing unit, the processing unit and the storage unit are interconnected through a bus unit. The storage unit stores a computer program. When the computer program is executed by the processing unit, the charging robot path planning method described in any one of claims 1-8 is implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by the processor, the charging robot path planning method described in any one of claims 1-8 is implemented.

Images