Indoor nuclear environment multi-target-point inspection path planning method for mobile robot

Abstract

According to the indoor nuclear environment multi-target-point inspection path planning method for the mobile robot, a grid map of an indoor nuclear environment is prior information, and each grid in the grid map comprises radiation field information and obstacle information; the method comprises the following steps: 1, multi-target point inspection path mathematical modeling: considering path length, accumulated radiation dose and turning energy consumption, and respectively establishing an inspection path quality evaluation model and a paired path quality evaluation model; 2, acquiring an inspection path based on an upper-layer algorithm; 3, constructing a lower-layer algorithm to calculate pairwise paths; and 4, circularly operating the two-layer algorithm to obtain an optimal solution. The method has the advantages that aiming at the MTIPP problem of the mobile robot in the indoor nuclear environment, the path length, the accumulated radiation dose and the turning energy consumption are comprehensively considered, a double-layer mathematical model and a double-layer hybrid algorithm are established, and compared with path planning which only considers a single factor in the prior art, the routing inspection path obtained through planning has the advantage that the routing inspection efficiency is improved. And the method has better security and lower global cost.

Description

Technical Field

[0001] The invention relates to the technical field of path planning, and in particular to a method for planning a multi-target point inspection path for a mobile robot in an indoor nuclear environment. Background Art

[0002] In the indoor environment of nuclear power plants, it is often necessary to perform multi-target inspection tasks. Under the premise that the indoor map of the nuclear power plant (the map contains radiation field and obstacle information) is known, planning a suitable inspection path is the basis for the safe and smooth execution of the multi-target inspection tasks. The above multi-target inspection tasks usually adopt the method of manual inspection. The path planning of manual inspection mostly only considers the cumulative radiation dose, and the path planning difficulty is relatively low.

[0003] It is rare to use mobile robots to perform multi-target inspection tasks. The path planning of multi-target inspection tasks is difficult. It is necessary to consider the turning energy consumption of the mobile robot (as few turns as possible and as small turning angle as possible), the cumulative radiation dose (as low as possible without exceeding the tolerance limit of the mobile robot) and the path length (the shorter the better). Currently, the industry has not seen a path planning method for mobile robots to perform multi-target inspection tasks. Summary of the invention

[0004] The purpose of the present invention is to overcome the shortcomings of the prior art and provide a method for planning inspection paths for multiple target points in an indoor nuclear environment. It solves the problem of how to reasonably plan inspection paths for multiple target points inspection tasks of a mobile robot under the premise that the indoor nuclear environment map is known.

[0005] The technical solution of the present invention is: a method for mobile robot indoor nuclear environment multi-target point inspection path planning, wherein the grid map of the indoor nuclear environment is prior information, and each grid in the grid map contains radiation field information and obstacle information;

[0006] Here’s how:

[0007] S01, mathematical modeling of multi-target inspection paths: considering the path length, cumulative radiation dose and turning energy consumption, the inspection path quality evaluation model and paired path quality evaluation model are established respectively;

[0008] S02, obtain the inspection path based on the upper algorithm: use the GA algorithm to obtain the rough optimization solution of the inspection path; use the rough optimization solution of the inspection path to process the initial pheromone inhomogeneity in the ACO algorithm; consider the path length, cumulative radiation dose and turning energy consumption, construct the heuristic function and selection probability function of the ACO algorithm, and optimize the pheromone update rule of the ACO algorithm; complete the construction of the upper algorithm, run the upper algorithm, and obtain the precise optimization solution of the inspection path;

[0009] S03, construct the lower-level algorithm to calculate the paired paths: consider the path length, cumulative radiation dose and turning energy consumption, and establish the heuristic function of the A* algorithm; complete the construction of the lower-level algorithm, run the lower-level algorithm, and plan all paired paths in the inspection path optimization solution;

[0010] S04, the two-layer algorithm runs in a loop to obtain the optimal solution: the inspection path quality evaluation model is used to compare the quality of the inspection path optimization solution obtained in the current iteration with that obtained in the previous iteration. If the inspection path optimization solution obtained in the current iteration is better, the inspection path optimization solution obtained in the current iteration will be used in the next iteration; if the inspection path optimization solution obtained in the current iteration is worse or consistent, the iteration is terminated and the inspection path optimization solution obtained in the previous iteration is output.

[0011] A further technical solution of the present invention is: Step S01 is described in detail as follows:

[0012] A. Establish an inspection path quality evaluation model: The multi-target inspection task requires the establishment of a Hamiltonian circuit that passes each inspection target point only once and eventually returns to the starting point. Considering the three dimensions of path length, cumulative radiation dose, and turning energy consumption, an inspection path quality evaluation model, or an inspection target point traversal sequence quality evaluation model, is established. See Formula 1.

[0013] Formula 1:

[0014] Where λ is the total cost of the three dimensions of all paired paths in the inspection path; T is the set of inspection target points, T = {T1, T2, T3, …, Tn}, and the subscript n is the total number of inspection target points; is the inspection target point T i and T j The path length cost matrix between is the inspection target point T i and T j The cumulative radiation dose cost matrix between is the inspection target point T i and T j The turning energy cost matrix between ,j is a decision variable. If the value is 1, it means that the inspection target point T is selected. i To the inspection target point T j If the value is 0, it means that the inspection target point T is not selected. i To the inspection target point T j The path between them; ω1 is the weight coefficient of the path length, ω2 is the weight coefficient of the cumulative radiation dose, and ω3 is the weight coefficient of the turning energy consumption;

[0015] B. Establish a pairwise path cost calculation model: T i and Tj are any two inspection target points, let T i to T j The set of grids that the path with the smallest total cost value in the three dimensions passes through in sequence is P, P = {p1, p, p3…p m}, then T i The grid where it is located is p1, then T j The grid is p m ; Considering the three dimensions of path length, cumulative radiation dose and turning energy consumption, a pairwise path cost calculation model is established, as shown in Formula 2;

[0016] Formula 2:

[0017] Where M(P) is the pairwise path T i to T j The total cost of the three dimensions, L(P) is T i to T j The path length cost value, R(P) is T i to T j The cumulative radiation dose cost of T i to T j The turning energy consumption cost value; ε1 is the weight coefficient of the steering angle variable, ε2 is the weight coefficient of the number of turns, l(pa-1,pa) is the grid p α-1 With grid p α The center point of the line, l(pa,pa +1 ) is the grid p α With grid p α+1 The center point of the line, ψ[l(pa -1 ,pa),l(pa,pa +1 )] is the line segment l(pa -1 ,pa) and line segment l(pa,pa +1 ), Nturn is the number of turns; (xα, yα) is the grid p α The coordinates of (xα +1 ,yα +1 ) is the grid p α+1 The coordinates of the grid are (xα-1, yα-1) α-1 Coordinates of R(pα,pα +1 ) is the mobile robot from the grid p a Travel to grid p a+1 The cumulative radiation dose received, is the average radiation dose rate of grid pα, is the grid pα +1 The average radiation dose rate, l is the average radiation dose rate of the grid from pα to pα +1If the distance between grids pα and pα +1 If the grids pα and pα are adjacent, then l is equal to 1 times the grid side length. +1 End point contact is equal to times the grid side length, v is the distance from grid pα to grid pα +1 speed.

[0018] A further technical solution of the present invention is: Step S02 is described in detail as follows:

[0019] A. Use GA algorithm to obtain the rough optimization solution of the inspection path:

[0020] The chromosomes in the GA algorithm correspond to the inspection target point traversal order, the genes in the GA algorithm correspond to the inspection target points, and the fitness in the GA algorithm corresponds to the inspection target point traversal order evaluation function; the GA algorithm process is as follows:

[0021] a1, real number coding: the inspection target points are deployed in sequence according to the numbers 1, 2, 3…n, indicating that the mobile robot should traverse n inspection target points and randomly generate a traversal order of inspection target points as the initial population;

[0022] a2, create a taboo list: for each true gene code, the selected taboo list is marked as 1, and the unselected taboo list is marked as 0;

[0023] a3, population initialization: set the evolutionary generation counter G and the maximum evolutionary generation G max , randomly generate the initial population;

[0024] a4, individual evaluation: evaluate the traversal order of all inspection target points according to the Hamiltonian circuit quality evaluation function;

[0025] a5, individual selection: select the traversal order of multiple inspection target points with the smallest generation value as the operator, apply it to the existing population, so that individuals with good genes are retained for the next generation;

[0026] a6, crossover operation: according to the preset crossover probability P c Select individuals from the existing population and generate new offspring individuals through the crossover operator, thus generating a new inspection target point traversal order;

[0027] a7, evolution termination judgment: when G≤G max Repeat steps a3 to a6; when G>G max When , the evolution terminates, and the individual with the smallest generation value in the population is the rough optimization solution of the inspection path, which is recorded as path GA ;

[0028] B. Non-uniform treatment of the initial pheromone in the ACO algorithm: The initial pheromone in the ACO algorithm is non-uniformly treated by using the rough optimization solution of the inspection path obtained by the GA algorithm, as shown in Formula 3;

[0029] Formula 3:

[0030] In the formula, τ 0 (i,j) represents the inspection target point T i With T j The initial pheromone concentration after the uneven treatment between the inspection target point T i With T j is a pair of paths, C is the initial pheromone weight value, τ is the set initial pheromone concentration, and st is the constraint condition;

[0031] C. Optimize the heuristic function of the ACO algorithm: Consider the impact of path length, cumulative radiation dose and turning energy consumption on the heuristic function, and introduce an adaptive adjustment coefficient to balance the convergence speed and global search capability of the ACO algorithm to construct the heuristic function of the ACO algorithm, as shown in Formula 4;

[0032] Formula 4:

[0033] Where η t (i, j) is the tth iteration from the inspection target point T i to T j The heuristic function value of c is the current iteration number, I max is the maximum number of iterations; ξ is the adaptive adjustment coefficient; ω1 is the weight coefficient of the path length cost value, ω2 is the weight coefficient of the cumulative radiation dose cost value, and ω3 is the weight coefficient of the turning energy consumption cost value. From the inspection target point T i to T j The path length cost matrix, From the inspection target point T i to T j The cumulative radiation dose cost matrix, From the inspection target point T i to T j The turning energy cost matrix;

[0034] D. Optimize the transition probability function of the ACO algorithm: Consider the pheromone concentration and the heuristic function comprehensively, and construct the transition probability function of the ACO algorithm, as shown in Formula 5;

[0035] Formula 5:

[0036] Where P t(i, j) is the tth iteration from the inspection target point T i to T j The probability of selection; α is the pheromone importance factor, β is the heuristic function importance factor; τ t (i,j) indicates the tth iteration from the inspection target point T i to T j The pheromone concentration of t (i,u) represents the tth iteration from the inspection target point T i to T u The pheromone concentration of T t ′ is the set of inspection target points that have not been traversed in the tth iteration;

[0037] E. Optimize the pheromone update rule of the ACO algorithm: Consider the path length, cumulative radiation dose, and turning energy consumption, and optimize the pheromone update rule of the ACO algorithm, as shown in Formula 6. This completes the construction of the upper-level algorithm, and runs the upper-level algorithm to obtain the precise optimization solution of the inspection path.

[0038] Formula 6:

[0039] In the formula, τt +1 (i, j) indicates the t+1th iteration from the inspection target point T i to T j , τt(i,j) represents the tth iteration from the inspection target point T i to T j ρ is the pheromone concentration, Q is the pheromone deposition factor, υ is the pheromone concentration, t-mi is the total cost of the three dimensions of the optimal inspection path of the tth iteration, υ t-ma is the total cost of the three dimensions of the worst inspection path in the tth iteration; I t is the optimal inspection path of the tth iteration; υ is the total cost of the three dimensions of the inspection path; assuming that there are n inspection target points in the inspection path, then “Ti,Ti +1 " are two consecutive inspection target points in the inspection path, and "Tn, T1" are two consecutive inspection target points in the inspection path. The inspection path requires that each inspection target point is passed only once and finally returns to the starting point; is the inspection target point T i and T i+1 The path length cost matrix between is the inspection target point T i and T i+1 The cumulative radiation dose cost matrix between is the inspection target point T i and T i+1 The turning energy cost matrix between is the inspection target point T n The path length cost matrix between and T1, is the inspection target point T n and T1, the cumulative radiation dose cost matrix, is the inspection target point T n The turning energy cost matrix between T1 and T2.

[0040] A further technical solution of the present invention is: Step S03 is described in detail as follows:

[0041] Considering the path length, cumulative radiation dose and turning energy consumption, the heuristic function of the A* algorithm is established. The A* algorithm is used to plan all paired paths in the precise optimization solution of the inspection path. The starting point and the end point of the paired path are named the starting inspection target point and the end inspection target point respectively. According to the cost values ​​of the three dimensions involved in the paired path evaluation model, the heuristic function of the A* algorithm is shown in Formula 7. At this point, the construction of the lower-level algorithm is completed, and the lower-level algorithm is run multiple times to obtain the specific routes and three-dimensional cost matrices of all paired paths in the precise optimization solution of the inspection path.

[0042] Formula 7:

[0043] In the formula, (g(p)+h(p)) is equivalent to L(P) in formula 2, and both are used to represent the path length cost between two inspection target points; R(p 1-m ) is equivalent to R(P) in formula 2, and both are used to represent the cumulative radiation dose cost between two inspection target points; E(p 1-m ) is equivalent to E(P) in Formula 2, and both are used to represent the turning energy consumption cost between two inspection target points; f(x) is the cost value of the three dimensions between the starting inspection target point and the end inspection target point; g(p) is the cost value of the path length from the starting inspection target point to the current grid, h(p) is the cost value of the path length from the current grid to the end inspection target point, (x1, y1) is the coordinate of the grid p1 where the starting inspection target point is located, (x i ,y i ) is the coordinate of the current grid pi, (x m ,y m ) is the coordinate of the grid pm where the terminal inspection target point is located.

[0044] A further technical solution of the present invention is: the A* algorithm in step S03 also includes a high radiation dose rate danger zone avoidance strategy;

[0045] During the path search process, when the ant perceives that the radiation dose rate of the next grid exceeds the tolerance limit of the mobile robot, the radiation dose rate cost value of the next grid is defined as infinity, prompting the ant to find other passable grids, refer to formula 8;

[0046] Formula 8:

[0047] Among them, P(pi) is the radiation dose rate substitute value of grid pi, Inf is infinity, and RImax is the maximum radiation dose rate that the mobile robot can withstand.

[0048] A further technical solution of the present invention is: Step S04 is described in detail as follows:

[0049] The “inspection path precise optimization solution” obtained by running the upper algorithm each time is marked as A1, A2, A3…Az in sequence; the “three-dimensional cost matrix of all paired paths in the inspection path precise optimization solution” obtained by running the lower algorithm each time is marked as C1, C2, C3…Cz in sequence; the result A1 obtained by running the upper algorithm for the first time and the result C1 obtained by running the lower algorithm for the first time are substituted into formula 1 to calculate the total cost value of the three dimensions corresponding to A1, which is recorded as λ1;

[0050] Take A1 as the initial population of the GA algorithm, and repeat steps S02-S03 to obtain A2 and C2; substitute A2 and C2 into formula 1 to calculate the total cost of the three dimensions corresponding to A2, recorded as λ2; the first iteration is completed; compare λ1 and λ2, if λ2≥λ1, then A1 corresponding to λ1 is the global optimal solution of the inspection path, if λ2<λ1, then A2 is put into the next iteration;

[0051] Take A2 as the initial population of the GA algorithm, repeat steps S02-S03, and obtain A3 and C3; substitute A3 and C3 into formula 1 to calculate the total cost of the three dimensions corresponding to A3, recorded as λ3; the second iteration is completed; compare λ2 and λ3, if λ3≥λ2, then A2 corresponding to λ2 is the global optimal solution of the inspection path, if λ3<λ2, then A3 is put into the next iteration;

[0052] This process repeats itself until the global optimal solution for the inspection path is finally obtained.

[0053] A further technical solution of the present invention is: the grids in the grid map are square; the radiation field information contained in each grid in the grid map is the average radiation dose rate recorded in each grid, and the obstacle information contained in each grid in the grid map is in each grid with binary codes "1" and "0" to indicate "no passage" and "allowed passage" respectively.

[0054] Compared with the prior art, the present invention has the following advantages:

[0055] 1. Aiming at the MTIPP (multi-target point inspection path planning) problem of mobile robots in indoor nuclear environments, the paper comprehensively considers the path length, cumulative radiation dose and turning energy consumption, and establishes a double-layer mathematical model and a double-layer hybrid algorithm. The inspection path planned by the paper has better safety and lower global cost than the path planning in the existing technology that only considers a single factor.

[0056] 2. In the two-layer mathematical model, the upper model (inspection path evaluation model) is used to quantitatively evaluate the quality of the inspection target point traversal sequence, and the lower model (paired path cost calculation model) proposes a calculation method for the three-dimensional cost values ​​of the paired paths. In the two-layer hybrid algorithm, the upper algorithm (ACO-GA hybrid algorithm) is used to determine the optimal inspection target point traversal sequence (inspection path), and the lower algorithm (A* (improved)) is used to calculate the three-dimensional cost matrix (cost map) of the paired paths; the upper and lower algorithms are iteratively run (loop operation), and the cost map calculated by the lower algorithm is returned to the upper algorithm to guide the upper algorithm to find a lower-cost inspection path until the upper algorithm converges, that is, the global optimal inspection path is obtained.

[0057] 3. The A* (improved) algorithm introduces multiple constraints in the heuristic function. Compared with the traditional A* algorithm, the total cost of planning paired paths is lower, which makes it more suitable for solving the MTIPP problem of mobile robots in indoor nuclear environments.

[0058] 4. The ACO-GA hybrid algorithm uses the rough optimization solution of the inspection path obtained by the GA algorithm to process the initial pheromone non-uniformity in the ACO algorithm, thereby improving the convergence of the ACO algorithm in the process of optimizing the traversal order of inspection target points. The ACO algorithm heuristic function, selection probability function and pheromone update rule are designed with comprehensive consideration of the path length, cumulative radiation dose and turning energy consumption, giving it dual advantages in solution quality and convergence speed, and is especially suitable for solving the MTIPP problem of mobile robots in indoor nuclear environments.

[0059] The present invention is further described below in conjunction with the figures and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 It is a technical route flow chart of the present invention;

[0061] Figure 2 Schematic diagram of the description of the MTIPP problem for mobile robots in indoor nuclear environments;

[0062] Figure 3 Schematic diagram of the eight-direction search strategy adopted by the mobile robot;

[0063] Figure 4This is a comparison chart before and after the initial pheromone non-uniformity processing in the ACO algorithm;

[0064] Figure 5 This is a box-line comparison chart of the four algorithms in scenario 1 under the shortest path;

[0065] Figure 6 This is a box-line comparison diagram of the four algorithms in scenario 1 under the minimum radiation dose path;

[0066] Figure 7 This is a box-line comparison diagram of the four algorithms in scenario 1 under the minimum turning energy consumption path;

[0067] Figure 8 This is a box-line comparison diagram of the four algorithms in scenario 1 under the paths formed by considering the three constraints;

[0068] Fig. 9 This is a box-line comparison chart of the four algorithms in scenario 2 under the shortest path;

[0069] Fig.10 This is a box-line comparison diagram of the four algorithms in scenario 2 under the minimum radiation dose path;

[0070] Fig.11 This is a box-line comparison diagram of the four algorithms in scenario 2 under the minimum turning energy consumption path;

[0071] Fig.12 This is a box-line comparison diagram of the four algorithms in scenario 2 under the paths formed by considering the three constraints;

[0072] Fig.13 The inspection path diagram obtained by using ACO-GA-A*, GA-ACO-A* and GA-A* algorithms in scenario 1;

[0073] Fig.14 The optimal inspection path diagram obtained by using ACO-A* in scenario 1;

[0074] Fig.15 This is the inspection path diagram obtained by using the ACO-GA-A* algorithm in scenario 2;

[0075] Fig.16 This is the inspection path diagram obtained by using the GA-ACO-A* and GA-A* algorithms in scenario 2;

[0076] Fig.17 This is the inspection path diagram obtained by using the ACO-A* algorithm in scenario 2;

[0077] Fig.18 It is a comparison chart of the convergence curves of ACO-GA-A*, GA-ACO-A*, GA-A* and ACO-A* algorithms when they obtain the optimal solution in scenario 1;

[0078] Fig.19 Comparison of the convergence curves of ACO-GA-A*, GA-ACO-A*, GA-A* and ACO-A* algorithms when obtaining the optimal solution in scenario 2. DETAILED DESCRIPTION

[0079] Embodiment 1:

[0080] A mobile robot indoor nuclear environment multi-target point inspection path planning method, wherein the map of the indoor nuclear environment is prior information. The map is in the form of a grid map, and the grids in the grid map are squares; each grid in the grid map contains radiation field information and obstacle information, the radiation field information is the average radiation dose rate recorded in each grid, and the obstacle information is in each grid with binary codes "1" and "0" to indicate "not passable" and "allowed to pass" respectively.

[0081] Combination Figure 2-3 Describe the MTIPP problem for mobile robots in indoor nuclear environments. Figure 2 The grid map of the indoor nuclear environment is displayed in the figure. In order to display the radiation field information more intuitively, different colors are used to distinguish the average radiation dose rates in different numerical ranges (the average radiation dose rate gradually increases according to the change rule of lake blue-beige-bright yellow-pink-dark red). In order to display the obstacle information more intuitively, black is used to indicate that the grid is "inaccessible", and all other grids except black are "allowed to pass". Figure 2 The CPC marked 11 inspection target points (T1, T2, T3…T 11 ), the 11 inspection target points are connected in sequence according to a specific traversal order to form a Hamiltonian circuit. Figure 3 It is shown that the mobile robot uses an eight-direction search strategy for path search, that is, when the mobile robot moves from the current grid point to the adjacent grid, it can only move along the straight line path of "current grid center point-adjacent grid center point".

[0082] like Figure 1 As shown, the method is as follows:

[0083] S01, mathematical modeling of multi-target inspection paths:

[0084] A. Establish an inspection path quality evaluation model: The multi-target inspection task requires the establishment of a Hamiltonian circuit that passes through each inspection target point only once (the starting and ending inspection target points are the same point, which is equivalent to passing through twice, so it is the only exception) and finally returns to the starting point. Considering the three dimensions of path length, cumulative radiation dose, and turning energy consumption, an inspection path quality evaluation model is established (the model aims to find the minimum value), or the inspection target point traversal order quality evaluation model, see formula 1;

[0085] Formula 1:

[0086] Where λ is the total cost of the three dimensions of all paired paths in the inspection path (the path between two consecutive inspection target points is called a paired path); T is the set of inspection target points, T = {T1, T2, T3, …, Tn}, and the subscript n is the total number of inspection target points; is the inspection target point T i and T j The path length cost matrix between is the inspection target point T i and T j The cumulative radiation dose cost matrix between is the inspection target point T i and T j The turning energy cost matrix between ,j is a decision variable. If the value is 1, it means that the inspection target point T is selected. i To the inspection target point T j If the value is 0, it means that the inspection target point T is not selected. i To the inspection target point T j The path between them; ω1 is the weight coefficient of the importance of path length, ω2 is the weight coefficient of the importance of cumulative radiation dose, and ω3 is the weight coefficient of the importance of turning energy.

[0087] B. Establish a pairwise path cost calculation model: T i and T j are any two inspection target points, let T i to T j The set of grids that the path with the smallest total cost value in the three dimensions passes through in sequence is P, P = {p1, p, p3…p m}, then T i The grid where it is located is p1, then T j The grid is p m ; Considering the three dimensions of path length, cumulative radiation dose and turning energy consumption, a pairwise path cost calculation model is established (the model aims to find the minimum value), as shown in Formula 2;

[0088] Formula 2:

[0089] Among them, M(P) is a paired path (the path between two inspection target points is called a paired path) T i to T j The total cost of the three dimensions, L(P) is T i to T j The path length cost value, R(P) is T i to Tj The cumulative radiation dose cost of T i to T j The turning energy consumption cost value; ε1 is the weight coefficient of the steering angle variable, ε2 is the weight coefficient of the number of turns, l(pa-1,pa) is the grid p α-1 With grid p α The center point of the line, l(pa,pa +1 ) is the grid p α With grid p α+1 The center point of the line, ψ[l(pa -1 ,pa),l(pa,pa +1 )] is the line segment l(pα -1 ,pa) and line segment l(pa,pa +1 ), Nturn is the number of turns; (xα, yα) is the grid p α The (map coordinate system) coordinates, (xα +1 ,yα +1 ) is the grid p α+1 The (map coordinate system) coordinates, (xα -1 ,yα -1 ) is the grid p α-1 (Map coordinate system) coordinates; R(pα,pα +1 ) is the mobile robot from the grid p a Travel to grid p a+1 The cumulative radiation dose received, is the average radiation dose rate of grid pα, is the grid pα +1 The average radiation dose rate, l is the average radiation dose rate of the grid from pα to pα +1 If the distance between grids pα and pα +1 If the grids pα and pα are adjacent, then l is equal to 1 times the grid side length. +1 End point contact is equal to times the grid side length, v is the distance from grid pα to grid pα +1 The speed (the initial speed of the mobile robot at the position of grid pα and the speed of grid pα +1 The final velocity of the position is consistent, and the speed of the mobile robot during the movement process is also constant). The average radiation dose rate of a grid refers to: the average radiation dose rate in the three-dimensional space formed by the height of 1m above the ground where the grid is located.

[0090] S02, obtain the inspection path based on the upper layer algorithm:

[0091] A. Use GA algorithm to obtain the rough optimization solution of the inspection path:

[0092] The chromosome in the GA algorithm corresponds to the inspection target point traversal order (i.e., candidate solution), the gene in the GA algorithm corresponds to the inspection target point (i.e., optimization variable), and the fitness in the GA algorithm corresponds to the inspection target point traversal order evaluation function (see formula 1); the GA algorithm process is as follows:

[0093] a1, real number coding: the inspection target points are deployed in sequence according to the numbers 1, 2, 3…n, indicating that the mobile robot should traverse n inspection target points and randomly generate a traversal order of inspection target points as the initial population;

[0094] a2, create a taboo list: for each true gene code, the selected taboo list is marked as 1, and the unselected taboo list is marked as 0;

[0095] a3, population initialization: set the evolutionary generation counter G and the maximum evolutionary generation G max , randomly generate the initial population;

[0096] a4, individual evaluation: evaluate the traversal order of all inspection target points according to the Hamiltonian circuit quality evaluation function (see formula 1);

[0097] a5, individual selection: select the traversal order of multiple inspection target points with the smallest generation value as the operator, apply it to the existing population, so that individuals with good genes are retained for the next generation;

[0098] a6, crossover operation: according to the preset crossover probability P c Select individuals from the existing population and generate new offspring individuals through the crossover operator, thus generating a new inspection target point traversal order;

[0099] a7, evolution termination judgment: when G≤G max Repeat steps a3 to a6; when G>G max When , the evolution terminates, and the individual with the smallest generation value in the population is the rough optimization solution of the inspection path, which is recorded as path GA .

[0100] B. Treatment of the initial pheromone inhomogeneity in the ACO algorithm: The rough optimization solution of the inspection path obtained by the GA algorithm is used to treat the initial pheromone inhomogeneity in the ACO algorithm. The treatment method is shown in Formula 3. The treatment effect is shown in Figure 4 As shown; Formula 3 aims to improve the convergence of the ACO algorithm in the process of optimizing the traversal order of inspection target points by guiding the non-uniform pheromone;

[0101] Formula 3:

[0102] In the formula, τ 0 (i,j) represents the inspection target point T iWith T j The initial pheromone concentration after the uneven treatment between the inspection target point T i With T j is a paired path, C is the initial pheromone weight value, τ is the set initial pheromone concentration (recommended to be set to 0.5), and st is the constraint condition; Formula 3 means that if the inspection target point T i and T j Belong to path GA The inspection target point in the inspection target point is T, then the initial pheromone weight value is added on the basis of the initial pheromone concentration as the initial pheromone concentration after the uneven processing. i and T j Not in path GA For the inspection target point in the image, the set initial pheromone concentration of 0.5 is used as the initial pheromone value after the uneven processing.

[0103] C. Optimize the heuristic function of the ACO algorithm: Consider the impact of path length, cumulative radiation dose and turning energy consumption on the heuristic function, and introduce an adaptive adjustment coefficient to balance the convergence speed and global search capability of the ACO algorithm to construct the heuristic function of the ACO algorithm, as shown in Formula 4;

[0104] Formula 4:

[0105] Where η t (i, j) is the tth iteration from the inspection target point T i to T j The heuristic function value of I is as follows: the larger the value, the better the corresponding path; the smaller the value, the worse the corresponding path; c is the current iteration number, I max is the maximum number of iterations; ξ is the adaptive adjustment coefficient; ω1 is the weight coefficient of the path length cost value, ω2 is the weight coefficient of the cumulative radiation dose cost value, and ω3 is the weight coefficient of the turning energy consumption cost value.

[0106] From the inspection target point T i to T j The path length cost matrix, From the inspection target point T i to T j The cumulative radiation dose cost matrix, From the inspection target point T i to T j The turning energy cost matrix;

[0107] D. Optimize the transition probability function of the ACO algorithm: Consider the pheromone concentration and the heuristic function comprehensively, and construct the transition probability function of the ACO algorithm, as shown in Formula 5;

[0108] Formula 5:

[0109] Where P t (i, j) is the tth iteration from the inspection target point T i to T j The probability of selection; α is the pheromone importance factor, β is the heuristic function importance factor; τ t (i,j) indicates the tth iteration from the inspection target point T i to T j The pheromone concentration of t (i,u) represents the tth iteration from the inspection target point T i to T u The pheromone concentration of T t ′ is the set of inspection target points that have not been traversed in the tth iteration; the denominator in Formula 5 is the number of inspection target points T from the tth iteration. i The sum of the selection probabilities of all untraversed inspection target points; if the inspection target point T in formula 5 j is the inspection target point that has not been traversed, and then T i to T j The selection probability calculation is as follows: if the inspection target point T j Has been traversed, then T i to T j The probability of selection is 0.

[0110] E. Optimize the pheromone update rule of the ACO algorithm: Consider the path length, the cumulative radiation dose (suffered by the mobile robot), and the turning energy consumption, and optimize the pheromone update rule of the ACO algorithm, as shown in Formula 6. This completes the construction of the upper-level algorithm (ACO-GA hybrid algorithm), and runs the upper-level algorithm to obtain the precise optimization solution of the inspection path.

[0111] Formula 6:

[0112] In the formula, τt +1 (i, j) indicates the t+1th iteration from the inspection target point T i to T j The pheromone concentration of (this path), τt(i,j) represents the t-th iteration from the inspection target point T i to T j (this path) is the pheromone concentration, ρ is the pheromone volatility factor, Q is the pheromone deposition factor, υ t-mi is the total cost of the three dimensions of the optimal inspection path of the tth iteration (υ value is the smallest), υ t-ma is the total cost of the three dimensions of the worst inspection path in the tth iteration (υ value is the largest); I tis the optimal inspection path of the tth iteration; υ is the total cost of the three dimensions of the inspection path; assuming that there are n inspection target points in the inspection path I, then “Ti,Ti +1 " are two consecutive inspection target points in inspection path I, then "Tn, T1" are two consecutive inspection target points in inspection path I. Inspection path I requires that each inspection target point is passed only once (the starting point and the end point inspection target point are the same point, which is equivalent to passing twice, so it is the only exception) and finally returns to the starting point; is the inspection target point T i and T i+1 The path length cost matrix between is the inspection target point T i and T i+1 The cumulative radiation dose cost matrix between is the inspection target point T i and T i+1 The turning energy cost matrix between is the inspection target point T n The path length cost matrix between and T1, is the inspection target point T n and the cumulative radiation dose cost matrix between T1, is the inspection target point T n The turning energy cost matrix between T1 and T2.

[0113] S03, construct the lower layer algorithm to calculate the paired paths:

[0114] Considering the path length, cumulative radiation dose and turning energy consumption, the heuristic function of the A* (improved) algorithm is established, and all paired paths in the inspection path precision optimization solution are planned by the A* (improved) algorithm. The starting point and the end point of the paired path are named as the starting inspection target point and the end inspection target point respectively; according to the three-dimensional cost value involved in the paired path quality evaluation model (Formula 2), the heuristic function of the A* (improved) algorithm is shown in Formula 7; the construction of the lower-level algorithm (A* (improved) algorithm) is completed, and the lower-level algorithm is run multiple times to obtain the specific routes and three-dimensional cost matrices of all paired paths in the inspection path precision optimization solution;

[0115] Formula 7:

[0116] In the formula, (g(p)+h(p)) is equivalent to L(P) in formula 2, and both are used to represent the path length cost between two inspection target points; R(p 1-m ) is equivalent to R(P) in formula 2, and both are used to represent the cumulative radiation dose cost between two inspection target points; E(p 1-m) is equivalent to E(P) in Formula 2, and both are used to represent the turning energy consumption cost between two inspection target points; f(x) is the cost value of the three dimensions between the starting inspection target point and the end inspection target point; g(p) is the cost value of the path length from the starting inspection target point to the current grid, h(p) is the cost value of the path length from the current grid to the end inspection target point, (x1, y1) is the coordinate of the grid p1 where the starting inspection target point is located, (x i ,y i ) is the coordinate of the current grid pi, (x m ,y m ) is the coordinate of the grid pm where the final inspection target point is located (the subscript m indicates that there are m grids between the starting inspection target point and the final inspection target point).

[0117] S04, the two-layer algorithm runs in a loop to obtain the global optimal solution:

[0118] The "inspection path precise optimization solution" obtained by running the upper algorithm each time is marked in sequence as A1, A2, A3...Az; the "three-dimensional cost matrix of all paired paths in the inspection path precise optimization solution" obtained by running the lower algorithm each time is marked in sequence as C1, C2, C3...Cz; substitute the result A1 obtained by the first run of the upper algorithm and the result C1 obtained by the first run of the lower algorithm into formula 1, and calculate the total cost value of the three dimensions corresponding to A1, which is recorded as λ1.

[0119] Take A1 as the initial population of the GA algorithm, and repeat steps S02-S03 to obtain A2 and C2; substitute A2 and C2 into formula 1 to calculate the total cost of the three dimensions corresponding to A2, recorded as λ2; the first iteration is completed; compare λ1 and λ2, if λ2≥λ1, then A1 corresponding to λ1 is the global optimal solution of the inspection path, if λ2<λ1, then A2 is put into the next iteration.

[0120] Take A2 as the initial population of the GA algorithm, repeat steps S02-S03, and obtain A3 and C3; substitute A3 and C3 into formula 1 to calculate the total cost of the three dimensions corresponding to A3, recorded as λ3; the second iteration is completed; compare λ2 and λ3, if λ3≥λ2, then A2 corresponding to λ2 is the global optimal solution of the inspection path, if λ3<λ2, then A3 is put into the next iteration.

[0121] This process repeats itself until the global optimal solution for the inspection path is finally obtained.

[0122] As a preferred solution, in step S03, the A* (improved) algorithm also includes a high radiation dose rate danger zone avoidance strategy; when the mobile robot passes through a grid area where the radiation dose rate exceeds its tolerance limit, it will be damaged by radiation and even cause component failure, so it is necessary to avoid such grid areas; in the path search process, when the ant perceives that the radiation dose rate of the next grid (the current grid) exceeds the tolerance limit of the mobile robot, the radiation dose rate cost value of the next grid is defined as infinity, prompting the ant to find other passable grids, refer to formula 8;

[0123] Formula 8:

[0124] Wherein, R(pi) is the radiation dose rate cost value of grid pi, Inf is infinity, and RImax is the maximum radiation dose rate that the mobile robot can withstand. Formula 8 indicates that if the radiation dose rate of grid pi does not exceed the maximum radiation dose rate that the mobile robot can withstand, the radiation dose rate of grid pi will not be changed. If the radiation dose rate of grid pi does not exceed the maximum radiation dose rate that the mobile robot can withstand, the radiation dose rate of grid pi is defined as infinity.

[0125] Technical effect analysis of the present invention: ACO-GA-A* (a two-layer hybrid algorithm, i.e., an algorithm composed of an upper-layer algorithm and a lower-layer algorithm) is compared with other algorithms in two indoor kernel environments of different complexity to verify the performance of ACO-GA-A* in terms of sensitivity, reliability and convergence.

[0126] Ⅰ. Sensitivity analysis:

[0127] When solving the MTIPP problem, three constraints, namely path length, risk level and turning energy consumption, are considered comprehensively. Therefore, it is necessary to verify whether the optimal path preferred by the user can be found by adjusting the cost coefficient weights (ω1, ω2, ω3). To this end, according to the actual working conditions of the mobile robot in the nuclear environment, the sensitivity of the ACO-GA-A* two-layer hybrid algorithm and other algorithms under four sets of weights are compared, where the coefficient weight combination (1, 0, 0) represents the shortest path; the coefficient weight combination (0, 1, 0) represents the minimum cumulative radiation dose path; the coefficient weight combination (0, 0, 1) represents the minimum turning energy consumption path; the coefficient weight combination (1, 1, 1) represents the optimal path that considers the three constraints at the same time. The comparison algorithms under different coefficient weight combinations were run 20 times in two indoor nuclear environment scenarios (scenario 1 and scenario 2) of different complexity. The box-line comparison diagram of different algorithms under different preferred paths in scenario 1 was obtained, as shown in the figure below. Figure 5-8 The box line comparison diagram of different algorithms under different preference paths in scenario 2 is obtained, as shown in Figure 9-12 shown.

[0128] analyze Figure 5-12 It can be seen that: 1. ACO-GA-A* is better than other algorithms in finding the shortest path in different nuclear environment scenarios, and the solution set is relatively concentrated (small deviation). 2. ACO-GA-A* is better than other algorithms in finding the path with the minimum cumulative radiation in different nuclear environment scenarios, and the solution set is relatively concentrated (small deviation). 3. When considering the three constraints, the total cost of the optimal inspection path planned by ACO-GA-A* in different scenarios is less than that of other hybrid algorithms, indicating that it has obvious advantages in solution quality. The above analysis shows that changes in weight coefficients and environmental complexity will not affect the effectiveness of ACO-GA-A*.

[0129] Ⅱ. Reliability Analysis

[0130] The reliability analysis of ACO-GA-A* was conducted based on two simulated indoor nuclear environment scenarios with different complexity (Scenario 1 and Scenario 2). The weight coefficient combination of the objective function was determined by the single-objective standardization method. The simulation was performed 15 times in Scenario 1 and Scenario 2 respectively, and the reliability data of different algorithms sampled in Scenario 1 and Scenario 2 were obtained, as shown in Table 1. The optimal inspection path diagram obtained by different algorithm planning in Scenario 1 is shown in Table 1. Figure 13-14 As shown (because the optimal inspection paths planned by ACO-GA-A*, GA-ACO-A* and GA-A* in scenario 1 are consistent, Fig.13 is the optimal inspection path diagram corresponding to the above three algorithms, Fig.14 The optimal inspection path diagram planned by ACO-A* in scenario 1 is shown in Figure 2. The optimal inspection path diagram planned by different algorithms in scenario 2 is shown in Figure 2. Figure 15-17 As shown (since the optimal inspection paths planned by GA-ACO-A* and GA-A* in scenario 2 are consistent, Fig.16 is the optimal inspection path diagram corresponding to the above two algorithms, Fig.15 and Fig.17 They are the optimal inspection path diagrams planned by ACO-GA-A* and ACO-A* in scenario 2).

[0131] Table 1: Reliability data of different algorithms in different scenarios

[0132]

[0133] As can be seen from Table 1, in nuclear environment scenarios of different complexity, the average total cost and standard deviation of the paths planned by ACO-GA-A* are significantly better than other algorithms, and the more complex the environment, the more obvious the advantage. The probability of obtaining the optimal path in scenarios 1 and 2 is 100% and 30% respectively, which is significantly higher than other algorithms. It can be seen that the ACO-GA-A* algorithm has higher reliability and robustness in solving the MTIPP problem than other algorithms. In terms of calculation time, ACO-GA-A* and GA-ACO-A* take about the same time, which is slightly longer than ACO-A* and GA-A*. The reason is that optimizing the initial pheromone distribution of ACO through GA or optimizing the initial population of GA through ACO requires more time cost. However, considering that the MTIPP problem is an offline path planning performed by the mobile robot before entering the inspection site, it is acceptable to obtain a better path with a small amount of time cost.

[0134] exist Figure 13-17 In order to more intuitively display the radiation field information, different colors are used to distinguish the average radiation dose rate in different numerical ranges (the average radiation dose rate gradually increases according to the blue-yellow-red change rule). In order to more intuitively display the obstacle information, black is used to indicate that the grid is "inaccessible", and all grids except black are "allowed to pass". Figure 13-17 It can be seen that the optimal paths planned by ACO-GA-A* in different nuclear environment scenarios have good traversal order, no path intersection, and none of the formed paths passes through the high radiation danger zone (red area).

[0135] III. Convergence Analysis

[0136] Convergence curve comparison of different algorithms when obtaining the optimal solution in scenario 1, as shown in Fig.18 As shown, Fig.18 The upper right corner is Fig.18 The comparison of the convergence curves of different algorithms when obtaining the optimal solution in scenario 2 is shown in Figure 2. Fig.19 As shown, Fig.19 The upper right corner shows the algorithms represented by lines of different colors ( Fig.18 Same as above).

[0137] from Figure 18-19It can be seen that the ACO-GA-A* algorithm converges to the optimal solution faster, while the other algorithms converge relatively slowly and cannot plan the inspection path with the minimum total cost. As the complexity of the environment increases, the performance of the ACO-A* algorithm decreases rapidly, especially in terms of the optimal solution. In addition, as the complexity of the environment increases, the number of iterations of the GA-A* algorithm increases most significantly. Comparing the convergence curves of ACO-GA-A* and ACO-A*, it can be confirmed that the enhancement of the initial pheromone guidance of ACO by GA has a significant effect on improving the convergence speed. Comparing the convergence of GA-ACO-A* and GA-A*, it can be seen that the initial population optimization of GA based on ACO significantly improves the convergence speed of GA. In summary, the ACO-GA-A* algorithm is easier to find the optimal solution.

Claims

1. A mobile robot indoor nuclear environment multi-target point inspection path planning method, which is characterized by: The grid map of the indoor nuclear environment is prior information, and each grid in the grid map contains radiation field information and obstacle information; Here’s how: S01, mathematical modeling of multi-target inspection paths: considering the path length, cumulative radiation dose and turning energy consumption, the inspection path quality evaluation model and paired path quality evaluation model are established respectively; S02, obtaining the inspection path based on the upper layer algorithm: using the GA algorithm to obtain a rough optimization solution for the inspection path; The rough optimization solution of the inspection path is used to deal with the initial pheromone inhomogeneity in the ACO algorithm; considering the path length, cumulative radiation dose and turning energy consumption, the heuristic function and selection probability function of the ACO algorithm are constructed to optimize the pheromone update rule of the ACO algorithm; the upper-level algorithm is constructed and run to obtain the precise optimization solution of the inspection path; S03, construct the lower-level algorithm to calculate the paired paths: consider the path length, cumulative radiation dose and turning energy consumption, and establish the heuristic function of the A* algorithm; complete the construction of the lower-level algorithm, run the lower-level algorithm, and plan all paired paths in the inspection path optimization solution; S04, the two-layer algorithm is cyclically run to obtain the optimal solution: the inspection path quality evaluation model is used to compare the quality of the inspection path optimization solution obtained in the current iteration with that obtained in the previous iteration. If the inspection path optimization solution obtained in the current iteration is better, the inspection path optimization solution obtained in the current iteration is used in the next iteration. If the inspection path optimization solution obtained in the current iteration is worse or consistent, the iteration is terminated and the inspection path optimization solution obtained in the previous iteration is output.

2. The method for mobile robot indoor nuclear environment multi-target point inspection path planning as claimed in claim 1 is characterized by: The S01 step is described in detail as follows: A. Establish an inspection path quality evaluation model: The multi-target inspection task requires the establishment of a Hamiltonian circuit that passes each inspection target point only once and eventually returns to the starting point. Considering the three dimensions of path length, cumulative radiation dose, and turning energy consumption, an inspection path quality evaluation model, or an inspection target point traversal sequence quality evaluation model, is established. See Formula 1. Formula 1: Among them, λ is the total cost of the three dimensions of all paired paths in the inspection path; T is the set of inspection target points, T = {T1, T2, T3, …, T n }, the subscript n is the total number of inspection target points; is the inspection target point T i and T j The path length cost matrix between is the inspection target point T i and T j The cumulative radiation dose cost matrix between is the inspection target point T i and T j The turning energy cost matrix between i,j is a decision variable. If the value is 1, it means that the inspection target point T is selected. i To the inspection target point T j If the value is 0, it means that the inspection target point T is not selected. i To the inspection target point T j The path between them; ω1 is the weight coefficient of the path length, ω2 is the weight coefficient of the cumulative radiation dose, and ω3 is the weight coefficient of the turning energy consumption; B. Establish a pairwise path cost calculation model: T i and T j are any two inspection target points, let T i to T j The set of grids that the path with the smallest total cost value in the three dimensions passes through in sequence is P, P = {p1, p, p3…p m }, then T i The grid where it is located is p1, then T j The grid is p m ; Considering the three dimensions of path length, cumulative radiation dose and turning energy consumption, a pairwise path cost calculation model is established, as shown in Formula 2; Formula 2: Where M(P) is the pairwise path T i to T j The total cost of the three dimensions, L(P) is T i to T j The path length cost value, R(P) is T i to T j The cumulative radiation dose cost of T i to T j The turning energy consumption cost value; ε1 is the weight coefficient of the steering angle variable, ε2 is the weight coefficient of the number of turns, l(p a -1,p a ) is the grid p α-1 With grid p α The center point of the line, l(p a ,p a+1 ) is the grid p α With grid p α+1 The center point of the line, ψ[l(p a-1 ,p a ),l(p a ,p a+1 )] is the line segment l(p a-1 ,p a ) and line segment l(p a ,p a+1 ) steering angle variable, N turn is the number of turns; (x α ,y α ) is the grid p α The coordinates of (x α+1 ,y α+1 ) is the grid p α+1 The coordinates of (x α-1 ,y α-1 ) is the grid p α-1 Coordinates of R(p α ,p α+1 ) is the mobile robot from the grid p a Travel to grid p a+1 The cumulative radiation dose received, is the grid p α The average radiation dose rate, is the grid p α+1 The average radiation dose rate of the grid p α to p α+1 If the grid p α With p α+1 If the grid edges are adjacent, then l is equal to 1 times the grid edge length. α With p α+1 End point contact is equal to times the grid side length, v is the length of the grid p α Move to grid p α+1 speed.

3. The method for mobile robot indoor nuclear environment multi-target point inspection path planning as claimed in claim 2 is characterized by: The S02 step is described in detail as follows: A. Use GA algorithm to obtain the rough optimization solution of the inspection path: The chromosomes in the GA algorithm correspond to the inspection target point traversal order, the genes in the GA algorithm correspond to the inspection target points, and the fitness in the GA algorithm corresponds to the evaluation function of the inspection target point traversal order. The GA algorithm process is as follows: a1, real number coding: the inspection target points are deployed in sequence according to the numbers 1, 2, 3…n, indicating that the mobile robot should traverse n inspection target points and randomly generate a traversal order of inspection target points as the initial population; a2, create a taboo list: for each true gene code, the selected taboo list is marked as 1, and the unselected taboo list is marked as 0; a3, population initialization: set the evolutionary generation counter G and the maximum evolutionary generation G max , randomly generate the initial population; a4, individual evaluation: evaluate the traversal order of all inspection target points according to the Hamiltonian circuit quality evaluation function; a5, individual selection: select the traversal order of multiple inspection target points with the smallest generation value as the operator, apply it to the existing population, so that individuals with good genes are retained for the next generation; a6, crossover operation: according to the preset crossover probability P c Select individuals from the existing population and generate new offspring individuals through the crossover operator, thus generating a new inspection target point traversal order; a7, evolution termination judgment: when G≤G max Repeat steps a3 to a6; when G>G max When , the evolution terminates, and the individual with the smallest generation value in the population is the rough optimization solution of the inspection path, which is recorded as path GA ; B. Processing the initial pheromone inhomogeneity in the ACO algorithm: The rough optimization solution of the inspection path obtained by the GA algorithm is used to process the initial pheromone inhomogeneity in the ACO algorithm, as shown in Formula 3; Formula 3: In the formula, τ 0 (i,j) represents the inspection target point T i With T j The initial pheromone concentration after the uneven treatment between the inspection target point T i With T j is a pair of paths, C is the initial pheromone weight value, τ is the set initial pheromone concentration, and st is the constraint condition; C. Optimize the heuristic function of the ACO algorithm: Consider the impact of path length, cumulative radiation dose and turning energy consumption on the heuristic function, and introduce an adaptive adjustment coefficient to balance the convergence speed and global search capability of the ACO algorithm to construct the heuristic function of the ACO algorithm, as shown in Formula 4; Formula 4: Where η t (i, j) is the tth iteration from the inspection target point T i to T j The heuristic function value of c is the current iteration number, I max is the maximum number of iterations; ξ is the adaptive adjustment coefficient; ω1 is the weight coefficient of the path length cost value, ω2 is the weight coefficient of the cumulative radiation dose cost value, and ω3 is the weight coefficient of the turning energy consumption cost value. From the inspection target point T i to T j The path length cost matrix, From the inspection target point T i to T j The cumulative radiation dose cost matrix, From the inspection target point T i to T j The turning energy cost matrix; D. Optimize the transition probability function of the ACO algorithm: Consider the pheromone concentration and the heuristic function comprehensively, and construct the transition probability function of the ACO algorithm, as shown in Formula 5; Formula 5: Where P t (i, j) is the tth iteration from the inspection target point T i to T j The probability of selection; α is the pheromone importance factor, β is the heuristic function importance factor; τ t (i,j) indicates the tth iteration from the inspection target point T i to T j The pheromone concentration of t (i,u) represents the tth iteration from the inspection target point T i to T u The pheromone concentration of T t ′ is the set of inspection target points that have not been traversed in the tth iteration; E. Optimize the pheromone update rule of the ACO algorithm: Consider the path length, cumulative radiation dose, and turning energy consumption, and optimize the pheromone update rule of the ACO algorithm, as shown in Formula 6. This completes the construction of the upper-level algorithm, and runs the upper-level algorithm to obtain the precise optimization solution of the inspection path. Formula 6: In the formula, τ t+1 (i, j) indicates the t+1th iteration from the inspection target point T i to T j The pheromone concentration, τ t (i,j) indicates the tth iteration from the inspection target point T i to T j ρ is the pheromone concentration, Q is the pheromone deposition factor, υ is the pheromone volatilization factor, t-mi is the total cost of the three dimensions of the optimal inspection path of the tth iteration, υ t-ma is the total cost of the three dimensions of the worst inspection path in the tth iteration; I t is the optimal inspection path for the tth iteration; υ is the total cost of the three dimensions of the inspection path; assuming that there are n inspection target points in the inspection path, then "T i ,T i+1 " are two consecutive inspection target points in the inspection path, then "T n ,T1” are the two consecutive inspection target points in the inspection path. The inspection path requires that each inspection target point is passed only once and finally returns to the starting point; is the inspection target point T i and T i+1 The path length cost matrix between is the inspection target point T i and T i+1 The cumulative radiation dose cost matrix between is the inspection target point T i and T i+1 The turning energy cost matrix between is the inspection target point T n The path length cost matrix between and T1, is the inspection target point T n and the cumulative radiation dose cost matrix between T1, is the inspection target point T n The turning energy cost matrix between T1 and T2.

4. The method for mobile robot indoor nuclear environment multi-target point inspection path planning as claimed in claim 3 is characterized by: The S03 step is described in detail as follows: Considering the path length, cumulative radiation dose and turning energy consumption, the heuristic function of the A* algorithm is established. The A* algorithm is used to plan all paired paths in the inspection path optimization solution. The starting point and end point of the paired path are named the starting inspection target point and the end inspection target point respectively. According to the three-dimensional cost values ​​involved in the paired path evaluation model, the heuristic function of the A* algorithm is shown in Formula 7. At this point, the construction of the lower-level algorithm is completed, and the lower-level algorithm is run multiple times to obtain the specific routes and three-dimensional cost matrices of all paired paths in the inspection path optimization solution. Formula 7: In the formula, (g(p)+h(p)) is equivalent to L(P) in formula 2, and both are used to represent the path length cost between two inspection target points; R(p 1-m ) is equivalent to R(P) in formula 2, and both are used to represent the cumulative radiation dose cost between two inspection target points; E(p 1-m ) is equivalent to E(P) in Formula 2, and both are used to represent the turning energy consumption cost between two inspection target points; f(x) is the cost value of the three dimensions between the starting inspection target point and the end inspection target point; g(p) is the cost value of the path length from the starting inspection target point to the current grid, h(p) is the cost value of the path length from the current grid to the end inspection target point, (x1, y1) is the coordinate of the grid p1 where the starting inspection target point is located, (x i ,y i ) is the current grid p i The coordinates of (x m ,y m ) is the grid p where the target point of the terminal inspection is located m The coordinates of .

5. The method for mobile robot indoor nuclear environment multi-target point inspection path planning as claimed in claim 4 is characterized by: The A* algorithm in step S03 also includes a high radiation dose rate danger zone avoidance strategy; During the path search process, when the ant perceives that the radiation dose rate of the next grid exceeds the tolerance limit of the mobile robot, the radiation dose rate cost value of the next grid is defined as infinity, prompting the ant to find other passable grids, refer to formula 8; Formula 8: Among them, R(p i ) is the grid p i The radiation dose rate substitute value, Inf is infinite, RI max is the maximum radiation dose rate that the mobile robot can withstand.

6. The method for mobile robot indoor nuclear environment multi-target point inspection path planning as claimed in claim 5 is characterized by: The S04 step is described in detail as follows: The "inspection path precise optimization solution" obtained by running the upper algorithm each time is marked as A1, A2, A3...Az in sequence; the "three-dimensional cost matrix of all paired paths in the inspection path precise optimization solution" obtained by running the lower algorithm each time is marked as C1, C2, C3...Cz in sequence; the result A1 obtained by running the upper algorithm for the first time and the result C1 obtained by running the lower algorithm for the first time are substituted into formula 1 to calculate the total cost value of the three dimensions corresponding to A1, which is recorded as λ1; Take A1 as the initial population of the GA algorithm, and repeat steps S02-S03 to obtain A2 and C2; substitute A2 and C2 into formula 1 to calculate the total cost of the three dimensions corresponding to A2, recorded as λ2; the first iteration is completed; compare λ1 and λ2, if λ2≥λ1, then A1 corresponding to λ1 is the global optimal solution of the inspection path, if λ2<λ1, then A2 is put into the next iteration; Take A2 as the initial population of the GA algorithm, repeat steps S02-S03, and obtain A3 and C3; substitute A3 and C3 into formula 1 to calculate the total cost of the three dimensions corresponding to A3, recorded as λ3; the second iteration is completed; compare λ2 and λ3, if λ3≥λ2, then A2 corresponding to λ2 is the global optimal solution of the inspection path, if λ3<λ2, then A3 is put into the next iteration; This process repeats itself until the global optimal solution for the inspection path is finally obtained.

7. The method for mobile robot indoor nuclear environment multi-target point inspection path planning as claimed in claim 6 is characterized by: The grids in the grid map are square; the radiation field information contained in each grid in the grid map is the average radiation dose rate recorded in each grid, and the obstacle information contained in each grid in the grid map is binary coded "1" and "0" in each grid to indicate "no passage" and "passage allowed" respectively.

Images