AGV guiding path network design optimization method considering strong connectivity

Through the general variable neighborhood search algorithm and the deep search method of the graph, combined with ear decomposition and local search of the strong connectivity graph, the computational complexity problem under strong connectivity constraints in AGV-oriented path network design is solved, and efficient path planning is achieved.

CN119067550AInactive Publication Date: 2024-12-03XIANGNAN UNIV
View PDF 1 Cites 0 Cited by

Patent Information

Application Number
CN202411292787.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-09-14
Publication Date
2024-12-03
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The existing AGV-oriented path network design has high computational complexity when meeting the strong connectivity constraints, making it difficult to effectively improve path planning efficiency, especially in large-scale problems, which is a long time to solve.

Method used

The general variable neighborhood search algorithm combined with the deep search method of graphs is used to generate initial solutions through the ear decomposition sequence of the strongly connected graph, and the neighborhoods of the redirection of the sub-graph are performed on the premise of satisfying the strongly connected constraints. Action, combined with local search algorithms to improve the effectiveness of solutions.

Benefits of technology

It significantly improves the efficiency and quality of AGV-oriented path network design, can generate initial solutions in linear time, and find the optimal solutions through local search, which reduces the computational complexity and improves the efficiency of path planning.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119067550A_ABST
    Figure CN119067550A_ABST
Patent Text Reader

Abstract

An AGV guiding path network design optimization method considering strong connectivity comprises the following steps that S0, under the condition that the loading and unloading point position of each unit and the logistics quantity between the units are known and under the condition that it is ensured that a guiding path can meet the strong connectivity constraint condition, the driving direction of each AGV lane is determined; constructing a guiding path network mathematical model by taking minimization of a total transportation distance formed by an AGV load distance and a no-load distance as a target; s1, solving a guiding path of the guiding path network mathematical model by using a general variable neighborhood search algorithm; understanding and identifying a strong connectivity structure of the graph through a deep search method of the graph; s2, obtaining an initial solution of an ear decomposition sequence of the strongly connected graph according to an algorithm for generating the ear decomposition sequence of the strongly connected graph; s3, performing neighborhood reduction search, and outputting an optimal solution in the whole process; and S4, combining the optimal solution in the neighborhood reduction search step into a general variable neighborhood search algorithm to enhance the search capability. The method can improve the efficiency of path planning.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to logistics automation technology, and in particular to an AGV guide path network design optimization method considering strong connectivity. Background Art

[0002] like Figure 2 As shown in the figure, it is an intelligent workshop that relies on the AGV storage and transportation system to realize material flow. The intelligent material flow system of the workshop combines the AGV storage and transportation system with the production and manufacturing system to achieve efficient material transportation and processing flow. The production and manufacturing system consists of multiple processing areas, each of which is responsible for material processing in different processes. The AGV storage and transportation system consists of multiple unidirectional AGVs, and unidirectional tracks are laid on the logistics channels in the workshop to form an AGV track transportation network. The AGV travels unidirectionally along the track, connecting the various processing areas in the workshop to realize the flow of materials between different processes. The design of this system not only helps to reduce energy consumption and time costs, but also improves the stability and reliability of the overall system. Through effective path planning and determination of driving direction, the transportation of AGV between workstations becomes more efficient and reliable, providing strong support for the smooth progress of the production process.

[0003] The goal of the AGV one-way guidance path network design problem is to determine the driving direction of each AGV lane while ensuring that the guidance path can satisfy the strong connectivity constraints when the location of each unit loading and unloading point and the amount of logistics between units are known.

[0004] The path network design problem is an NP-hard problem. Although the 0-1 integer programming method can obtain an exact solution, the computational complexity is relatively high, which will make the model construction more difficult as the constraints increase. When solving larger-scale problems, the solution time will be longer. Other existing methods also have some shortcomings. In order to solve the path network design problem, it is necessary to develop a method that is conducive to improving the efficiency of path planning. Summary of the invention

[0005] In view of the above-mentioned defects, the purpose of the present invention is to propose an AGV guidance path network design optimization method considering strong connectivity, which can improve the efficiency of path planning.

[0006] To achieve this object, the present invention adopts the following technical solutions:

[0007] A method for optimizing the design of an AGV guided path network considering strong connectivity includes the following steps:

[0008] S0. Under the condition that the location of each unit loading and unloading point and the amount of logistics between units are known, the driving direction of each AGV lane is determined to ensure that the guidance path can meet the strong connectivity constraint condition; the total transportation distance formed by the AGV load distance and empty distance is minimized, and the guidance path network mathematical model is constructed;

[0009] S1. Use the general variable neighborhood search algorithm to obtain the guidance path of the guidance path network mathematical model; understand and identify the strong connectivity structure of the graph through the deep search method of the graph;

[0010] S2. According to the algorithm for generating the ear decomposition sequence of the strongly connected graph, the initial solution of the ear decomposition sequence of the strongly connected graph can be obtained in linear time;

[0011] S3, neighborhood search, setting neighborhood actions including reversal circles, reversal paths and subgraph redirection under the premise of satisfying strong connectivity constraints, exploring one or more neighborhoods; at the same time, introducing local search algorithms to find better solutions; if a better solution is found, updating the current solution and adjusting the search strategy as needed; after reaching the preset termination condition, outputting the optimal solution in the whole process;

[0012] S4. Combine the optimal solution in the descending neighborhood search step into the general variable neighborhood search algorithm to enhance the search capability.

[0013] Preferably, the mathematical model in step S0 is as follows:

[0014] In order to describe the problem and accurately establish the corresponding mathematical model, the following definitions are made:

[0015] There is no upper limit to AGV transport capacity;

[0016] Each unit has a unique loading and unloading port, and the position is known;

[0017] The transport paths of AGVs are all the shortest paths;

[0018] It does not consider whether the AGV will be blocked;

[0019] Where N is the number of units; S is the vertex set; P n is the feeding port, n∈[1,N]; D n is the unloading port, n∈[1,N]; E(i,j) is the lane from vertex i to vertex j; T ij is the lane direction; H ij is the lane length; F bd From the feed port D b To the feeding port P d Number of handling times; G db From the feeding port P d To the feeding port D bNo-load times; L ij is the shortest path from vertex i to vertex j; Is lane E(i,j) located at the unloading port D? b To the feeding port P d on the transport route; Is lane E(i,j) located at the loading port P? d To the feeding port D b On the empty route; q bd is the load flow matrix between unit b and unit d; is the no-load flow matrix between unit b and unit d;

[0020]

[0021] T ij +T ji =1 Formula (2.12)

[0022] Σ i∈S T ij ≥1 Formula (2.13)

[0023] Σ j∈S T ij ≥1 Formula (2.14)

[0024] Σ d∈N F bd =Σ d∈N G db Formula (2.15)

[0025]

[0026] Formula (2.1) of the mathematical programming model is the objective function. The optimization goal is to make the total running distance of the AGV as short as possible, and the total running distance of the AGV includes two parts: the empty distance and the loaded distance. Constraint formulas (2.2) and (2.3) ensure that the length of the route selected by the calculation is equal to the empty distance, that is, from the loading port P b To the feeding port D d , from the feed port D d To the feeding port P b The empty route distance is equal to the selected lane length and;

[0027] Constraints (2.4)-(2.5) are to ensure that there is only one inlet and outlet for the feeder P. b Lanes;

[0028] Constraints (2.6)-(2.7) are to ensure that the inlet and outlet D d The number of lanes is only one each;

[0029] Constraints (2.8)-(2.9) are to ensure that the number of times the AGV enters all non-loading and unloading point vertices is consistent with the number of times it leaves the vertex;

[0030] Constraints (2.10)-(2.11) are to ensure that the lanes in each empty transport route will not be repeated;

[0031] The constraint (2.12) is to ensure that each lane is one-way;

[0032] Constraints (2.13)-(2.14) are to ensure that the guidance path network is strongly connected. It is necessary to ensure that for any vertex in the vertex set, there is a guidance path starting from and ending at the vertex;

[0033] The role of constraints (2.15)-(2.16) is to ensure logistics balance, that is, to ensure that the number of times the AGV is loaded is equal to the number of times it is empty;

[0034] In addition, constraint (2.17) is a variable constraint, ensuring that each variable in it is a 0-1 variable; from the mathematical model, it contains a large number of 0-1 variables and constraints, which will cause the solution space to grow exponentially with the increase in the scale of decision variables.

[0035] Preferably, in step S2:

[0036] The ear decomposition of a directed graph is to decompose it into a set of paths and cycles, and there are no common edges between all paths and cycles. Its specific definition is as follows: For a directed multigraph D, its ear decomposition is a sequence ε = {P 0 ,P 1 ,…P t}, where P 0 Represents a circle, each P i (1≤i≤t) is a path or a cycle with the following properties

[66] :

[0037] (1) When i≠j, P i and P j The arcs do not intersect;

[0038] (2) For each i = 1, 2, ..., t, if P i is a circle, then P i With V(D i-1 ) have only one common vertex; otherwise, P i The endpoint is V(D i-1 ) are not the same vertices, and P i The remaining vertices of V(D i-1 ); here, D i Indicates that it has a vertex set and arc set The directed graph formed;

[0039] (3)

[0040] Each P i is called an ear of ε, the number of ears in ε is t+1, if an ear |A(P j )|=1, it is called ordinary ear; otherwise it is called extraordinary ear;

[0041] Theorem 1: A directed graph D = (V, A) is a strongly connected graph if and only if it has at least one ear decomposition sequence; in addition, if D is a strongly connected graph, for each vertex v, each cycle C containing v can be used as the initial cycle P of the ear decomposition sequence of D. 0 ;

[0042] From Theorem 1, we know that every strongly connected directed graph D has an ear decomposition sequence. Selecting different vertices v can get different ear decomposition sequences of D. Obviously, every strongly connected directed graph D has multiple ear decomposition sequences, but the ear decomposition sequences of two different strongly connected directed graphs must be different.

[0043] Corollary 1: For a strongly connected directed graph with n vertices and m arcs, each ear decomposition sequence contains m-n+1 ears, and there exists a linear time algorithm that can find the ear decomposition of a strongly connected directed graph D;

[0044] Through the proof process of Theorem 1, it is easy to prove that each ear decomposition sequence contains m-n+1 ears; we can use a depth-first search algorithm based on the directed graph D to construct an ear in each backtracking process; since the entire backtracking only traverses each vertex and arc once, the algorithm is a linear time algorithm.

[0045] Preferably, in step S2:

[0046] By using the graph depth search method, we can determine the strongly connected directed graph D of the undirected graph G. Then, according to the algorithm for generating the ear decomposition sequence, we can find the ear decomposition sequence ε of the strongly connected directed graph D in linear time.

[0047] First, the workshop AGV transport track network to be directed is modeled as an undirected graph G, and then the strongly connected directed graph D of G is determined, and finally it is converted into an ear decomposition sequence ε. In this way, an initial solution can be generated within a linear time range.

[0048] Further, in step S3:

[0049] For the strong connectivity constraint, we design neighborhood actions that satisfy this constraint from the perspective of graph theory;

[0050] Reversal circle:

[0051] Theorem 2: If D' and D" are two k-strongly connected orientations of an undirected graph G = (V, E), then there exists a set of k-strongly connected orientations D' = D 0 ,D 1 ,…,D k =D", so that every D i (i=1,2,…,k) can be obtained by reversing a directed loop or path. i-1 ;

[0052] If D = (V, A) is a strongly connected directed graph, and C is a circle in D, then Theorem 2 shows that if the direction of each arc on circle C is reversed, the resulting new directed graph D' is still a strongly connected directed graph. Let the sequence ε = {P 0 ,P 1 ,P 2 ,…,P t} is an ear sequence of a strongly connected directed graph D, and the reverse ear P 0 Mid-arc direction, use If The corresponding directed graph D' is a strongly connected directed graph; therefore, the reverse direction circle obtained by randomly selecting a circle and reversing the direction of each arc on the circle is still a strongly connected directed graph; therefore, it is possible to set the random reversal of the directions of all arcs on a circle of a strongly connected directed graph as a neighborhood action;

[0053] Reverse Road:

[0054] Theorem 3: Assume that D = (V, A ∪ P) is a strongly connected directed graph, where P is a path from vertex a to vertex b. Let is the directed graph obtained after reversing P, then The necessary and sufficient condition for a strongly connected directed graph is that there are at least two paths from vertex a to vertex b in the directed graph D and these paths do not intersect;

[0055] For a path in a strongly connected directed graph D, all arcs on the reversed path can still maintain the strong connectivity of the directed graph; therefore, it is only necessary to determine whether there are at least two non-intersecting paths between two vertices to determine the neighborhood structure formed by the operation of reversing the path and ensure that its neighborhood structure satisfies strong connectivity;

[0056] Subgraph redirection:

[0057] Theorem 4: Assume that the directed graph D = (V, A) is a strongly connected directed graph; we Figure 2Decomposed into two parts: D1 = (V1, A1) and D2 = (V2, A2); the base graph of D1 is an undirected connected graph; now, if we redirect D1 to obtain a strongly connected graph D 1 '=(V1,A 1 '), then the union (D 1 '∪D 2 )The new graph will be a strongly connected directed graph; so according to Theorem 4, first a strongly connected directed graph is decomposed into two subgraphs by binary decomposition, one of the subgraphs is randomly selected, and the directions of all arcs in the subgraph are flipped to obtain a new strongly connected directed subgraph, which is then merged with the other original subgraph to obtain a new strongly connected directed graph.

[0058] Furthermore, the local search algorithm introduced in step S3 is the local search algorithm of the best improved search strategy or the local search algorithm of the first improved search strategy.

[0059] One of the above technical solutions includes the following beneficial effects: This solution is to minimize the material handling cost and optimize the AGV guidance path network, which reduces this problem to the problem of solving the optimal strongly connected orientation of the graph. Therefore, the total transportation cost consisting of the AGV's carrying cost and the empty cost is minimized as the optimization goal, and the corresponding mathematical model is established. Then, in order to improve the local search capability, based on the characteristics of the reversal path, reversal circle, and subgraph redirection in the strong connectivity of the directed graph to maintain strong connectivity, three methods for generating neighborhood structures are proposed to ensure the effectiveness of the solutions generated in the neighborhood solution search process and improve the efficiency and quality of the solution. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 is a flow chart of the overall steps of an embodiment of the present invention;

[0061] Figure 2 is a manufacturing plant distribution diagram of an embodiment of the present invention;

[0062] Figure 3 is a four-unit undirected graph of one embodiment of the present invention;

[0063] Figure 4 is the generation of a strongly connected directed graph according to an embodiment of the present invention;

[0064] Figure 5 is a schematic diagram of a reversal circle of an embodiment of the present invention;

[0065] Figure 6 is a schematic diagram of a reverse path of an embodiment of the present invention;

[0066] Figure 7is a decomposed schematic diagram of a directed graph of an embodiment of the present invention;

[0067] Figure 8 It is a schematic diagram of redirection of a sub-graph according to an embodiment of the present invention. DETAILED DESCRIPTION

[0068] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and cannot be understood as limiting the present invention.

[0069] like Figure 1 As shown, a method for optimizing the design of an AGV guided path network considering strong connectivity includes the following steps:

[0070] S0. Under the condition that the location of each unit loading and unloading point and the amount of logistics between units are known, the driving direction of each AGV lane is determined to ensure that the guidance path can meet the strong connectivity constraint condition; the total transportation distance formed by the AGV load distance and empty distance is minimized, and the guidance path network mathematical model is constructed;

[0071] S1. Use the general variable neighborhood search algorithm to obtain the guidance path of the guidance path network mathematical model; understand and identify the strong connectivity structure of the graph through the deep search method of the graph;

[0072] S2. According to the algorithm for generating the ear decomposition sequence of the strongly connected graph, the initial solution of the ear decomposition sequence of the strongly connected graph can be obtained in linear time;

[0073] S3, neighborhood search, setting neighborhood actions including reversal circles, reversal paths and subgraph redirection under the premise of satisfying strong connectivity constraints, exploring one or more neighborhoods; at the same time, introducing local search algorithms to find better solutions; if a better solution is found, updating the current solution and adjusting the search strategy as needed; after reaching the preset termination condition, outputting the optimal solution in the whole process;

[0074] S4. Combine the optimal solution in the descending neighborhood search step into the general variable neighborhood search algorithm to enhance the search capability.

[0075] In order to minimize the material handling cost, this scheme optimizes the AGV guided path network and reduces this problem to the problem of solving the optimal strongly connected orientation of the graph. Therefore, the total transportation cost consisting of the AGV's carrying cost and empty cost is minimized as the optimization goal, and the corresponding mathematical model is established. Then, in order to improve the local search ability, based on the characteristics of the strong connectivity of the directed graph, the reverse path, the reverse circle, and the subgraph redirection to maintain strong connectivity, three methods for generating neighborhood structures are proposed to ensure the effectiveness of the solutions generated in the neighborhood solution search process and improve the efficiency and quality of the solution. Then, based on the various neighboring orders composed of these three neighborhoods and the two search strategies, they are combined with different VND algorithms (variable neighborhood descent method) to form a variety of VND variants, and experimental solutions are carried out for the unidirectional guided path direction orientation problem, and the two VND variants with the best solution effect are obtained. Finally, based on the combination of the two optimal VND variants with the GVNS algorithm, the optimal neighborhood search algorithm for solving the AGV guided path problem is obtained. The neighborhood search algorithm can effectively find the optimal solution or a solution close to the optimal solution of the guided path problem, and can be used to solve the guided path problem well. In the process of using the neighborhood search algorithm, the generated neighborhood structure must ensure that it has strong connectivity, and the directed graph ear decomposition can provide an intuitive and effective method to identify the strongly connected components in the graph. Through topological sorting and ear merging, the strongly connected structure of the graph can be clearly understood and identified, and the required neighborhood structure that satisfies strong connectivity is generated for the algorithm according to the conditions, thus forming a complete neighborhood search algorithm and effectively solving the guided path problem. Therefore, the variable neighborhood descent search algorithm based on the directed graph ear decomposition sequence can effectively solve the AGV guided path problem under the premise of ensuring that the generated solution has strong connectivity.

[0076] like Figure 3 As shown, the mathematical model in step S0 is as follows:

[0077] In order to describe the problem and accurately establish the corresponding mathematical model, the following definitions are made:

[0078] There is no upper limit to AGV transport capacity;

[0079] Each unit has a unique loading and unloading port, and the position is known;

[0080] The transport paths of AGVs are all the shortest paths;

[0081] It does not consider whether the AGV will be blocked;

[0082] Where N is the number of units; S is the vertex set; P n is the feeding port, n∈[1,N]; D n is the unloading port, n∈[1,N]; E(i,j) is the lane from vertex i to vertex j; T ijis the lane direction; H ij is the lane length; F bd From the feed port D b To the feeding port P d Number of handling times; G db From the feeding port P d To the feeding port D b No-load times; L ij is the shortest path from vertex i to vertex j; Is lane E(i,j) located at the unloading port D? b To the feeding port P d on the transport route; Is lane E(i,j) located at the loading port P? d To the feeding port D b On the empty route; q bd is the load flow matrix between unit b and unit d; is the no-load flow matrix between unit b and unit d;

[0083]

[0084]

[0085] T ij +T ji =1 Formula (2.12)

[0086] ∑ i∈S T ij ≥1 Formula (2.13)

[0087] Σ j∈S T ij ≥1 Formula (2.14)

[0088] Σ d∈N F bd =Σ d∈N G db Formula (2.15)

[0089]

[0090] Formula (2.1) of the mathematical programming model is the objective function. The optimization goal is to make the total running distance of the AGV as short as possible, and the total running distance of the AGV includes two parts: the empty distance and the loaded distance. Constraint formulas (2.2) and (2.3) ensure that the length of the route selected by the calculation is equal to the empty distance, that is, from the loading port P b To the feeding port D d , from the feed port D d To the feeding port P b The empty route distance is equal to the selected lane length and;

[0091] Constraints (2.4)-(2.5) are to ensure that there is only one inlet and outlet for the feeder P. b Lanes;

[0092] Constraints (2.6)-(2.7) are to ensure that the inlet and outlet D d The number of lanes is only one each;

[0093] Constraints (2.8)-(2.9) are to ensure that the number of times the AGV enters all non-loading and unloading point vertices is consistent with the number of times it leaves the vertex;

[0094] Constraints (2.10)-(2.11) are to ensure that the lanes in each empty transport route will not be repeated;

[0095] The constraint (2.12) is to ensure that each lane is one-way;

[0096] Constraints (2.13)-(2.14) are to ensure that the guidance path network is strongly connected. It is necessary to ensure that for any vertex in the vertex set, there is a guidance path starting from and ending at the vertex;

[0097] The role of constraints (2.15)-(2.16) is to ensure logistics balance, that is, to ensure that the number of times the AGV is loaded is equal to the number of times it is empty;

[0098] In addition, constraint (2.17) is a variable constraint, ensuring that each variable in it is a 0-1 variable; from the mathematical model, it contains a large number of 0-1 variables and constraints, which will cause the solution space to grow exponentially with the increase in the scale of decision variables.

[0099] Wherein, in step S2:

[0100] The ear decomposition of a directed graph is to decompose it into a set of paths and cycles, and there are no common edges between all paths and cycles. Its specific definition is as follows: For a directed multigraph D, its ear decomposition is a sequence ε = {P 0 ,P 1 ,…P t}, where P 0 Represents a circle, each P i (1≤i≤t) is a path or a cycle with the following properties:

[0101] (1) When i≠j, P i and P j The arcs do not intersect;

[0102] (2) For each i = 1, 2, ..., t, if P i is a circle, then P iWith V(D i-1 ) have only one common vertex; otherwise, P i The endpoint is V(D i-1 ) are not the same vertices, and P i The remaining vertices of V(D i-1 ); here, D i Indicates that it has a vertex set and arc set The directed graph formed;

[0103] (3)

[0104] Each P i is called an ear of ε, the number of ears in ε is t+1, if an ear |A(P j )|=1, it is called ordinary ear; otherwise it is called extraordinary ear;

[0105] Theorem 1: A directed graph D = (V, A) is a strongly connected graph if and only if it has at least one ear decomposition sequence; in addition, if D is a strongly connected graph, for each vertex v, each cycle C containing v can be used as the initial cycle P of the ear decomposition sequence of D. 0 ;

[0106] From Theorem 1, we know that every strongly connected directed graph D has an ear decomposition sequence. Selecting different vertices v can get different ear decomposition sequences of D. Obviously, every strongly connected directed graph D has multiple ear decomposition sequences, but the ear decomposition sequences of two different strongly connected directed graphs must be different.

[0107] Corollary 1: For a strongly connected directed graph with n vertices and m arcs, each ear decomposition sequence contains m-n+1 ears, and there exists a linear time algorithm that can find the ear decomposition of a strongly connected directed graph D;

[0108] Through the proof process of Theorem 1, it is easy to prove that each ear decomposition sequence contains m-n+1 ears; we can use a depth-first search algorithm based on the directed graph D to construct an ear in each backtracking process; since the entire backtracking only traverses each vertex and arc once, the algorithm is a linear time algorithm.

[0109] In addition, in step S2:

[0110] By using the graph depth search method, we can determine the strongly connected directed graph D of the undirected graph G. Then, according to the algorithm for generating the ear decomposition sequence, we can find the ear decomposition sequence ε of the strongly connected directed graph D in linear time.

[0111] First, the workshop AGV transport track network to be directed is modeled as an undirected graph G, and then the strongly connected directed graph D of G is determined, and finally it is converted into an ear decomposition sequence ε. In this way, an initial solution can be generated within a linear time range.

[0112] The initial solution generation steps are:

[0113]

[0114]

[0115] by Figure 3 Taking the workshop unit shown in the figure as an example, the undirected graph G of the AGV track network is as follows Figure 4 According to step 3 of the initial population generation algorithm, we first randomly select a vertex as the starting point, and then use the graph depth search method to label each vertex from small to large (label from small to large in the search order), as shown in Figure 4 As shown in (b), after labeling the vertex labels, the edges in the undirected graph G are oriented from the vertex with smaller label values ​​to the vertex with larger label values ​​according to the vertex label numbers, and a strongly connected directed graph D is obtained, as shown in Figure 4 Finally, according to steps 4 and 5 of the initial population generation algorithm, the ear decomposition sequence ε = {P 0 ,P 1 ,P 2 ,P 3 ,P 4}.

[0116] In addition, in step S3:

[0117] For the strong connectivity constraint, we design neighborhood actions that satisfy this constraint from the perspective of graph theory;

[0118] Reversal circle:

[0119] Theorem 2: If D' and D" are two k-strongly connected orientations of an undirected graph G = (V, E), then there exists a set of k-strongly connected orientations D' = D 0 ,D 1 ,…,D k =D", so that every D i (i=1,2,…,k) can be obtained by reversing a directed loop or path. i-1 ;

[0120] If D = (V, A) is a strongly connected directed graph, and C is a circle in D, then Theorem 2 shows that if the direction of each arc on circle C is reversed, the resulting new directed graph D' is still a strongly connected directed graph. Let the sequence ε = {P 0 ,P 1 ,P2 ,…,P t} is an ear sequence of a strongly connected directed graph D, and the reverse ear P 0 Mid-arc direction, use If The corresponding directed graph D' is a strongly connected directed graph; therefore, randomly selecting a circle and reversing the direction of each arc on the circle to obtain a reverse direction circle is still a strongly connected directed graph; therefore, we can set the random reversal of the directions of all arcs on a circle of a strongly connected directed graph as a neighborhood action, as follows Figure 5 As shown;

[0121] Reverse Road:

[0122] Theorem 3: Assume that D = (V, A ∪ P) is a strongly connected directed graph, where P is a path from vertex a to vertex b. Let is the directed graph obtained after reversing P, then The necessary and sufficient condition for a strongly connected directed graph is that there are at least two paths from vertex a to vertex b in the directed graph D and these paths do not intersect;

[0123] For a path in a strongly connected directed graph D, reversing all arcs on the path can still maintain the strong connectivity of the directed graph; therefore, we only need to determine whether there are at least two disjoint paths between two vertices to determine the neighborhood structure formed by the reversal operation and ensure that its neighborhood structure satisfies strong connectivity, as shown in the following example: Figure 6 As shown;

[0124] Subgraph redirection:

[0125] Theorem 4: Assume that the directed graph D = (V, A) is a strongly connected directed graph; we Figure 2 Decomposed into two parts: D1 = (V1, A1) and D2 = (V2, A2); the base graph of D1 is an undirected connected graph; now, if we redirect D1 to obtain a strongly connected graph D 1 '=(V1,A 1 '), then the union (D 1 '∪D 2 ) The new graph will be a strongly connected directed graph; so according to Theorem 4, firstly, a strongly connected directed graph is decomposed into two subgraphs by binary decomposition, such as Figure 7 As shown, randomly select one of the subgraphs and flip the directions of all arcs in the subgraph to obtain a new strongly connected directed subgraph, which is then merged with the original subgraph to obtain a new strongly connected directed graph, as shown in Figure 8 shown.

[0126] In addition, the local search algorithm introduced in step S3 is a local search algorithm of the best improved search strategy or a local search algorithm of the first improved search strategy.

[0127] This scheme conducts comparative analysis experiments on VND search strategies: mainly targeting VND (Variable Neighborhood Descent) variants, and the main goal is to experimentally compare multiple VND strategies in existing literature to observe whether there are significant differences between them. The following aspects that may affect the effect of VND variants are considered:

[0128] 1. The order in which neighbors are selected may affect the quality of the final solution.

[0129] 2. Whether to use a first improvement strategy or a best improvement strategy may have an impact on the quality of the solution. After sorting the neighborhood structures into lists, there are several possibilities for performing the search within them. One question that needs to be considered for any local search heuristic is whether to use: (i) a first improvement strategy - once a solution is found to be an improvement, it is set as the new current solution; or (ii) a best improvement strategy - among all the improved solutions, the best solution is set as the new current solution.

[0130] 3. Move or not move decision (i.e., if an improvement is detected in a certain neighborhood, there are several options to continue searching after updating the current solution):

[0131] (1) Basic VND (B-VND)—Returns to the first neighbor in the list.

[0132] (2) Pipe VND (P-VND) - Continue searching within the same neighborhood.

[0133] (3) Cyclic VND (C-VND) - resume the search from the next neighborhood in the list.

[0134] (4) Union VND (U-VND) or multi-neighborhood search, where a single neighborhood is the union of all predefined neighborhoods—continuous search within the same large neighborhood. U-VND was proposed and used in tabu search in Lu et al. (2011) and Wu et al. (2012).

[0135] As can be seen from the introduction, there are several aspects that may affect the effect of VND variants. According to the random combination of these aspects, more than a dozen different VND variants can be obtained. Therefore, in order to select the most suitable VND variant to solve the AGV guidance path problem in the system, in this section, several VND variants will be matched with the above aspects that affect the VND effect, and they will be used to solve the guidance path problems of multiple different unit sizes, so as to compare the effects of different VND variants in solving the guidance path problem and obtain the VND variant with the best effect.

[0136] The experimental results for each VND variant are shown below. We use the following abbreviations to indicate the search strategy used:

[0137] BI: Search using the best improved search strategy.

[0138] FI: Search was conducted using the first improvement search strategy exploration.

[0139] To distinguish the six possible neighborhood orders, we use the following notation:

[0140] 1st: "reverse circle", "reverse road", "subgraph redirection";

[0141] 2nd: "reversal circle", "subgraph redirection", "reversal road";

[0142] 3rd: "reverse path", "reverse circle", "subgraph redirection";

[0143] 4th: "reverse path", "subgraph redirection", "reverse circle";

[0144] 5th: "Subgraph redirection", "reversal circle", "reversal road";

[0145] 6th: “Subgraph redirection”, “reversal path”, “reversal circle”;

[0146] Summarizing the experimental results, the following conclusions can be drawn:

[0147] 1. Through multiple examples and different objective functions, from the experimental data of each mobile decision (B-VND, P-VND, C-VND, U-VND), under the premise of the same neighborhood order, the solution obtained by using the optimal improved search strategy will be better than the solution obtained by using the first improved search strategy, and the average solution obtained is also relatively better.

[0148] 2. Under the same conditions, by adjusting the neighborhood order under different mobility decisions (B-VND, P-VND, C-VND, U-VND), the U-VND decision is more stable and resistant to interference when the neighborhood order changes. Compared with other decision strategies, no matter how the neighborhood order changes, the U-VND decision can ensure that the solution obtained is more stable, which shows that U-VND has strong adaptability to changes in external conditions and can better resist changes in the neighborhood order, ensuring that its performance is not affected. For other strategies, especially C-VND, it is most sensitive to changes in the neighborhood order. When the neighborhood order changes, the performance of C-VND will be significantly affected.

[0149] 3. Through actual experimental results, it is found that when the "reverse path" is set as the first neighborhood structure strategy, it has more obvious advantages than the other two neighborhood structures. The experimental results show that when the "reverse path" is used as the first neighborhood structure, it can not only ensure that this VND variant can obtain a better solution than other neighborhood structures in most cases, but also ensure that the obtained solution is a high-quality solution. In addition, the solution time using the 3rd neighborhood order will be slightly faster than the solution time required using the 4th neighborhood order. In contrast, when the "reverse circle" is used as the first neighborhood structure, the experimental results are obviously not ideal. Not only is it slightly insufficient in solving the problem, but also the quality rate of the solution is low.

[0150] 4. For the same example, the same neighbor order, and the same search strategy, the solutions obtained by the four mobile decisions (B-VND, P-VND, C-VND, and U-VND) are compared. Compared with P-VND and C-VND, B-VND and U-VND show better results in obtaining the optimal solution. Whether it is a simple or complex problem, B-VND can provide relatively high-quality solutions. On the other hand, U-VND has a strong optimization ability and a broad search space, which can better adapt to problems of various sizes and provide high-quality solutions.

[0151] 5. Comparing from the perspective of solution time, no matter what scale of the case is processed, the solution time of C-VND is significantly less than the other three. Relatively speaking, the solution time of B-VND and U-VND is relatively long, because their search range is wider and more time is needed to find a high-quality solution. This does not mean that these two strategies are inferior, but they may not be the best choice in an environment with time requirements or resource constraints. The solution time of B-VND is the longest among the four decisions, which means that if the scale of the case is large, or a better solution needs to be obtained in a short time, B-VND may not be the best choice. Considering that the solutions obtained by B-VND are all better, B-VND is more suitable for problems with higher solution quality requirements without considering the solution time.

[0152] 6. Comparing the time required for solving under the same conditions, it is found that no matter which move decision is executed, the first improved search strategy saves a lot of solving time than the optimal improved search strategy. This is because the optimal improved search strategy takes longer because it performs more comparisons and judgments during the operation. Although this will improve the quality of the solution to a certain extent, it also increases the complexity and time consumption of the calculation.

[0153] 7. When using B-VND decision making, the optimal solution can be obtained when it is combined with the optimal improved search strategy and the 2nd neighbor order; for P-VND, it shows the best problem-solving ability when it is combined with the optimal improved search strategy and the 5th neighbor order, and the solution obtained at this time will be the best; and for C-VND and U-VND, the solution obtained is the best when combined with the optimal improved search strategy and the 4th neighbor order.

[0154] 8. In terms of solution quality, the four decision strategies B-VND, U-VND, P-VND and C-VND show different advantages, among which B-VND has the highest solution quality, followed by U-VND, P-VND and C-VND. This shows that the B-VND strategy has a significant advantage in obtaining high-quality solutions. In terms of solution time, C-VND has the fastest execution speed, followed by P-VND, U-VND and B-VND. This shows that C-VND may be a better choice in scenarios where time is tight or a solution needs to be given quickly.

[0155] 9. U-VND can obtain a better solution while taking into account the solution time requirements. It can not only obtain the solution in a shorter time, but also the quality of the solution is relatively high. Therefore, U-VND may be the best choice as a mobile decision that can strike a balance between solution quality and solution time. Secondly, B-VND performs best in solution quality. Although it ranks last in solution time, if we prefer to obtain high-quality results rather than quickly get a solution.

[0156] The following are prior art citations:

[0157] Theorem 1 and Theorem 4 refer to "Research on Production Workshop Layout Method Considering Intelligent Material Storage and Transportation System";

[0158] Theorem 2, Theorem 3, Digraphs-theory, algorithms and applications[M]. Springer Publishing Company, Incorporated, 2010;

[0159] The deep search method of the graph, refer to "A Review of Research on Automated Guided Vehicle Scheduling Optimization";

[0160] Union VND, refer to Neighborhood analysis: a case study on curriculum-basedcourse timetabling[J].Journal ofHeuristics 17,2,97–118, Multi-neighborhoodtabu search for the maximum weight clique problem[J].Annals of OperationsResearch196,1,611–634.

[0161] The technical principle of the present invention is described above in conjunction with specific embodiments. These descriptions are only for explaining the principle of the present invention and cannot be interpreted as limiting the scope of protection of the present invention in any way. Based on the explanations herein, those skilled in the art can associate other specific implementations of the present invention without paying creative labor, and these methods will fall within the scope of protection of the present invention.

Claims

1. An AGV guided path network design optimization method considering strong connectivity, It is characterized in that The following steps are involved: S0. Under the condition that the location of each unit loading and unloading point and the amount of logistics between units are known, the driving direction of each AGV lane is determined to ensure that the guidance path can meet the strong connectivity constraint condition; the total transportation distance formed by the AGV load distance and empty distance is minimized, and the guidance path network mathematical model is constructed; S1. Use the general variable neighborhood search algorithm to obtain the guidance path of the guidance path network mathematical model; understand and identify the strong connectivity structure of the graph through the deep search method of the graph; S2. According to the algorithm for generating the ear decomposition sequence of the strongly connected graph, the initial solution of the ear decomposition sequence of the strongly connected graph can be obtained in linear time; S3, neighborhood search, setting neighborhood actions including reversal circles, reversal paths and subgraph redirection under the premise of satisfying strong connectivity constraints, exploring one or more neighborhoods; at the same time, introducing local search algorithms to find better solutions; if a better solution is found, updating the current solution and adjusting the search strategy as needed; after reaching the preset termination condition, outputting the optimal solution in the whole process; S4. Combine the optimal solution in the descending neighborhood search step into the general variable neighborhood search algorithm to enhance the search capability.

2. The AGV guidance path network design optimization method considering strong connectivity according to claim 1, It is characterized in that The mathematical model in step S0 is as follows: In order to describe the problem and accurately establish the corresponding mathematical model, the following definitions are made: There is no upper limit to AGV transport capacity; Each unit has a unique loading and unloading port, and the position is known; The transport paths of AGVs are all the shortest paths; Regardless of whether the AGV will be blocked; Where N is the number of units; S is the vertex set; P n is the feeding port, n∈[1,N]; D n is the unloading port, n∈[1,N]; E(i,j) is the lane from vertex i to vertex j; T ij is the lane direction; H ij is the lane length; F bd From the feed port D b To the feeding port P d Number of transports; G db From the feeding port P d To the feeding port D b No-load times; L ij is the shortest path from vertex i to vertex j; Is lane E(i, j) located at the unloading port D b To the feeding port P d on the transport route; Is lane E(i, j) located at the loading port P? d To the feeding port D b qbd is the load flow matrix between unit b and unit d; is the no-load flow matrix between unit b and unit d; T ij +T ji =1 Formula (2.12) ∑ i∈S T ij ≥1 Formula (2.13) ∑ j∈S T ij ≥1 Formula (2.14) ∑ d∈N F bd =∑ d∈N G db Formula (2.15) Formula (2.1) of the mathematical programming model is the objective function. The optimization goal is to make the total running distance of the AGV as short as possible, and the total running distance of the AGV includes two parts: the empty distance and the loaded distance. Constraints (2.2)-(2.3) ensure that the length of the route selected by the calculation is equal to the empty distance, that is, from the loading port Pb to the unloading port D d , from the feed port D d The empty route distance to the loading port Pb is equal to the length of the selected lane and; Constraints (2.4)-(2.5) are to ensure that there is only one lane for entering and exiting the loading port Pb; Constraints (2.6)-(2.7) are to ensure that the inlet and outlet D d The number of lanes is only one each; Constraints (2.8)-(2.9) are to ensure that the number of times the AGV enters all non-loading and unloading point vertices is consistent with the number of times it leaves the vertex; Constraints (2.10)-(2.11) are to ensure that the lanes in each empty transport route will not be repeated; The constraint (2.12) is to ensure that each lane is one-way; Constraints (2.13)-(2.14) are to ensure that the guidance path network is strongly connected. It is necessary to ensure that for any vertex in the vertex set, there is a guidance path starting from and ending at the vertex; The role of constraints (2.15)-(2.16) is to ensure logistics balance, that is, to ensure that the number of times the AGV is loaded is equal to the number of times it is empty; In addition, constraint (2.17) is a variable constraint, ensuring that each variable in it is a 0-1 variable; from the mathematical model, it contains a large number of 0-1 variables and constraints, which will cause the solution space to grow exponentially with the increase in the scale of decision variables.

3. The AGV guidance path network design optimization method considering strong connectivity according to claim 1, It is characterized in that In step S2: The ear decomposition of a directed graph is to decompose it into a set of paths and cycles, and there are no common edges between all paths and cycles. Its specific definition is as follows: For a directed multigraph D, its ear decomposition is a sequence ε = {P 0 , P 1 , …P t }, where P 0 Represents a circle, each P i (1≤i≤t) is a path or a cycle with the following properties[66]: (1) When i≠j, P i and P j The arcs do not intersect; (2) For each i = 1, 2, ..., t, if P i is a circle, then P i With V(D i-1 ) have only one common vertex; otherwise, P i The endpoint is V(D i-1 ) are not the same vertices, and P i The remaining vertices of V(D i-1 ); here, D i Indicates that it has a vertex set and arc set The directed graph formed; (3) Each P i is called an ear of e, and the number of ears in ε is t+1. If an ear |A(P j )|=1, it is called ordinary ear; otherwise it is called extraordinary ear; Theorem 1: A directed graph D = (V, A) is a strongly connected graph if and only if it has at least one ear decomposition sequence; in addition, if D is a strongly connected graph, for each vertex v, each cycle C containing v can be used as the initial cycle P of the ear decomposition sequence of D. 0 ; From Theorem 1, we know that every strongly connected directed graph D has an ear decomposition sequence. Selecting different vertices v can get different ear decomposition sequences of D. Obviously, every strongly connected directed graph D has multiple ear decomposition sequences, but the ear decomposition sequences of two different strongly connected directed graphs must be different. Corollary 1: For a strongly connected directed graph with n vertices and m arcs, each ear decomposition sequence contains m-n+1 ears, and there exists a linear time algorithm that can find the ear decomposition of a strongly connected directed graph D; Through the proof process of Theorem 1, it is easy to prove that each ear decomposition sequence contains m-n+1 ears; we can use a depth-first search algorithm based on the directed graph D to construct an ear in each backtracking process; since the entire backtracking only traverses each vertex and arc once, the algorithm is a linear time algorithm.

4. The AGV guidance path network design optimization method considering strong connectivity according to claim 3, It is characterized in that In step S2: By using the graph depth search method, we can determine the strongly connected directed graph D of the undirected graph G. Then, according to the algorithm for generating the ear decomposition sequence, we can find the ear decomposition sequence ε of the strongly connected directed graph D in linear time. First, the workshop AGV transport track network to be directed is modeled as an undirected graph G, and then the strongly connected directed graph D of G is determined, and finally it is converted into an ear decomposition sequence ε. In this way, an initial solution can be generated within a linear time range.

5. The AGV guidance path network design optimization method considering strong connectivity according to claim 1, It is characterized in that In step S3: For the strong connectivity constraint, we design neighborhood actions that satisfy this constraint from the perspective of graph theory; Reversal circle: Theorem 2: If D' and D" are two k-strongly connected orientations of an undirected graph G = (V, E), then there exists a set of k-strongly connected orientations D' = D 0 ,D 1 ,…,D k =D", so that every D i (i=1,2,…,k) can be obtained by reversing a directed loop or path. i-1 ; If D = (V, A) is a strongly connected directed graph, and C is a circle in D, Theorem 2 shows that by reversing the direction of each arc on circle C, the resulting new directed graph D' is still a strongly connected directed graph; Let the sequence ε={P 0 ,P 1 ,P 2 ,…,P t } is an ear sequence of a strongly connected directed graph D, and the reverse ear P 0 Mid-arc direction, use If The corresponding directed graph D' is a strongly connected directed graph; therefore, the reverse direction circle obtained by randomly selecting a circle and reversing the direction of each arc on the circle is still a strongly connected directed graph; therefore, it is possible to set the random reversal of the directions of all arcs on a circle of a strongly connected directed graph as a neighborhood action; Reverse Road: Theorem 3: Assume that D = (V, AU P) is a strongly connected directed graph, where P is a path from vertex a to vertex b. Let is the directed graph obtained after reversing P, then The necessary and sufficient condition for a strongly connected directed graph is that there are at least two paths from vertex a to vertex b in the directed graph D and these paths do not intersect; For a path in a strongly connected directed graph D, reversing all arcs on the path can still maintain the strong connectivity of the directed graph; Therefore, we only need to determine whether there are at least two non-intersecting paths between two vertices to determine the neighborhood structure formed by the operation of reversing the path, and ensure that its neighborhood structure satisfies strong connectivity; Subgraph redirection: Theorem 4: Assume that the directed graph D = (V, A) is a strongly connected directed graph; we decompose this graph into D 1 =(V 1 , A 1 ) and D 2 =(V 2 , A 2 ) two parts; among them, D 1 The base graph of D is an undirected connected graph; now, if we reorient D 1 Get the strongly connected graph D 1 ′=(V 1 , A 1 ′), then by D 2 and D 1 ′The union of the two (D 1 ′∪ D 2 )The new graph will be a strongly connected directed graph; so according to Theorem 4, first a strongly connected directed graph is decomposed into two subgraphs by binary decomposition, one of the subgraphs is randomly selected, and the directions of all arcs in the subgraph are flipped to obtain a new strongly connected directed subgraph, which is then merged with the other original subgraph to obtain a new strongly connected directed graph.

6. The AGV guidance path network design optimization method considering strong connectivity according to claim 1, It is characterized in that The local search algorithm introduced in step S3 is a local search algorithm of the best improved search strategy or a local search algorithm of the first improved search strategy.

Citation Information

Patent Citations

  • Job shop motion orbit path guiding optimization method based on hybrid genetic algorithm

    CN114444239A