Optimal coordination method for logistics warehousing based on distributed differential game
Through the distributed differential game method and artificial potential field method, collision avoidance rules are designed in the multi-automatic guide car system, the coordination problem under limited perception and communication is solved, and more efficient task completion and better scalability are achieved.
Patent Information
- Application Number
- CN202211729139.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-12-30
- Publication Date
- 2024-12-03
- Estimated Expiration
- 2042-12-30
AI Technical Summary
The existing differential game method is difficult to effectively coordinate collision avoidance problems in multi-automatic guided car systems, resulting in low task efficiency and cannot be applied to warehousing and logistics scenarios with perception and communication limitations.
The distributed differential game method is adopted, combined with the artificial potential field method to design collision avoidance rules, and the Pontriajin minimum principle is used to analyze the existence and uniqueness of the local optimal coordination strategy, which is transformed into the optimal control problem to ensure the local optimal Global convergence of the strategy.
Under the conditions of limited perception and communication, the time for automatically guiding the trolley to the target point is reduced, the task completion efficiency is improved, and it has good scalability, which reduces the time for executing tasks by 16% compared to existing methods.
Smart Images

Figure CN116107276B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of coordinated control of multiple automatic guided vehicle systems, and in particular to an optimal coordination method for logistics warehousing based on distributed differential games. Background Art
[0002] With the development of economy, automated guided vehicles have become the core component of modern warehousing logistics, which can reduce transportation time and reduce the burden of manpower. In recent years, the research on effective coordination strategies of multiple automated guided vehicle systems has attracted the attention of many researchers. While considering the global goals, this coordination strategy can effectively achieve local goal requirements, maximize the economic benefits of warehousing logistics, and improve the transportation efficiency of warehousing logistics when automated guided vehicles perform their respective given tasks in warehousing logistics. That is, while avoiding collisions, the transportation tasks are completed efficiently. Considering the limited perception and communication of automated guided vehicles in warehousing logistics, it is very practical to analyze the coordination problem of automated guided vehicles under limited communication.
[0003] In order to improve the transportation efficiency of warehouse logistics, automated guided vehicles should ensure the safe and efficient completion of the given package transportation task. Therefore, a good coordination strategy needs to optimize the trajectory, which is the key to improving the transportation efficiency of automated guided vehicles. At present, there are mainly two types of coordination strategies, negotiation-based strategy and planning-based strategy. The negotiation-based strategy is a greedy algorithm, and the final strategy cannot achieve optimal coordination. The planning-based strategy is usually the best, but with the increase in the number of automated guided vehicles, the amount of calculation increases and the scalability is poor.
[0004] Game theory can optimize the behavior of each automated guided vehicle from the perspective of group performance to achieve Nash equilibrium and further improve group performance. In particular, differential games can quantify the dynamic interactions and conflict processes between multiple automated guided vehicles, and are often used to solve coordination problems, such as pursuit and formation problems. The literature (Mylvaganam T, Sassano M, Astolfi A. A differential game approach to multi-agent collision avoidance [J]. IEEE Transactions on Automatic Control, 2017, 62 (8): 4229-4235.) first used the differential game method to solve the problem of coordinated collision avoidance strategy. Including the literature (Lin W, Qu Z, Simaan MA. Nash strategies for pursuit-evasion differential games involving limited observations [J]. IEEE Transactions on Aerospace and Electronic Systems, 2015, 51 (2): 1347-1356.) proposed a method for constructing a feedback pursuit strategy that does not rely on the global state information of the agent. The paper (de la Cruz N, Jimenez-Lizarraga M. Finite time robust feedback Nash equilibrium for linear quadratic games [J]. IFAC-Papers On Line, 2017, 50 (1): 11794-11799.) established a centralized differential game model with external interference, treating the external interference as a virtual player that maximizes the cost function, but did not consider the limited communication capability of the agent. The paper (Fu Y, Chai T. Online solution of two-player zero-sum games for continuous-time nonlinear systems with completely unknown dynamics [J]. IEEE transactions on neural networks and learning systems, 2015, 27 (12): 2577-2587.) constructed a distributed uncertain zero-sum differential game and obtained a local robust Nash equilibrium, but there was no strict theoretical guarantee.In order to achieve the coordination of multi-agent global tasks, the global convergence of local robust Nash equilibrium is required. Although the above can successfully solve the multi-agent collision avoidance problem, it cannot be directly applied to solve the coordinated collision avoidance problem of automatic guided vehicles in warehousing logistics. The main reason is that, first of all, the local optimal solution obtained cannot improve the working efficiency of the automatic guided vehicle; secondly, the communication assumption of the automatic guided vehicle is perfect and not subject to communication restrictions.
[0005] Therefore, in order to solve the problem of low efficiency in completing tasks when differential game methods are introduced into the collision avoidance problem, the artificial potential field method can be introduced in combination with the distributed optimization method to design collision avoidance rules. Design a distributed differential game coordination collision avoidance strategy method. The differential game solutions proposed by the current existing technology mainly focus on the situation where global information is known. Distributed differential game methods are still rarely used in the collision avoidance problem of automatic guided vehicles in logistics warehousing, and cannot provide a suitable solution. Summary of the invention
[0006] In view of this, the purpose of the present invention is to provide a logistics warehousing optimal coordination method based on distributed differential game, aiming to solve the above problems.
[0007] To achieve the above object, the present invention adopts the following technical solution:
[0008] An optimal coordination method for logistics warehousing based on distributed differential game includes the following steps:
[0009] Step S1: Based on the warehousing logistics scenario, using graph theory, establish the communication topology structure between the guided vehicles in the multi-automatic guided vehicle system, including the game participant model and the obstacle environment model;
[0010] Step S2: using the artificial potential field method to design the operating cost function of the automatic guided vehicle in the game participant model;
[0011] Step S3: The logistics warehousing coordination problem of multiple automated guided vehicle systems subject to perception and communication constraints is regarded as a distributed differential game problem, and a distributed differential game model is established;
[0012] Step S4: Using the Pontryagin minimum principle, analyze the existence and uniqueness of the local optimal coordination strategy, and solve the expression of the local optimal coordination strategy, give the convergence conditions from the local optimal coordination to the global Nash equilibrium, and obtain the optimal coordination solution.
[0013] Furthermore, the warehousing and logistics scenario specifically includes
[0014] 1) Task allocation: In the work environment, the staff assigns the order to transport the package to the corresponding appropriate AGV;
[0015] 2) Package collection: After receiving the command, the AGV runs from the charging station to the corresponding shelf to pick up the package;
[0016] 3) Package transportation: After the automated guided vehicle picks up all assigned packages, it transports the packages to the designated required location;
[0017] 4) Return to the charging station: After the AGV completes the package delivery mission, it returns to the nearest charging station.
[0018] Furthermore, the game participant model is specifically:
[0019] The specific form of the dynamic equation of the automated guided vehicle is:
[0020]
[0021] In the formula, p i (t) and are the position and speed information of the i-th automated guided vehicle at time t, u i (t) is the control input of the i-th automated guided vehicle at time t, and Bii is the corresponding constant matrix;
[0022] The specific form of the position error of the automated guided vehicle is:
[0023]
[0024] In the formula, is the expected transportation target point position of the i-th automated guided vehicle, is the position error between the current transport position of the i-th automated guided vehicle at time t and the expected transport target point;
[0025] The specific form of the dynamic equation of the multi-automatic guided vehicle system is:
[0026]
[0027] In the formula, is the state of the automatic guided vehicle system at time t, is the state change rate of the automatic guided vehicle system at time t, N is the number of automatic guided vehicles, B i is the corresponding constant matrix,
[0028] Establish a directed interactive topological graph G(v,ε) of N automated guided vehicles; where v = {v1,...,v N} represents the set of automated guided vehicles, represents the set of edges, e ijrepresents the connection relationship between automatic guided vehicle i and automatic guided vehicle j, e ij ∈ε means that the automated guided vehicle i can receive information from the automated guided vehicle j. The neighbor set of the automated guided vehicle i is
[0029] The neighbor information matrix of automated guided vehicle i is defined as:
[0030]
[0031] Where I2 is a 2×2 unit matrix, is the Kronecker product, is the information matrix of neighboring guided vehicles j except for the automated guided vehicle i, where only the matrix F i The i-th row and the j-th row in are 1, and the rest are 0;
[0032] The local dynamic equation of the automated guided vehicle i is defined as:
[0033]
[0034] In the formula, is the local state information of the automated guided vehicle i at time t, is the corresponding constant matrix.
[0035] Furthermore, the obstacle environment model is specifically:
[0036] Expand the outlines of obstacles of various shapes in logistics warehousing into ellipses and define the collision avoidance area S ik for:
[0037]
[0038] In the formula, r i is the safety radius of the automated guided vehicle i, is the radius of obstacle k, c k (t) is the center of mass position of obstacle k at time t, E k =I2 is the identity matrix;
[0039] Define the sensing area D ik for:
[0040]
[0041] In the formula, R i is the sensing range of the automated guided vehicle i;
[0042] Define the free area M ik for:
[0043]
[0044] Further, the step S2 is specifically:
[0045] Given the following preconditions:
[0046] (1): The automatic guided vehicle will not slip during operation;
[0047] (2): The directed interaction topology graph G(v,ε) between the AGVs is fixed and strongly connected;
[0048] Design collision avoidance rules based on artificial potential field method:
[0049]
[0050] In the formula, the constant χ i (0<χ i <1) and is a constant, is the obstacle penalty function of the automated guided vehicle i at time t, which is defined as follows:
[0051]
[0052] is the distance penalty function of the automated guided vehicle i; for the trajectory optimization problem of the automated guided vehicle, the penalty function is defined as follows:
[0053]
[0054] In the formula, γ i is the deviation angle between the current trajectory of the automated guided vehicle i and the predefined reference trajectory.
[0055] Furthermore, the step S3 is specifically as follows:
[0056] The expression of the distributed differential game cost function is as follows:
[0057]
[0058] In the formula, t f is the final time of the game of the automated guided vehicle i, u -i (t) is the control strategy set of neighboring automated guided vehicles except automated guided vehicle i at time t, is the operating cost function of the automatic guided vehicle i at time t, which is defined as u ij (t) is the control strategy of the neighboring automated guided vehicle j of automated guided vehicle i at time t, and U(t) is the control cost at time t, which is defined as Where Rii ,R ij are the adjustable positive definite weight matrices corresponding to the control strategies of the automated guided vehicle i and the neighboring automated guided vehicle j respectively;
[0059] The goal of the coordinated control problem of distributed multi-AGV systems is to design the optimal coordination strategy for each AGV, and to ensure that the AGVs can safely reach the target point within a preset time. At the same time, AGV i and its neighbor AGVs can converge to a local Nash equilibrium, that is, the strategy set at time t is satisfy:
[0060]
[0061] In the formula, is the optimal cost function of the automated guided vehicle i at time t.
[0062] Furthermore, the step S4 specifically includes:
[0063] Assume that the additional state at time t is:
[0064]
[0065] In the formula, is the cost function at time t, is the initial condition, is the terminal moment cost;
[0066] Assume the expansion state at time t is:
[0067]
[0068] The corresponding dynamic equation is as follows:
[0069]
[0070] In the formula,
[0071] The distributed differential game problem is transformed into an optimal control problem, which can be expressed as follows:
[0072]
[0073] According to the Pontryagin minimum principle, the Hamiltonian function is defined as
[0074]
[0075] In the formula, λ i (t) is the Lagrange multiplier at time t;
[0076] The local optimal coordination strategy satisfies the following set of differential equations,
[0077]
[0078] And its boundary conditions:
[0079] p * (0) = p0,
[0080] λ i (t f )=0 (20)
[0081] Where p0 is the initial position of the automatic guided vehicle system, λ i (t f ) is the Lagrange multiplier at the final moment;
[0082] Assume that the only condition for the existence of a local optimal coordination strategy is: for a given distributed differential game problem of automatic guided vehicles in warehouse logistics, the existing local optimal coordination strategy The satisfied form is transformed into the following matrix form:
[0083]
[0084] In the formula, For a given matrix, such as is a positive definite matrix, then there exists a unique local coordination strategy;
[0085] Analyze the convergence of the local optimal coordination strategy to the global optimal strategy, which is defined as follows: For a differential game with N participants, if the following inequality holds, then the strategy set {u1(t),...,u i (t),...,u N (t)} converges to the global Nash equilibrium solution,
[0086]
[0087] In the formula, is the strategy of all the remaining automated guided vehicles except the automated guided vehicle i at time t. In addition, the strategy The following inequality is also satisfied:
[0088]
[0089] is the local optimal coordination strategy of the automated guided vehicle relative to its neighbors. Assuming that the topology graph between the automated guided vehicles is strongly connected, then: 1) for automated guided vehicle i, each local optimal coordination strategy is equal, 2) the local optimal coordination of the automated guided vehicle can converge to the global Nash equilibrium if and only if the communication topology graph of the automated guided vehicle is strongly connected.
[0090] Compared with the prior art, the present invention has the following beneficial effects:
[0091] The present invention is aimed at an automatic guided vehicle with a first-order linear model with limited perception and communication, and transforms the coordination control problem of multiple automatic guided vehicle systems in warehousing and logistics into a distributed differential game problem. Considering the existing differential game method that only considers obstacle avoidance goals, based on the artificial potential field method, a trajectory optimization goal is introduced to punish the degree to which the automatic guided vehicle deviates from the target point, weighing the distance between the automatic guided vehicle reaching the target point and the distance from the obstacle, and reducing the time for the automatic guided vehicle to reach the target point; based on the optimal control principle, the expression form of the local optimal strategy and the unique existence analysis are given; under the assumption of a fixed strongly connected topological graph, the global convergence of the local optimal strategy is guaranteed. This method can reduce the time it takes for the automatic guided vehicle to reach the transportation point. In addition, the introduction of a distributed architecture differential game model has good scalability compared to a centralized architecture. Compared with the existing more popular warehousing and logistics coordinated collision avoidance methods, this method reduces the time to execute tasks by 16%. BRIEF DESCRIPTION OF THE DRAWINGS
[0092] Figure 1 is a method principle block diagram of an embodiment of the present invention;
[0093] Figure 2 1 is a schematic diagram of a collision avoidance rule based on an artificial potential field method according to an embodiment of the present invention, wherein 1 is an automatic guided vehicle charging station, 2 is a shelf, 3 is a workbench, 4 is an obstacle, 5 is an automatic guided vehicle in operation, and 6 is a package transportation point;
[0094] Figure 3 is a position deviation diagram of a multi-automatic guided vehicle system on a horizontal axis in a simulation example implemented by the present invention;
[0095] Figure 4 is a position deviation diagram of a multi-automatic guided vehicle system on a longitudinal axis in a simulation example implemented by the present invention;
[0096] Figure 5 The communication topology diagram of the multi-automatic guided vehicle system adopted in the simulation example implemented by the present invention;
[0097] Figure 6 A diagram showing the transport trajectory of multiple automatic guided vehicle systems in a simulation example implemented by the present invention;
[0098] Figure 7 A diagram showing a comparison of the transport time of the algorithms in a simulation example implemented in the present invention. DETAILED DESCRIPTION
[0099] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.
[0100] Please refer to Figure 1 The present invention provides a logistics warehousing optimal coordination method based on distributed differential game, comprising the following steps:
[0101] Describe the warehousing logistics scenario, use graph theory to establish the communication topology between guided vehicles in the multi-automatic guided vehicle system, and use the guided vehicle and its neighbors as game participants to deal with possible obstacles in the work of the guided vehicle. A first-order linear integrator is established as the model of the guided vehicle. Considering that the working environment of the guided vehicle is complex and it is inevitable to encounter obstacles, the collision area, sensing area, and free area are designed for the working environment of the guided vehicle, and various obstacles of different shapes in the logistics warehouse are expanded into ellipses.
[0102] The collision avoidance rules are designed using the artificial potential field method as the operating cost function of the automated guided vehicle in the game model;
[0103] The logistics warehousing coordination problem of multiple automated guided vehicle systems subject to perception and communication constraints is regarded as a distributed differential game problem; a distributed differential game model is established, which includes the operation cost function and the control cost;
[0104] Using the Pontryagin minimum principle, the existence and uniqueness of the local optimal coordination strategy are analyzed, and the expression form of the local optimal coordination strategy is solved; the convergence of the global Nash equilibrium of the local optimal coordination is analyzed.
[0105] In this embodiment, the package transport task performed by the automated guided vehicle is to enable the automated guided vehicle to reach the target point without collision under the condition of limited perception and communication, thereby reducing the completion time of the automated guided vehicle transport task. That is, the position deviation of each automated guided vehicle approaches zero, and the transport task is completed.
[0106] In this embodiment, the information acquired by the automated guided vehicle includes the following categories: location information and strategy information of neighboring automated guided vehicles. The local state information of the automated guided vehicle and the first-order linear model of the automated guided vehicle are calculated through the above information:
[0107]
[0108] In the formula, p i (t) and are the position and speed information of the i-th automated guided vehicle at time t, u i (t) is the control input of the i-th automated guided vehicle at time t, B ii is the corresponding constant matrix;
[0109] The specific form of the position error of the automated guided vehicle is:
[0110]
[0111] In the formula, is the expected transport position of the i-th automated guided vehicle, is the position error between the current transport position of the i-th automated guided vehicle at time t and the expected transport target point;
[0112] The specific form of the dynamic equation of the multi-automatic guided vehicle system is:
[0113]
[0114] In the formula, is the state of the automatic guided vehicle system at time t, is the state change rate of the automatic guided vehicle system at time t, N is the number of automatic guided vehicles, B i is the corresponding constant matrix,
[0115] Establish a directed interactive topological graph G(v,ε) of N automated guided vehicles, where v = {v1,...,v N} represents the set of automated guided vehicles, represents the set of edges, e ij represents the connection relationship between automatic guided vehicle i and automatic guided vehicle j, e ij ∈ε means that the automated guided vehicle i can receive information from the automated guided vehicle j, and the neighbor set of the automated guided vehicle i is
[0116] The neighbor information matrix of automated guided vehicle i is defined as:
[0117]
[0118] Where I2 is a 2×2 unit matrix, is the information matrix of other neighboring guided vehicles j except automated guided vehicle i, where only matrix F i The i-th row and the j-th row in are 1, and the rest are 0;
[0119] The local dynamic equation of the automated guided vehicle i is defined as:
[0120]
[0121] In the formula, is the local state information of the automated guided vehicle i at time t, is the corresponding constant matrix;
[0122] Build an obstacle environment model:
[0123] Expand obstacles of various shapes in warehouse logistics into ellipses and define the collision avoidance area S ik for:
[0124]
[0125] In the formula, r i is the safety radius of the automated guided vehicle i, is the radius of obstacle k, c k (t) is the center of mass position of obstacle k at time t, E k =I2 is the identity matrix;
[0126] Define the sensing area D ik for:
[0127]
[0128] In the formula, R i is the sensing range of the automated guided vehicle i;
[0129] Define the free area M ik for:
[0130]
[0131] Given the following assumptions:
[0132] Assumption 1: The automated guided vehicle will not slip during operation;
[0133] Assumption 2: The directed interaction topology graph G(v,ε) between automated guided vehicles is fixed and strongly connected;
[0134] Design collision avoidance rules based on artificial potential field method:
[0135]
[0136] In the formula, the constant χ i (0<χ i <1) and is a constant, is the obstacle penalty function of the automated guided vehicle i at time t, is the distance penalty function of the automated guided vehicle i at time t;
[0137] The representation of the distance penalty function is as follows:
[0138]
[0139] In order to optimize the trajectory of the automated guided vehicle i, a distance penalty function is introduced to penalize the agent's deviation from the target point as follows:
[0140]
[0141] In the formula, γ i is the deviation angle of the automated guided vehicle i, which is the angle between the current trajectory of the automated guided vehicle i and the predefined reference trajectory;
[0142] The expression of the distributed differential game cost function is as follows:
[0143]
[0144] In the formula, t f is the final time of the game of the automated guided vehicle i, u -i (t) is the control strategy set of neighboring automated guided vehicles except automated guided vehicle i at time t, is the operating cost function of the automatic guided vehicle i at time t, which is defined as u ij (t) is the control strategy of the neighboring automated guided vehicle j of automated guided vehicle i at time t, and U(t) is the control cost at time t, which is defined as Where R ii ,R ij are the adjustable positive definite weight matrices corresponding to the control strategies of the automated guided vehicle i and the neighboring automated guided vehicle j respectively;
[0145] The goal of the coordinated control problem of distributed multi-AGV systems is to design the optimal coordination strategy for each AGV, and to ensure that the AGVs can reach the target point safely in the shortest possible time. At the same time, AGV i and its neighbor AGVs can converge to a local Nash equilibrium, that is, the strategy set at time t is satisfy:
[0146]
[0147] In the formula, is the optimal cost function of the automated guided vehicle i at time t;
[0148] The following theorems and expressions about the local optimal coordination strategy are given:
[0149] The additional state at time t is given as:
[0150]
[0151] In the formula, is the cost function at time t, is the initial state, is the terminal cost, and the expansion state at time t is given as:
[0152]
[0153] The corresponding dynamic equation is as follows:
[0154]
[0155] In the formula,
[0156] The distributed differential game problem is transformed into an optimal control problem, which can be expressed as follows:
[0157]
[0158] According to the Pontryagin principle, the Hamiltonian function is defined as
[0159]
[0160] In the formula, λ i (t) is the Lagrange multiplier at time t;
[0161] The local optimal coordination strategy satisfies the following set of partial differential equations:
[0162]
[0163] And boundary conditions:
[0164] p * (0) = p0,
[0165] λ i (t f )=0 (20)
[0166] Where p0 is the initial position of the automatic guided vehicle system, λ i (t f ) is the Lagrange multiplier at the final moment;
[0167] Given the unique condition for the existence of a local optimal coordination strategy: for a given distributed differential game problem of automated guided vehicles in warehouse logistics, the existing local optimal coordination strategy The satisfied form is transformed into the following matrix form:
[0168] The only law that exists for the local optimal coordination strategy is given: For a given distributed differential game problem of automated guided vehicles in warehouse logistics, the local optimal coordination strategy can be transformed into the following form:
[0169]
[0170] In the formula, For a given matrix, such as Positive definite matrix, then there exists a unique local coordination strategy;
[0171] In order to achieve the global optimal coordination of the transportation tasks of multiple automated guided vehicle systems, it is necessary to analyze the local
[0172] The convergence of the local optimal coordination strategy to the global optimal strategy is defined as follows: For a differential game with N players, if the following inequality holds, then the strategy set {u1(t),...,u i (t),...,u N (t)} converges to the global Nash equilibrium solution.
[0173]
[0174] In the formula, The strategies of all the remaining automated guided vehicles except automated guided vehicle i. In addition, the strategy So that the following equation holds;
[0175]
[0176] The proposition of global convergence of the local optimal coordination strategy is given: is the local optimal coordination strategy of the automated guided vehicle relative to its neighbors. Assuming that the topology graph between the automated guided vehicles is strongly connected, then: 1) for automated guided vehicle i, each local optimal coordination strategy is equal; 2) the local optimal coordination of the automated guided vehicle can converge to the global Nash equilibrium if and only if the communication topology graph of the automated guided vehicle is strongly connected.
[0177] In this embodiment, two specific examples are given to demonstrate the effectiveness and superiority of the proposed distributed differential game method in solving the multi-automatic guided vehicle coordination problem and verify the advantages of the method through experiments.
[0178] In order to prove that the obtained local optimal coordination strategy can reduce the time for the automated guided vehicle to complete the transportation task, this example conducts a simulation experiment and compares it with the existing centralized differential game method that only considers the obstacle penalty target. The specific model expression of the multi-automatic guided vehicle system is given as follows:
[0179]
[0180] In the formula, p1(t) and are the position and speed information of the first automated guided vehicle at time t, p2(t) and are the position and speed information of the second automated guided vehicle at time t, u1(t) and u2(t) represent the control strategies of automated guided vehicles 1 and 2 respectively;
[0181] The initial position and transportation target point position of each automatic guided vehicle (AGV) are:
[0182] p1(0)=[80,80]T, p2(0)=[360,360] T , The specific form of each automatic guided vehicle benefit function is as follows:
[0183]
[0184] according to Figure 3-4 It can be seen that compared with the proposed method, although both methods can make the position deviation of the agent tend to 0, the convergence time of the centralized differential game method with the introduction of trajectory optimization objectives is 105s, and the convergence time of the centralized differential game method with only obstacle penalties is 125s. Therefore, the centralized differential game method with the introduction of trajectory optimization objectives can reduce the time for the agent to complete the task, and the transportation time is reduced by 16% compared with the comparison method.
[0185] In order to illustrate the effectiveness of the proposed method in distributed mode, this example provides a directed communication topology diagram of three automatic guided vehicles, as shown in Figure 5 This example conducts a simulation experiment and gives a specific model expression of multiple automatic guided vehicles as follows:
[0186]
[0187] In the formula, p1(t) and are the position and speed information of the first automated guided vehicle at time t, p2(t) and are the position and speed information of the second automated guided vehicle at time t, p3(t) and They are the position and speed information of the third automated guided vehicle at time t, u1(t), u2(t) and u3(t) respectively represent the control strategies of automated guided vehicles 1, 2 and 3.
[0188] The initial position and transportation target position of each AGV are: p1(0) = [30,30] T , p2(0)=[370,370] T , p3(0)=[370,30] T , p3(0)=[30,370] T The sensing radius is R1=R2=48.
[0189] The specific form of each automatic guided vehicle benefit function is as follows:
[0190]
[0191] according to Figure 6 As shown in the figure, the proposed distributed differential game method can make the position deviation of each automatic guided vehicle approach 0, indicating that the task is completed.
[0192] In order to further illustrate the optimal coordination of the proposed distributed differential game method (DDG-TO), it is compared with the existing popular logistics warehousing methods, namely the negotiation-based method (FCFS) and the planning-based method (ETCEN-MCS).
[0193] Each AGV uses the above three methods for 10 times respectively and takes the average value. Figure 7 It can be seen that when the number of automated guided vehicles is 4, the transportation time of the three methods is similar. As the number of automated guided vehicles increases, the transportation time of the FCFS method slows down, followed by the ETCEN-MCS method, and the proposed distributed differential game method (DDG-TO) has the shortest transportation time, which is reduced by 15.7% and 9.2% respectively compared with the first two methods. In summary, the proposed distributed differential game method can enable each automated guided vehicle to complete the transportation task safely and can reduce the transportation time of each automated guided vehicle.
[0194] The above description is only a preferred embodiment of the present invention. All equivalent changes and modifications made according to the scope of the patent application of the present invention should fall within the scope of the present invention.
Claims
1. An optimal coordination method for logistics warehousing based on distributed differential game, characterized in that: The following steps are involved: Step S1: Based on the warehousing logistics scenario, using graph theory, establish the communication topology structure between the guided vehicles in the multi-automatic guided vehicle system, including the game participant model and the obstacle environment model; Step S2: using the artificial potential field method to design the operating cost function of the automatic guided vehicle in the game participant model; Step S3: The logistics warehousing coordination problem of multiple automated guided vehicle systems subject to perception and communication constraints is regarded as a distributed differential game problem, and a distributed differential game model is established; Step S4: using the Pontryagin minimum principle, analyzing the existence and uniqueness of the local optimal coordination strategy, solving the expression of the local optimal coordination strategy, giving the convergence conditions from the local optimal coordination to the global Nash equilibrium, and obtaining the optimal coordination solution; The game participant model is specifically: The specific form of the dynamic equation of the automated guided vehicle is: In the formula, p i (t) and are the position and speed information of the i-th automated guided vehicle at time t, u i (t) is the control input of the i-th automated guided vehicle at time t, B ii is the corresponding constant matrix; The specific form of the position error of the automated guided vehicle is: In the formula, is the expected transportation target point position of the i-th automated guided vehicle, is the position error between the current transport position of the i-th automated guided vehicle at time t and the expected transport target point; The specific form of the dynamic equation of the multi-automatic guided vehicle system is: In the formula, is the state of the automatic guided vehicle system at time t, is the state change rate of the automatic guided vehicle system at time t, N is the number of automatic guided vehicles, B i is the corresponding constant matrix, Establish a directed interactive topological graph G(v,ε) of N automated guided vehicles; Where v = {v1,...,v N } represents the set of automated guided vehicles, represents the set of edges, e ij represents the connection relationship between automatic guided vehicle i and automatic guided vehicle j, e ij ∈ε means that the automated guided vehicle i can receive information from the automated guided vehicle j, and the neighbor set of the automated guided vehicle i is The neighbor information matrix of automated guided vehicle i is defined as: Where I2 is a 2×2 unit matrix, is the Kronecker product, is the information matrix of neighboring guided vehicles j except for the automated guided vehicle i, where only the matrix F i The i-th row and the j-th row in are 1, and the rest are 0; The local dynamic equation of the automated guided vehicle i is defined as: In the formula, is the local state information of the automated guided vehicle i at time t, is the corresponding constant matrix; The obstacle environment model is specifically: Expand the outlines of obstacles of various shapes in logistics warehousing into ellipses and define the collision avoidance area S ik for: In the formula, r i is the safety radius of the automated guided vehicle i, is the radius of obstacle k, c k (t) is the center of mass position of obstacle k at time t, E k =I2 is the identity matrix; Define the sensing area D ik for: In the formula, R i is the sensing range of the automated guided vehicle i; Define the free area M ik for: The step S2 is specifically: Given the following preconditions: (1): The automatic guided vehicle will not slip during operation; (2): The directed interaction topology graph G(v,ε) between the AGVs is fixed and strongly connected; Design collision avoidance rules based on artificial potential field method: In the formula, the constant χ i (0<χ i <1) and is a constant, is the obstacle penalty function of the automated guided vehicle i at time t, which is defined as follows: is the distance penalty function of the automated guided vehicle i; for the trajectory optimization problem of the automated guided vehicle, the penalty function is defined as follows: In the formula, γ i is the deviation angle between the current trajectory of the automatic guided vehicle i and the predefined reference trajectory; The step S3 is specifically as follows: The expression of the distributed differential game cost function is as follows: In the formula, t f is the final time of the game of the automated guided vehicle i, u -i (t) is the control strategy set of neighboring automated guided vehicles except automated guided vehicle i at time t, is the operating cost function of the automatic guided vehicle i at time t, which is defined as u ij (t) is the control strategy of the neighboring automated guided vehicle j of automated guided vehicle i at time t, and U(t) is the control cost at time t, which is defined as Where R ii ,R ij are the adjustable positive definite weight matrices corresponding to the control strategies of the automated guided vehicle i and the neighboring automated guided vehicle j respectively; The goal of the coordinated control problem of distributed multi-AGV systems is to design the optimal coordination strategy for each AGV, and to ensure that the AGVs can safely reach the target point within a preset time. At the same time, AGV i and its neighbor AGVs can converge to a local Nash equilibrium, that is, the strategy set at time t is satisfy: In the formula, is the optimal cost function of the automated guided vehicle i at time t; The step S4 specifically includes: Assume that the additional state at time t is: In the formula, is the cost function at time t, is the initial condition, is the terminal moment cost; Assume the expansion state at time t is: The corresponding dynamic equation is as follows: In the formula, The distributed differential game problem is transformed into an optimal control problem, which can be expressed as follows: According to the Pontryagin minimum principle, the Hamiltonian function is defined as In the formula, λ i (t) is the Lagrange multiplier at time t; The local optimal coordination strategy satisfies the following set of differential equations, And its boundary conditions: p * (0)=p0, l i (t f )=0 (20) Where p0 is the initial position of the automatic guided vehicle system, λ i (t f ) is the Lagrange multiplier at the end; suppose the condition that the local optimal coordination strategy exists only: for the distributed differential game problem of automatic guided vehicles in a given warehouse logistics, the existing local optimal coordination strategy The satisfied form is transformed into the following matrix form: In the formula, For a given matrix, such as is a positive definite matrix, then there exists a unique local coordination strategy; Analyze the convergence of the local optimal coordination strategy to the global optimal strategy, which is defined as follows: For a differential game with N participants, if the following inequality holds, then the strategy set {u1(t),...,u i (t),...,u N (t)} converges to the global Nash equilibrium solution, In the formula, is the strategy of all the remaining automated guided vehicles except the automated guided vehicle i at time t. In addition, the strategy The following inequality is also satisfied: is the local optimal coordination strategy of the automated guided vehicle relative to its neighbors. Assuming that the topology graph between the automated guided vehicles is strongly connected, then: 1) for automated guided vehicle i, each local optimal coordination strategy is equal, 2) the local optimal coordination of the automated guided vehicle can converge to the global Nash equilibrium if and only if the communication topology graph of the automated guided vehicle is strongly connected.
2. The optimal coordination method for logistics warehousing based on distributed differential game according to claim 1 is characterized in that: The warehousing and logistics scenarios specifically include: 1) Task allocation: In the work environment, the staff assigns the order to transport the package to the corresponding appropriate AGV; 2) Package collection: After receiving the command, the AGV runs from the charging station to the corresponding shelf to pick up the package; 3) Package transportation: After the automated guided vehicle picks up all assigned packages, it transports the packages to the designated required location; 4) Return to the charging station: After the AGV completes the package delivery mission, it returns to the nearest charging station.
Citation Information
Patent Citations
Intelligent warehousing optimization method based on game theory
CN108171459A
Obstacle avoidance optimization method for warehouse logistics dispatching based on KM algorithm and artificial potential field method
CN108829105A
Intelligent storage route identification device and method used for multiple AGVs
CN110262408A
Cloud manufacturing multi-task scheduling optimization method based on game theory
CN110751293A
Intelligent unmanned logistics transportation method and system based on optimization efficiency evaluation algorithm
CN112686609A
