Dynamic decision making and planning method and system for multi-objective operation of train virtual coupling

Abstract

Disclosed are a dynamic decision making and planning method and system for multi-objective operation of train virtual coupling. The method comprises the steps: S1, designing an overall objective function and single objective functions of a lead vehicle and a following vehicle, and designing target weights in the overall objective function and the single objective functions, to obtain a weighted multi-objective function; S2, on the basis of non-convex optimization set decomposition of a Monte Carlo decision tree, converting a non-convex optimization problem into a convex optimization problem, to obtain a reachable set; and in the reachable set searched by the Monte Carlo decision tree, using an evolutionary multi-objective optimization algorithm to solve the weighted multi-objective function, obtaining a planning curve sequence of the lead vehicle and the following vehicle. The present invention can achieve overall multi-objective virtual coupling control, and has high control efficiency and accuracy.

Description

A dynamic decision-making and planning method and system for virtual train coupling and multi-objective operation

[0001] Cross-reference to related applications

[0002] This application is based on and claims priority to Chinese Patent Application No. 2024114077924, filed on October 9, 2024, entitled "A Dynamic Decision Planning Method and System for Virtual Coupling and Multi-Objective Operation of Trains", the full text of which is incorporated herein by reference as a part of this application. [Technical Field]

[0003] This invention mainly relates to the field of train virtual coupling technology, specifically to a dynamic decision-making and planning method and system for multi-objective operation of train virtual coupling. [Background Technology]

[0004] Virtual train coupling refers to controlling and managing two trains as a single unit without physical couplers. These two trains can accelerate, decelerate, and stop synchronously with a small distance between them. Virtual coupling cooperative control technology refers to the control techniques that enable two trains to accelerate, decelerate, and stop synchronously with a small distance between them. Train spacing protection can no longer be based on absolute braking distance; relative braking distance must be used. That is, when calculating the safe distance between the following train and the preceding train, the speed and rate of change of speed of the preceding train must be considered. Depending on the topology and control division of labor, it is mainly divided into centralized cooperative control and distributed cooperative control. Current virtual coupling planning technology mainly focuses on distributed cooperative control. This involves first defining the target of the lead train, designing a corresponding controller to control the lead train, and obtaining the lead train's planning curve. Then, based on this, the target of the following train is defined, and a corresponding controller is designed to control the following train.

[0005] Existing research has the following drawbacks: 1. Current virtual train coupling planning techniques mainly include centralized control and distributed control. Centralized control can only design overall objectives, while distributed control can only design individual objectives. However, in the planning process of virtual train coupling, both overall and individual objectives need to be considered to achieve multi-objective optimization. Therefore, a single overall or individual objective cannot achieve multi-objective optimization. 2. Virtual train coupling planning is a non-convex problem, and current train operation planning algorithms do not have good solutions for non-convex optimization problems. Discrete search methods explore the cost of each point, making it difficult to simultaneously consider the optimal sequence objective. Continuous function optimization methods are prone to getting trapped in local minima and producing poor solutions. Therefore, neither discrete search methods nor continuous function optimization methods can compute the optimal solution.

[0006] Therefore, fully considering the safety, punctuality, stability, and energy-saving objectives of virtual coupling from both a holistic and individual perspective, and achieving the optimal solution in a non-convex problem, so that the train has a good planning curve during virtual coupling, is a technical challenge in this field.

[0007] [Summary of the Invention]

[0008] To address the technical problems existing in the prior art, this invention provides a dynamic decision-making and planning method and system for multi-objective operation of virtual train coupling, which improves the control efficiency and accuracy of virtual train coupling.

[0009] To solve the above-mentioned technical problems, the technical solution proposed by this invention is as follows:

[0010] A dynamic decision-making and planning method for virtual train coupling and multi-objective operation includes the following steps:

[0011] S1. Design the overall objective function and individual objective functions of the lead vehicle and the follower vehicle, and design the weights of each objective in the overall objective function and the individual objective functions to obtain the weighted multi-objective function;

[0012] S2. Based on the decomposition of the non-convex optimization set using the Monte Carlo decision tree, the non-convex optimization problem is transformed into a convex optimization problem, and the reachable set is obtained. In the reachable set searched by the Monte Carlo decision tree, a multi-objective evolutionary optimization algorithm is used to solve the weighted multi-objective function, and the planning curve sequence of the lead car and the follower car is obtained.

[0013] Preferably, in step S1, the overall objective function includes energy saving J1, operating efficiency J2, and tracking efficiency J3.

[0014] Preferably, the energy efficiency J1 takes into account the work done by the train's traction force / electric braking force. The smaller this index is, the more energy-efficient the automatic driving operation is. The specific calculation formula is as follows:

[0015] Where M is the total running time; F i Traction / electric braking force during the operating period; S i This represents the distance traveled within the specified operating time period.

[0016] Preferably, the operating efficiency J2 takes into account the time required for the entire train journey. The shorter the time required, the higher the average train speed within that section, and the higher the train operating efficiency. The specific calculation formula is as follows:

[0017] Among them, V i V represents the current operating speed. i+1 The operating speed at the next moment; S i This represents the distance traveled at the current moment.

[0018] Preferably, the tracking efficiency J3 is calculated using the formula: J3 = min||S m -S d || 2 ,S d <S m

[0019] Among them, S m For safe protection distance, S d The parking interval between vehicles.

[0020] Preferably, in step S1, the single objective function includes safety J4 and stationarity J5.

[0021] Preferably, the formula for calculating the security J4 is:

[0022] Among them, V i V represents the current operating speed. i,LKJ To set the LKJ speed limit value.

[0023] Preferably, the formula for calculating the stability J5 is:

[0024] Where, ΔF i This represents the change in traction / braking force.

[0025] Preferably, in step S2, the nonconvex optimization set decomposition based on the Monte Carlo decision tree includes the following steps:

[0026] Selection: Starting from the root node, recursively select the optimal child node until a leaf node is reached;

[0027] Expansion: If the current leaf node is not the terminal node, then create one or more child nodes and select one of them to expand;

[0028] Simulation: Starting from the extended node, run the output of a simulation until the game ends;

[0029] Backtracking: Using the output of the simulation, backtracking propagates to update the current action sequence.

[0030] Preferably, in step S2, a multi-objective evolutionary optimization algorithm is used to solve the weighted multi-objective function to obtain the optimal acceleration a and the final velocity v. Z Guarantee minimum runtime T opt Minimum operating energy consumption E opt And stability indicators, thus obtaining the planning curve sequence of the lead car and the follower car.

[0031] Preferably, the specific process of solving the weighted multi-objective function using a multi-objective evolutionary optimization algorithm is as follows:

[0032] Given state constraints, control constraints, and control objectives, solve the constrained multi-objective function to obtain the optimal solution; transform the time domain and position domain into the velocity domain, with the velocity domain corresponding to the position, i.e., the control variable;

[0033] For solving the fitness function f of a multi-objective evolutionary algorithm fit The process of finding the Pareto optimal solution set is a typical process of solving for the Pareto solution set; based on this theory, the Pareto optimal solution set X is derived. opt This is also known as the Pareto front; then, the fitness function f satisfying the iterative genetic algorithm is solved in the Pareto front. fit The optimal solution.

[0034] The present invention also discloses a computer program product, comprising a computer program that, when executed by a processor, performs the steps of the method described above.

[0035] The present invention further discloses a computer-readable storage medium having a computer program stored thereon, the computer program executing the steps of the method described above when run by a processor.

[0036] The present invention also discloses a computer device including a memory and a processor interconnected thereon, wherein the memory stores a computer program that, when run by the processor, performs the steps of the method described above.

[0037] Compared with the prior art, the advantages of the present invention are as follows:

[0038] This invention, in constructing a hierarchical multi-objective function, comprehensively considers both the overall performance and individual characteristics of the virtual train coupling formation. At the first level, from an individual perspective, objective functions are constructed for the lead car and following cars to ensure safety, improve efficiency, maintain stability, and save energy. At the second level, both are considered as a unified formation, and another set of objective functions for safety, efficiency, stability, and energy saving is constructed from an overall perspective. By assigning appropriate weights to these objectives, a weighted multi-objective function is formed to achieve coordinated control of the overall formation, thereby optimizing the operational effect of the virtual coupling.

[0039] In terms of dynamic decision-making and programming, this invention employs a coarse search method based on Monte Carlo decision trees. This method can effectively decompose the non-convex solution set into an optimization problem in a convex space, reducing the complexity of the problem. Subsequently, an evolutionary algorithm is applied to the convex solution set for detailed optimization to solve the hierarchical multi-objective function. This innovative solution process not only obtains the dynamic programming curve sequence of the lead car and the follower car, but also solves the non-convex multi-objective optimization problem, improving the control efficiency and accuracy of virtual train coupling. [Attached Image Description]

[0040] Figure 1 is a flowchart of the dynamic decision planning method of the present invention in an embodiment.

[0041] Figure 2 is a schematic diagram of the layered multi-target tracking distance of the virtual train coupling of the present invention.

[0042] Figure 3 is a schematic diagram of the non-convex optimization set decomposition based on Monte Carlo decision tree in this invention.

[0043] Figure 4 is a schematic diagram of the multi-objective evolutionary algorithm in this invention for solving the optimal sequence.

Detailed Implementation Methods

[0044] The present invention will be further described below with reference to the accompanying drawings and specific embodiments.

[0045] As shown in Figure 1, in terms of virtual coupling, this embodiment of the invention provides a hierarchical multi-objective dynamic decision-making and programming method for virtual train coupling. It constructs a safe, efficient and stable hierarchical multi-objective function from the overall and individual levels of the lead car and the following car. In the dynamic decision-making and programming part, it studies the coarse search method based on Monte Carlo decision tree to decompose the non-convex solution set, and uses an evolutionary algorithm to solve the two-level multi-objective function to obtain the dynamic programming curve sequence of the lead car and the following car.

[0046] The aforementioned hierarchical multi-objective dynamic decision-making programming method for virtual train coupling specifically includes the following steps:

[0047] S1. Design a hierarchical multi-objective function, including the following steps:

[0048] 1) Design of the overall objective function for the lead vehicle and the follower vehicle.

[0049] The virtual coupled train formation is viewed as a whole, with overall goals including energy efficiency, operational efficiency, and tracking efficiency.

[0050] Energy efficiency targets take into account the energy consumption during train operation; the energy consumption index mainly considers the work done by traction force / electric braking force. The smaller the index, the more energy-efficient the automatic driving operation.

[0051] The operational efficiency index mainly considers the time required for the entire operation. The shorter the time required, the higher the average speed of the train in that section, and the higher the train operation efficiency.

[0052] To achieve the goal of optimal tracking efficiency at a safe distance, it is necessary to design a tracking efficiency objective function.

[0053] 2) Design of individual objective functions for the lead vehicle and the follower vehicle.

[0054] For each individual unit in the virtual linkage, namely the lead vehicle and the follower vehicle, a multi-objective function for safety and stability must be satisfied. It is important to note that the objective functions for safety and stability need to be constructed for both the lead vehicle and the follower vehicle.

[0055] 3) Multi-objective weight design and weighted multi-objective function design.

[0056] Since the absolute values ​​of the factors considered by each indicator are different and the emphasis may be on specific performance under different conditions, we consider setting weights for each indicator and adding them as coefficients to each indicator to prevent the solution process from getting trapped in a local optimum due to the excessive prominence of a certain indicator value.

[0057] S2, Dynamic Decision Programming, includes the following steps:

[0058] 1) Non-convex optimization set decomposition based on Monte Carlo decision tree.

[0059] The planning objective of virtual train coupling is to calculate an optimal speed curve that satisfies the constraints. The planning of virtual train coupling is a mathematical solution process, and the objective function and constraints are both non-convex. Therefore, virtual train coupling planning is a non-convex problem. Non-convex problems can have multiple possible solutions, and these solutions are relatively independent. However, obtaining a globally optimal solution during the search is difficult, and it is easy to enter local optima.

[0060] The Monte Carlo decision tree is designed using a selection-expansion-simulation-backtracking pattern. A heuristic function is designed to progressively search the solution space, decomposing the non-convexity and transforming the optimization of the non-convex space into the optimization of the convex space, thus obtaining the reachable set.

[0061] 2) Multi-objective evolutionary optimization algorithm.

[0062] A multi-objective evolutionary optimization algorithm is designed to solve the hierarchical multi-objective function in the reachable set of Monte Carlo decision tree search, so as to obtain the optimal acceleration and final velocity, ensuring the minimum running time, minimum running energy consumption and stability index.

[0063] In the construction of hierarchical multi-objective functions, this invention includes the overall and individual aspects of the virtual train coupling formation. The first layer considers the objectives of safe, efficient, stable, and energy-saving train operation, and constructs the objective functions for the lead car and following cars from the perspective of individual cars. The second layer treats the lead car and following cars as a whole formation and constructs the objective function for safe, efficient, stable, and energy-saving from the overall perspective. Finally, the weights of multiple objectives are designed to form a weighted multi-objective function, which can realize the overall multi-objective virtual coupling collaborative control.

[0064] In designing dynamic decision-making programming, this invention uses a coarse search method based on Monte Carlo decision trees to decompose the non-convex solution set into an optimization problem in a convex space. An evolutionary algorithm is then used to solve the hierarchical multi-objective function on the convex solution set, resulting in a dynamic programming curve sequence for the lead car and the follower car. This allows for the solution of the multi-objective function and the resolution of the non-convex multi-objective optimization problem.

[0065] To better understand the above technical solution, the following will provide a detailed explanation of the technical solution in conjunction with the accompanying drawings and specific implementation methods:

[0066] The first step is to design the overall objective function for the lead vehicle and the follower vehicles:

[0067] Virtual train platooning requires calculating the "optimal" speed-distance operating curve for the platoon, taking into account constraints such as CTC information, track information, and vehicle information, while comprehensively considering punctuality, comfort, and energy conservation. Virtual platooning is a multi-objective optimization process; therefore, a multi-objective function for energy conservation, operational efficiency, and tracking efficiency needs to be designed for its overall implementation. Specifically:

[0068] (1) Energy efficiency J1

[0069] The energy consumption index considers the energy consumption during train operation. It primarily considers the work done by traction / electric braking force. The smaller this index, the more energy-efficient the automatic driving operation. Specifically:

[0070] In the formula: M is the total running time; F i Traction / electric braking force during the operating period; S i This represents the distance traveled within the specified operating time period.

[0071] (2) Operating efficiency J2

[0072] The operational efficiency index mainly considers the time required for the entire operation. The shorter the time required, the higher the average train speed within that section, and the higher the train operation efficiency. Specifically:

[0073] In the formula: V iV represents the current operating speed. i+1 The operating speed at the next moment; S i This represents the distance traveled at the current moment.

[0074] (3) Tracking efficiency J3

[0075] As shown in Figure 2, to achieve safe distance tracking of the target, the tracking target function is designed as follows: J3=min||S m -S d || 2 ,S d <S m (3)

[0076] In the formula: S m For safe protection distance; S d The parking interval between vehicles.

[0077] The second step is to design the individual objective functions for the lead vehicle and the follower vehicle:

[0078] For each individual unit in the virtual coupling, namely the lead vehicle and the follower vehicle, a multi-objective function for safety and stability must be satisfied. It should be noted that the following objective functions need to be constructed for both the lead vehicle and the follower vehicle, specifically:

[0079] (1) Security J4

[0080] From the perspective of train operation safety, this planning method is not allowed if any speeding occurs during operation. Therefore, if a speeding point is found in the planned control curve, the safety index is 1; otherwise, the safety index is 0.

[0081] In the formula: V i V represents the current operating speed. i,LKJ To set the LKJ speed limit value.

[0082] (2) Stability J5

[0083] Train stability primarily considers the changes in traction / braking force during train operation. Drastic changes in traction / braking force, especially under heavy loads, can easily cause jerking and surging. Therefore, research on train stability evaluation methods emphasizes that the handling stability index considers changes in traction / braking force during operation. A smaller index indicates a smoother change in train acceleration, better retention of the steering wheel position, and more stable operation. Specifically:

[0084] In the formula: ΔF i This represents the change in traction / braking force.

[0085] The third step is the design of multi-objective weights and hierarchical multi-objective functions:

[0086] Since the absolute values ​​of the factors considered by each indicator are different and the emphasis may be on specific performance under different conditions, we consider setting weights for each indicator and adding them as coefficients to each indicator to prevent the solution process from getting trapped in a local optimum due to the excessive prominence of a certain indicator value.

[0087] In summary, the overall evaluation index J is set as follows: J=k1J1+k2J2+k3J3+k4J4+k5J5 (6)

[0088] In the formula, k1, k2, k3, k4, and k5 represent the weights of safety, stability, energy efficiency, operating efficiency, and tracking efficiency indicators, respectively. If any speeding occurs, the control curve is considered unreasonable, so k1 is set to infinity. When speeding is prevented, stability, energy consumption, and operating efficiency are the primary considerations.

[0089] Step 4: Non-convex optimization set decomposition based on Monte Carlo decision tree:

[0090] The non-convex optimization set decomposition based on Monte Carlo decision trees can be divided into the following steps:

[0091] 1) Selection: Starting from the root node, recursively select the optimal child node, eventually reaching a leaf node.

[0092] 2) Expansion: If the current leaf node is not the terminal node, then create one or more child nodes and select one of them to expand.

[0093] 3) Simulation: Starting from the expansion node, run a simulation output until the game ends. For example, if you start from this expansion node, simulate ten times, and win nine times, then the expansion node will have a higher score, and vice versa.

[0094] 4) Backtracking: Using the results of the simulation in step 3), the backpropagation is used to update the current action sequence.

[0095] The specific implementation is shown in Figure 3. There exists a coordinate system, and the horizontal coordinate of the coordinate system is represented as S = [s0, s1, ..., s2]. i ,…,s N The horizontal axis represents the control level position U = [u0, u1, ..., u] i ,…,u N ]. Thus, all decisions u i The resulting vector is called the decision vector U. (i) The space consisting of feasible decision vectors is called the feasible decision vector space Ω.

[0096] Decision vector U (i) In each s i Corresponding to a velocity v i Therefore, for any decision variable U (i) Once determined, the state vector V = [v0, v1, ..., v] can be determined according to formula (1). i ,…,v N For example: based on the finite set sequence U of the initial decision vectors. (0) =[u0,u1,…,u i ,…,u N V is obtained by iterative calculation using formula (1). (0) =[v0,v1,…,v i ,…,u N ].

[0097] Step 5: Multi-objective evolutionary optimization algorithm:

[0098] A multi-objective evolutionary optimization algorithm is designed to solve the hierarchical multi-objective function within the reachable set searched by a Monte Carlo decision tree, thereby obtaining the optimal acceleration *a* and the final velocity *v*. Z Guarantee minimum runtime T opt Minimum operating energy consumption E opt And stability indicators.

[0099] Given state constraints, control constraints, and control objectives, an evolutionary algorithm is designed to solve the constrained multi-objective function and obtain the optimal solution. The time domain and position domain are converted into the velocity domain, and the velocity domain corresponds to the position, i.e., the control variable.

[0100] For solving the fitness function f of a multi-objective evolutionary algorithm fit The process of finding the Pareto solution set is a typical process. If... It is the constraint set of a multi-objective optimization model. If there is a solution x1, And for any k = 1, 2, 3, ..., we have f k (x1)≤f k If x1 is better than x2, then x1 is better than x2. If x1 is better than all other solutions in X, then x1 is the optimal solution for the objective function model. When solution x1 cannot satisfy all f... i When all (x) are optimal, but no solution is superior to x1, then x1 is the Pareto optimal solution of the objective optimization model. Based on this theory, the Pareto optimal solution set X can be derived. opt This can also be called the Pareto front; then, the fitness function f satisfying the iterative genetic algorithm is solved in this Pareto front. fit The optimal solution is shown in Figure 4.

[0101] This invention, in constructing a hierarchical multi-objective function, comprehensively considers both the overall performance and individual characteristics of the virtual train coupling formation. At the first level, from an individual perspective, objective functions are constructed for the lead car and following cars to ensure safety, improve efficiency, maintain stability, and save energy. At the second level, both are considered as a unified formation, and another set of objective functions for safety, efficiency, stability, and energy saving is constructed from an overall perspective. By assigning appropriate weights to these objectives, a weighted multi-objective function is formed to achieve coordinated control of the overall formation, thereby optimizing the operational effect of the virtual coupling.

[0102] In terms of dynamic decision-making and programming, this invention employs a coarse search method based on Monte Carlo decision trees. This method can effectively decompose the non-convex solution set into an optimization problem in a convex space, reducing the complexity of the problem. Subsequently, an evolutionary algorithm is applied to the convex solution set for detailed optimization to solve the hierarchical multi-objective function. This innovative solution process not only obtains the dynamic programming curve sequence of the lead car and the follower car, but also solves the non-convex multi-objective optimization problem, improving the control efficiency and accuracy of virtual train coupling.

[0103] This invention also discloses a computer program product, comprising a computer program that, when run by a processor, performs the steps of the method described above. This invention further discloses a computer-readable storage medium storing a computer program that, when run by a processor, performs the steps of the method described above. This invention also discloses a computer device comprising a memory and a processor interconnected, wherein the memory stores a computer program that, when run by a processor, performs the steps of the method described above. The products, media, and devices of this invention, corresponding to the methods described above, also possess the advantages described above.

[0104] The present invention can implement all or part of the processes in the methods of the above embodiments, or it can be implemented by hardware related to computer program instructions. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor, it can implement the steps of the above method embodiments. The computer program includes computer program code, which can be in the form of source code, object code, executable file, or some intermediate form. The computer-readable storage medium includes: any entity or device capable of carrying computer program code, recording media, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. The memory is used to store computer programs and / or modules. The processor implements various functions by running or executing the computer programs and / or modules stored in the memory, and by calling data stored in the memory. The memory may include high-speed random access memory, as well as non-volatile memory, such as hard disks, RAM, plug-in hard disks, smart media cards (SMC), secure digital (SD) cards, flash cards, at least one disk storage device, flash memory device, or other volatile solid-state storage devices.

[0105] The above are merely preferred embodiments of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should be considered within the scope of protection of the present invention.

Claims

1. A dynamic decision-making and planning method for multi-objective virtual train coupling operation, characterized in that, Including the following steps: S1. Design the overall objective function and individual objective functions of the lead vehicle and the follower vehicle, and design the weights of each objective in the overall objective function and the individual objective functions to obtain the weighted multi-objective function; S2. Based on the decomposition of the non-convex optimization set using the Monte Carlo decision tree, the non-convex optimization problem is transformed into a convex optimization problem, and the reachable set is obtained. In the reachable set searched by the Monte Carlo decision tree, a multi-objective evolutionary optimization algorithm is used to solve the weighted multi-objective function, and the planning curve sequence of the lead car and the follower car is obtained.

2. The dynamic decision-making and planning method for multi-objective virtual train coupling operation according to claim 1, characterized in that, In step S1, the overall objective function includes energy saving J1, operating efficiency J2, and tracking efficiency J3.

3. The dynamic decision-making and planning method for multi-objective virtual train coupling operation according to claim 2, characterized in that, The energy efficiency J1 considers the work done by the train's traction force / electric braking force. The smaller this index is, the more energy-efficient the automatic driving operation is. The specific calculation formula is as follows: Where M is the total running time; F i Traction / electric braking force during the operating period; S i This represents the distance traveled within the specified operating time period.

4. The dynamic decision-making and planning method for multi-objective virtual train coupling operation according to claim 2, characterized in that, The operating efficiency J2 takes into account the time required for the entire train journey. The shorter the time required, the higher the average train speed within that section, and the higher the train operating efficiency. The specific calculation formula is as follows: Among them, V i V represents the current operating speed. i+1 The operating speed at the next moment; S i This represents the distance traveled at the current moment.

5. The dynamic decision-making and planning method for multi-objective virtual train coupling operation according to claim 2, characterized in that, The formula for calculating the tracking efficiency J3 is: J3 = min||S m -S d || 2 ,S d <S m Among them, S m For safe protection distance, S d The parking interval between vehicles.

6. The dynamic decision-making and planning method for multi-objective virtual train coupling operation according to any one of claims 1-5, characterized in that, In step S1, the single objective function includes security J4 and stationarity J5.

7. The dynamic decision-making and planning method for multi-objective virtual train coupling operation according to claim 6, characterized in that, The formula for calculating the security J4 is as follows: Among them, V i V represents the current operating speed. i,LKJ To set the LKJ speed limit value.

8. The dynamic decision-making and planning method for multi-objective virtual coupling operation of trains according to claim 6, characterized in that, The formula for calculating the stability J5 is as follows: Where, ΔF i This represents the change in traction / braking force.

9. The dynamic decision-making and planning method for multi-objective virtual train coupling operation according to any one of claims 1-5, characterized in that, In step S2, the nonconvex optimization set decomposition based on the Monte Carlo decision tree includes the following steps: Selection: Starting from the root node, recursively select the optimal child node until a leaf node is reached; Expansion: If the current leaf node is not the terminal node, then create one or more child nodes and select one of them to expand; Simulation: Starting from the extended node, run the output of a simulation until the game ends; Backtracking: Using the output of the simulation, backtracking propagates to update the current action sequence.

10. The dynamic decision-making and planning method for multi-objective virtual coupling operation of trains according to any one of claims 1-5, characterized in that, In step S2, a multi-objective evolutionary optimization algorithm is used to solve the weighted multi-objective function to obtain the optimal acceleration a and the final velocity v. Z Guarantee minimum runtime T opt Minimum operating energy consumption E opt And stability indicators, thus obtaining the planning curve sequence of the lead car and the follower car.

11. The dynamic decision-making and planning method for multi-objective virtual train coupling operation according to claim 10, characterized in that, The specific process of solving the weighted multi-objective function using a multi-objective evolutionary optimization algorithm is as follows: Given state constraints, control constraints, and control objectives, solve the constrained multi-objective function to obtain the optimal solution; transform the time domain and position domain into the velocity domain, with the velocity domain corresponding to the position, i.e., the control variable; For solving the fitness function f of a multi-objective evolutionary algorithm fit The process of finding the Pareto optimal solution set is a typical process of solving for the Pareto solution set; based on this theory, the Pareto optimal solution set X is derived. opt Also known as the Pareto frontier; then, in the Pareto Solving for the fitness function f that satisfies the iterative genetic algorithm in the frontier fit The optimal solution.

12. A computer program product, comprising a computer program, characterized in that, The computer program is executed by a processor to perform the steps of the method as described in any one of claims 1-11.

13. A computer-readable storage medium having a computer program stored thereon, characterized in that, The computer program, when run by a processor, performs the steps of the method as described in any one of claims 1-11.

14. A computer device comprising a memory and a processor interconnected thereon, the memory storing a computer program, characterized in that, The computer program, when run by a processor, performs the steps of the method as described in any one of claims 1-11.

Images