Multi-vehicle dispatching method based on genetic algorithm

Abstract

The present invention relates to the technical fields of target scheduling and warehousing logistics, and in particular to a multi-vehicle scheduling method based on a genetic algorithm. The method comprises: generating an initial solution using an improved repair method; taking the total time as the target, performing a roulette wheel to select the parent generation according to fitness, performing crossover and mutation operations, and generating offspring; repeating the improved repair in the initial solution for the offspring; merging the parent generation and offspring, and retaining individuals with high fitness to enter the next generation; taking the probability of conflict as the target, optimizing multi-vehicle trajectory planning through timestamp estimation and artificial potential field; taking the conflict cost and the total distance as the optimization targets, and performing target optimization through a non-dominated sorting genetic algorithm to obtain a non-dominated solution; normalizing the time and conflict cost, calculating the Euclidean distance between each solution and the ideal point, and selecting the solution with the smallest distance as the optimal solution. The method has the advantages of achieving multi-target optimization through non-dominated sorting and congestion distance calculation; and considering conflict risks while planning globally to plan reasonable routes.

Description

Technical Field

[0001] The present invention relates to the technical field of target scheduling and warehousing logistics, and in particular to a multi-vehicle scheduling method based on genetic algorithm. Background Art

[0002] To effectively solve scheduling problems, existing technical solutions have widely utilized optimization algorithms. These algorithms use mathematical models to simulate various factors in the vehicle scheduling process, such as assigned distance, task balance, and delivery time, to achieve globally optimal or near-optimal scheduling results. For example, particle swarm optimization, clustering algorithms, and genetic algorithms are used to optimize vehicle task allocation and route selection, ensuring that total travel distance and time are minimized. This optimization-based approach allows for flexible adjustment of scheduling strategies in complex and dynamic environments to adapt to changing operational requirements.

[0003] The multi-vehicle planning problem is a complex optimization problem with uniqueness and completeness constraints. Existing methods often need to consider global scheduling and specific path planning separately, making it difficult to consider the possibility of conflict in the overall planning. In order to avoid potential conflicts between vehicles, the space-time A , distributed constraint optimization, conflict-oriented search and other advanced path planning algorithms are introduced into the intelligent vehicle scheduling system. It is a path planning algorithm that adds a time dimension to the traditional A algorithm and is used to solve the path planning problem of multiple agents in a dynamic environment. The traditional A algorithm mainly focuses on the shortest path search in space, while the spatiotemporal A This approach considers not only spatial position but also states at different points in time, enabling better handling of moving obstacles and potential conflicts between multiple agents. Each agent in the distributed constrained optimization algorithm maintains its own local view and attempts to find the optimal action plan within its local constraints. Agents exchange information through message passing, gradually adjusting their strategies to achieve global coordination. Conflict-guided search employs a hierarchical planning strategy, first generating an initial set of conflict-free paths for all agents. Then, when conflicts are detected, the problem is broken down into subproblems and solved recursively.

[0004] However, the efficiency of existing global scheduling algorithms still needs to be improved, and they struggle to account for metrics such as allocation balance and the delivery priority of each task. Current conflict resolution algorithms can ideally plan routes with minimal conflicts. However, these methods only judge the effectiveness of a solution based on the occurrence or number of conflicts, assuming that each vehicle's schedule is precisely predictable. They fail to consider the risk of conflicts arising when vehicles fail to meet their scheduled times. In multi-vehicle delivery tasks, each vehicle must deliver to multiple destinations, and their schedules are even more unpredictable. These conflict resolution algorithms can easily fall short of their intended effectiveness. Furthermore, these algorithms often require step-by-step planning and conflict avoidance, making it difficult to achieve global optimization. Furthermore, there is currently a lack of algorithms that can plan specific trajectories for each vehicle in a factory environment, taking both delivery time and conflict risk into account. In summary, these aforementioned methods can only address either multi-vehicle scheduling or obstacle avoidance planning, but cannot simultaneously address task allocation and execution. Summary of the Invention

[0005] In order to solve the above problems, the present invention provides a multi-vehicle scheduling method based on genetic algorithm.

[0006] The present invention aims at a multi-vehicle dispatching method based on a genetic algorithm, which specifically comprises the following steps:

[0007] S1. Initial solution generation phase: Generate the initial solution using an improved repair method;

[0008] S2. Single-objective optimization: With total time as the goal, a roulette wheel is used to select parents based on fitness, perform crossover and mutation operations, and generate offspring. The improvement and repair operations in the initial solution are repeated on the offspring. The parents and offspring are merged, and individuals with high fitness are retained for the next generation.

[0009] S3. Multi-objective optimization: Using collision probability as the objective, we optimize multi-vehicle trajectory planning through timestamp estimation and artificial potential fields.

[0010] S4. Set the conflict cost and total distance as optimization objectives, and optimize the objectives through the non-dominated sorting genetic algorithm to obtain a non-dominated solution; normalize the time and conflict cost, calculate the Euclidean distance between each solution and the ideal point, and select the solution with the smallest distance as the optimal solution.

[0011] Preferably, step S3 specifically includes the following sub-steps:

[0012] S301. Calculate the travel time between two adjacent points based on the distance between the target points on the path and the vehicle speed; combine the task sequence and accumulate the time to obtain the estimated arrival timestamp of each target point;

[0013] S302. Split the map according to the coordinates of the two adjacent target points on each route to obtain a local map;

[0014] S303. Generate an artificial potential field based on the difference between the midpoint position and timestamp of two adjacent target points, and perform trajectory planning using the A-star algorithm under the local map;

[0015] S304. After trajectory planning is completed, a binary map of target points with conflict risks is obtained; factors causing potential conflicts are obtained based on the binary map, and the possibility of conflict is determined based on these factors.

[0016] Preferably, the calculation process of the A-star algorithm in step S303 is as follows:

[0017] Input a local map and an artificial potential field; convert the repulsive force of the artificial potential field into the cost function of the A-star algorithm; use the A-star algorithm to search for the minimum cost path between two points in the local map; and finally output a smooth trajectory without conflict.

[0018] The artificial potential field is expressed as follows:

[0019]

[0020] Where, represents repulsion, k and α represent parameters for adjusting the repulsion strength, Indicates the distance between the current position and the waypoint. t r timestamp representing the waypoint, t o Indicates the timestamp of the other vehicle's destination point.

[0021] Preferably, the factors causing potential conflict in step S304 include the conflict type, the length of the conflicting road section, and the conflict time;

[0022] The possibility of conflict is mainly characterized by the conflict cost, which is expressed as follows:

[0023]

[0024] Where C represents the conflict cost, k c represents the lateral conflict cost coefficient, k r represents the opposite conflict cost coefficient, represents the time-correlation attenuation coefficient of lateral conflict, represents the time correlation attenuation coefficient of the opposite conflict, l represents the length of the conflict section, 、 Indicates the timestamp of the current car and other target points at the intersection, 、 Indicates the timestamp when two cars enter the straight road conflict section. 、 Indicates the timestamp when the two vehicles leave the conflicting section on the straight road.

[0025] Preferably, the conflict types are mainly divided into lateral conflicts and opposite conflicts.

[0026] Preferably, the ideal point in step S4 is the point with the shortest time and the lowest probability of conflict;

[0027] The non-dominated sorting genetic algorithm specifically includes the following steps:

[0028] S401. Confirm the optimization goal; set algorithm parameters and constraints;

[0029] S402. Randomly generate an initial parent population and calculate the fitness of each individual in the initial parent population;

[0030] S403. Perform non-dominated sorting on the population; after the non-dominated sorting, a crossover mutation operator generates a progeny population and calculates the fitness function value of the progeny individuals;

[0031] S404. Merge the initial parent population and the offspring population, and perform non-dominated sorting and crowding calculation on the merged population;

[0032] S405. Using the elite selection strategy, select the top N individuals from the merged population as the new generation population;

[0033] S406. Determine whether the iteration termination condition is met. If so, terminate the calculation; otherwise, continue to repeat steps S404 and S405.

[0034] Preferably, the parameters in step S401 include population size, crossover probability, mutation probability and maximum number of iterations; the constraint condition is that each target point must be reached once to avoid duplication or omission;

[0035] The fitness in step S402 includes the total time and conflict cost;

[0036] In step S403, the crossover mutation operator performs the following operations:

[0037] Crossover operation: select two parent individuals and generate two offspring individuals;

[0038] Mutation operation: Randomly perturb individuals to increase population diversity;

[0039] Repair operation: Repair the generated offspring individuals to ensure that each target point is reached only once.

[0040] Preferably, step S2 specifically includes the following sub-steps:

[0041] S201. Fitness calculation: Perform fitness calculation and convert the total path distance into a fitness value;

[0042] S202. Selection and Reproduction: Roulette wheel selection of parent individuals; crossover and mutation operations on the parent individuals to generate offspring;

[0043] S203. Offspring repair: Repeat the improved repair operation in the initial solution on the offspring to ensure the legitimacy of the gene;

[0044] S204. Population update: Merge the selected parent and offspring generations, and retain individuals with high fitness to enter the next generation.

[0045] Preferably, step S1 specifically includes the following sub-steps:

[0046] S101. Initial population generation: Randomly generate initial pathways, allowing for gene duplication or deletion;

[0047] S102. Improve repair operations: delete duplicate genes; for deleted genes, calculate the cost of all possible insertion positions;

[0048] S103. Roulette wheel selection: Select the optimal insertion position based on cost probability, which speeds up iteration efficiency and prevents local optimality to a certain extent.

[0049] Preferably, the cost includes time cost, high priority delay, and task balance.

[0050] Compared with the prior art, the present invention can achieve the following beneficial effects:

[0051] (1) The present invention improves the gene repair operation by using a destruction operation and an improved repair operation, which greatly improves the efficiency compared with the traditional genetic algorithm.

[0052] (2) The present invention comprehensively considers time, priority, and task balance in multi-vehicle scheduling and prevents local convergence to a certain extent.

[0053] (3) The present invention performs route planning by predicting time and distance, thereby preventing the traditional dynamic obstacle avoidance algorithm from repeatedly planning and failing to achieve global optimality, and saving computing resources.

[0054] (4) Existing methods can reduce the number of conflicts when the vehicle travel time can be accurately predicted, but they will fail when it cannot be predicted. The present invention determines the possibility of conflict by the type and length of potential conflicts and the arrival time difference between the two vehicles, which can better deal with the situation where the running time of smart vehicles cannot be accurately predicted.

[0055] (5) The present invention combines single-objective and multi-objective optimization, and considers the possibility of conflict when planning the overall scheduling. However, the algorithm that first performs multi-vehicle planning and then obstacle avoidance planning often finds it difficult to take both indicators into account. In addition, the present invention has a faster speed than the multi-objective algorithm and ensures the effect of multi-objective planning.

[0056] In summary, the advantage of the present invention is that it can perform multi-objective algorithms, mainly realizing multi-objective optimization through non-dominated sorting and congestion distance calculation mechanisms, and can also use other mechanisms to achieve multi-objective optimization; it can take into account conflict risks while performing global planning and plan the most reasonable route. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 4 is a flowchart of a repair operation used when generating an initial solution according to an embodiment of the present invention.

[0058] Figure 2 It is an overall flow chart of the multi-objective optimization algorithm provided according to an embodiment of the present invention.

[0059] Figure 3 This is a flow chart of path planning for a single-segment task provided according to an embodiment of the present invention.

[0060] Figure 4 2 is a schematic diagram of a local map and estimated timestamps provided according to an embodiment of the present invention.

[0061] Figure 5 4 is a schematic diagram of the timestamps of the path points provided according to an embodiment of the present invention.

[0062] Figure 6 1 is a schematic diagram of lateral conflict and oncoming conflict provided according to an embodiment of the present invention; A represents lateral conflict, and B represents oncoming conflict.

[0063] Figure 7 3 is a schematic diagram of optimal solution selection provided according to an embodiment of the present invention. DETAILED DESCRIPTION

[0064] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. In the following description, identical modules are denoted by identical reference numerals. In the case of identical reference numerals, their names and functions are also identical. Therefore, their detailed description will not be repeated.

[0065] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and do not constitute a limitation of the present invention.

[0066] The present invention provides a multi-vehicle scheduling method based on genetic algorithm (flow chart as shown in FIG. Figure 1 ), involving a multi-objective optimization algorithm for the multi-vehicle scheduling problem in a factory environment. The algorithm is based on the genetic algorithm (GA) framework and incorporates the ideas of the non-dominated sorting genetic algorithm II (NSGA-II algorithm). The specific steps include:

[0067] S1. Initial solution generation phase: Generate the initial solution using the improved repair method; specifically, it includes the following sub-steps:

[0068] S101. Initial population generation: Randomly generate the initial path plan (chromosome encoding), allowing for gene duplication or deletion;

[0069] S102. Improve repair operations: delete duplicate genes (target points); for missing genes, calculate the cost of all possible insertion positions;

[0070] The costs include time cost, high priority delay, task balance, etc.

[0071] S103. Roulette wheel selection: Selects the optimal insertion position based on cost probability, which speeds up iteration efficiency and prevents local optimality to a certain extent;

[0072] As attached Figure 1 As shown in the figure, the quality of the initial solution of the genetic algorithm has a great influence on the iteration result. When generating the initial solution, the above-mentioned repair method will be used to generate it, which not only ensures randomness but also makes the solution as good as possible.

[0073] S2. Single-objective optimization: With total time as the goal, a roulette wheel is used to select the parent generation based on fitness (i.e., total distance), and then crossover and mutation operations are performed to generate offspring. The improvement and repair operations in the initial solution are repeated on the offspring. The parent generation and offspring are merged, and individuals with high fitness are retained for the next generation. This includes the following sub-steps:

[0074] S201. Fitness calculation: Perform fitness calculation and convert the total route distance into a fitness value; the shorter the distance, the higher the fitness (the single-objective optimization fitness is the total distance of all delivery routes).

[0075] S202. Selection and Reproduction: Roulette wheel selection of parent individuals; crossover and mutation operations on the parent individuals to generate offspring;

[0076] S203. Offspring repair: Repeat the improved repair operation in the initial solution on the offspring to ensure the legitimacy of the gene;

[0077] S204. Population update: Merge the selected parent and offspring generations, and retain individuals with high fitness to enter the next generation.

[0078] S3. Multi-objective optimization: the overall flow chart is as attached Figure 2As shown in the figure, collision probability is used as the second objective. Multi-vehicle trajectory planning is optimized through timestamp estimation and artificial potential field. The specific sub-steps are as follows:

[0079] S301. Calculate the travel time between two adjacent points based on the distance between the target points on the path and the vehicle speed; combine the task sequence (i.e., the path sequence) and accumulate the time to obtain the estimated arrival timestamp for each target point;

[0080] S302. Split the map according to the coordinates of the two adjacent target points on each route to obtain a local map;

[0081] S303. Generate an artificial potential field based on the difference between the midpoint position and timestamp of two adjacent target points, and perform trajectory planning using the A-star algorithm under the local map; the process of path planning for a single-segment task is as shown in the attached figure. Figure 3 As shown;

[0082] Specifically, the calculation process of the A-star algorithm is as follows: first, the local map and the artificial potential field are input; the repulsive force of the artificial potential field is converted into the cost function of the A-star algorithm (the area with high repulsive force has high cost); the A-star algorithm searches for the minimum cost path (balancing distance and obstacle avoidance) between two points (such as from point A to point B) in the local map; and finally outputs a smooth trajectory without conflict (such as the attached Figure 4 Path from 0 to 4);

[0083] Attachment Figure 4 The trajectory of the car represented by the path points 0 to 4 is planned, and the map is divided by the positions of two adjacent target points. The numbers in the dots are timestamps obtained based on the distance from each point to the starting point; Figure 5 The numbers in the waypoints in this map represent timestamps generated based on the distance from the starting point of the car represented by waypoints 0 to 4;

[0084] The artificial potential field is expressed as follows:

[0085]

[0086] Where, represents repulsion, k and α represent parameters for adjusting the repulsion strength, Indicates the distance between the current position and the waypoint. t r timestamp representing the waypoint, t o Indicates the timestamp of the target point of other vehicles;

[0087] S304. After trajectory planning is completed, a binary map of target points with conflict risk is obtained. Based on the binary map, factors causing potential conflicts are obtained, and the possibility of conflict is determined based on these factors.

[0088] Factors that may lead to potential conflicts include the type of conflict, the length of the conflict section, and the duration of the conflict. Conflict types are mainly divided into lateral conflicts and opposite conflicts ( Figure 6 );

[0089] The possibility of conflict is mainly characterized by the conflict cost, which is expressed as follows:

[0090]

[0091] Where C represents the conflict cost, k c represents the lateral conflict cost coefficient, k r represents the opposite conflict cost coefficient, represents the lateral conflict time correlation attenuation coefficient, represents the time correlation attenuation coefficient of the opposite conflict, l represents the length of the conflict section, 、 Indicates the timestamp of the current car and other target points at the intersection, 、 Indicates the timestamp when two cars enter the straight road conflict section. 、 Indicates the timestamp when the two vehicles leave the conflicting section on the straight road.

[0092] In this step, the probability of conflict is used as the second goal. The estimated time stamp of reaching each point is obtained by using the distance between targets and the task sequence. The map is segmented according to the coordinates of the positions of two adjacent points on each route to reduce the amount of calculation. An artificial potential field is generated based on the difference between the position and time stamp of the midpoint of two adjacent target points, and trajectory planning is performed using the A-star algorithm under the local map.

[0093] S4. Optimize the conflict cost and total distance using the non-dominated sorting genetic algorithm (NSGA-II) to obtain a non-dominated solution. Normalize the time and conflict cost, calculate the Euclidean distance between each solution and the ideal point (shortest time, lowest conflict probability), and select the solution with the smallest distance as the optimal solution (e.g. Figure 7 The purpose of this step is to solve the optimization problem with multiple conflicting objectives;

[0094] The non-dominated sorting genetic algorithm specifically includes the following steps:

[0095] S401. Confirm the optimization goal; set algorithm parameters and constraints;

[0096] Specifically, the parameters include population size, crossover probability, mutation probability, maximum number of iterations, etc.; the constraints are: each target point must be reached once to avoid duplication or omission;

[0097] S402. Randomly generate an initial parent population and calculate the fitness of each individual in the initial parent population, including total time and conflict cost;

[0098] S403. Perform non-dominated sorting on the population; after the non-dominated sorting, a crossover mutation operator generates a progeny population and calculates the fitness function value of the progeny individuals;

[0099] The specific crossover mutation operator performs the following operations:

[0100] Crossover operation: select two parent individuals and generate two offspring individuals;

[0101] Mutation operation: Randomly perturb individuals to increase population diversity;

[0102] Repair operation: Repair the generated offspring individuals to ensure that each target point is reached only once;

[0103] S404. Merge the initial parent population and the offspring population, and perform non-dominated sorting and crowding calculation on the merged population;

[0104] Non-dominated sorting: The values ​​of the root total distance and conflict cost determine the dominance relationship between all individuals in the population (that is, when a solution is not inferior to another solution in terms of two objectives and is superior to another solution in terms of at least one objective, the solution dominates the other solution). Based on this dominance relationship, the population is divided into multiple non-dominated levels. The first level consists of individuals that are not dominated by any other solution and are considered to be the best in the current population. Then, solutions that are not dominated by other individuals are screened from the remaining individuals to form the second level, and so on, until all individuals are assigned to the corresponding levels.

[0105] Crowding calculation: For each solution, first sort its total distance and conflict cost; then, for each sorted individual, calculate the sum of the squares of the normalized differences between it and the solutions of the two adjacent individuals as the crowding degree (the crowding degree of the boundary individual is infinite); the elite strategy will first select individuals with higher dominance levels, and will give priority to individuals with higher crowding degrees within the same dominance level.

[0106] S405. Using the elite selection strategy, select the top N individuals from the merged population as the new generation population;

[0107] S406. Determine whether the iteration termination condition is met. If so, terminate the calculation; otherwise, continue to repeat steps S404 and S405.

[0108] The number of iterations for single-objective optimization is 200 generations, and the number of iterations for multi-objective optimization is 30 generations. The iteration is terminated when the number of iterations meets the conditions.

[0109] Application scenarios of the non-dominated sorting genetic algorithm: This algorithm is suitable for multi-vehicle scheduling problems in a factory environment, especially when there are many target points and complex paths. It can effectively optimize the scheduling plan, reduce the probability of conflict, and improve overall efficiency.

[0110] This paper proposes an algorithm for the multi-vehicle scheduling problem in a factory. It takes into account both the total time and the probability of conflict, and can plan a fast and safe route for each vehicle. Its key technical points are as follows:

[0111] To improve efficiency, the repair operation of the traditional genetic algorithm was improved, duplicate genes were destroyed, and repairs were performed taking into account the increased distance, delivery delay and balance, which increased the optimization speed and improved the optimization effect.

[0112] Combining single-objective and multi-objective optimization, it improves the speed of algorithm iteration while ensuring the effectiveness of conflict prediction;

[0113] Route planning by predicting time and distance;

[0114] The possibility of conflict is determined by the type of potential conflict, the length of the conflict and the arrival time difference between the two vehicles.

[0115] It should be understood that the various forms of the processes shown above can be used to reorder, add, or delete steps. For example, the steps described in the present disclosure can be performed in parallel, sequentially, or in a different order, as long as the desired results of the technical solutions disclosed in the present disclosure can be achieved. This is not limited herein.

[0116] The above specific embodiments do not limit the scope of protection of the present invention. Those skilled in the art will appreciate that various modifications, combinations, sub-combinations, and substitutions may be made based on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention are intended to be included within the scope of protection of the present invention.

Claims

1. A multi-vehicle scheduling method based on genetic algorithm, characterized by: The specific steps include: S1. Initial solution generation phase: Generate the initial solution using the improved repair method; specifically, it includes the following sub-steps: S101. Initial population generation: Randomly generate initial pathways, allowing for gene duplication or deletion; S102. Improve repair operations: delete duplicate genes; for deleted genes, calculate the cost of all possible insertion positions; S103. Roulette wheel selection: select the optimal insertion position based on cost probability; S2. Single-objective optimization: With total time as the goal, a roulette wheel is used to select parents based on fitness, perform crossover and mutation operations, and generate offspring. The improvement and repair operations in the initial solution are repeated on the offspring. The parents and offspring are merged, and individuals with high fitness are retained for the next generation. S3. Multi-objective optimization: Using collision probability as the second objective, we optimize multi-vehicle trajectory planning through timestamp estimation and artificial potential fields. S4. Optimize the conflict cost and total distance as the optimization objectives and obtain a non-dominated solution through the non-dominated sorting genetic algorithm. Normalize the time and conflict cost, calculate the Euclidean distance between each solution and the ideal point, and select the solution with the smallest distance as the optimal solution. The ideal point is the point with the shortest time and the lowest conflict probability.

2. The multi-vehicle scheduling method based on genetic algorithm according to claim 1, characterized in that: The step S3 specifically includes the following sub-steps: S301. Calculate the travel time between two adjacent points based on the distance between the target points on the path and the vehicle speed; combine the task sequence and accumulate the time to obtain the estimated arrival timestamp of each target point; S302. Split the map according to the coordinates of the two adjacent target points on each route to obtain a local map; S303. Generate an artificial potential field based on the difference between the midpoint position and timestamp of two adjacent target points, and perform trajectory planning using the A-star algorithm under the local map; S304. After trajectory planning is completed, a binary map of target points with conflict risks is obtained; factors causing potential conflicts are obtained based on the binary map, and the possibility of conflict is determined based on these factors.

3. The multi-vehicle scheduling method based on genetic algorithm according to claim 2, characterized in that: The calculation process of the A-star algorithm in step S303 is as follows: Input a local map and an artificial potential field; convert the repulsive force of the artificial potential field into the cost function of the A-star algorithm; use the A-star algorithm to search for the minimum cost path between two points in the local map; and finally output a smooth trajectory without conflict. The artificial potential field is expressed as follows: Where, represents repulsion, k and α represent parameters for adjusting the repulsion strength, Indicates the distance between the current position and the waypoint. t r timestamp representing the waypoint, t o Indicates the timestamp of the other vehicle's destination point.

4. The multi-vehicle scheduling method based on genetic algorithm according to claim 2, characterized in that: The factors causing potential conflict in step S304 include the conflict type, the length of the conflicting road section, and the conflict time; The possibility of conflict is characterized by the conflict cost, which is expressed as follows: Where C represents the conflict cost, k c represents the lateral conflict cost coefficient, k r represents the opposite conflict cost coefficient, represents the lateral conflict time correlation attenuation coefficient, represents the time correlation attenuation coefficient of the opposite conflict, l represents the length of the conflict section, 、 Indicates the timestamp of the current car and other target points at the intersection, 、 Indicates the timestamp when two cars enter the straight road conflict section. 、 Indicates the timestamp when the two vehicles leave the conflicting section on the straight road.

5. The multi-vehicle scheduling method based on genetic algorithm according to claim 4, characterized in that: The conflict types are divided into lateral conflicts and opposite conflicts.

6. The multi-vehicle scheduling method based on genetic algorithm according to claim 1, characterized in that: The non-dominated sorting genetic algorithm in step S4 specifically includes the following steps: S401. Confirm the optimization goal; set algorithm parameters and constraints; S402. Randomly generate an initial parent population and calculate the fitness of each individual in the initial parent population; S403. Perform non-dominated sorting on the population; after the non-dominated sorting, a crossover mutation operator generates a progeny population and calculates the fitness function value of the progeny individuals; S404. Merge the initial parent population and the offspring population, and perform non-dominated sorting and crowding calculation on the merged population; S405. Using the elite selection strategy, select the top N individuals from the merged population as the new generation population; S406. Determine whether the iteration termination condition is met. If so, terminate the calculation; otherwise, continue to repeat steps S404 and S405.

7. The multi-vehicle scheduling method based on genetic algorithm according to claim 6, characterized in that: The parameters in step S401 include population size, crossover probability, mutation probability and maximum number of iterations; the constraint condition is that each target point must be reached once to avoid duplication or omission; The fitness in step S402 includes the total time and conflict cost; In step S403, the crossover mutation operator performs the following operations: Crossover operation: select two parent individuals and generate two offspring individuals; Mutation operation: Randomly perturb individuals to increase population diversity; Repair operation: Repair the generated offspring individuals to ensure that each target point is reached only once.

8. The multi-vehicle scheduling method based on genetic algorithm according to claim 1, characterized in that: The step S2 specifically includes the following sub-steps: S201. Fitness calculation: Perform fitness calculation and convert the total path distance into a fitness value; S202. Selection and Reproduction: Roulette wheel selection of parent individuals; crossover and mutation operations on the parent individuals to generate offspring; S203. Offspring repair: Repeat the improved repair operation in the initial solution on the offspring to ensure the legitimacy of the gene; S204. Population update: Merge the selected parent and offspring generations, and retain individuals with high fitness to enter the next generation.

9. The multi-vehicle scheduling method based on genetic algorithm according to claim 1, characterized in that: The costs include time cost, high priority delay, and task balance.

Images