A Dynamic Task Scheduling Method, Product, Medium and Device for a Ring Track Transportation System

Abstract

The present application discloses a dynamic task scheduling method, product, medium and device for a ring track transportation system, relating to the technical field of track transportation. The method includes: obtaining basic data during the operation of the ring track transportation system; constructing a task scheduling framework by using a method based on periodic rescheduling according to the basic data; determining a dynamic task scheduling strategy for the ring track transportation system based on the task scheduling framework and a dynamic scheduling algorithm; the dynamic scheduling algorithm includes a reinforcement learning algorithm and a nearest neighbor strategy; operating a transport vehicle in the ring track transportation system according to the dynamic task scheduling strategy. The dynamic task scheduling method, product, medium and device for the ring track transportation system of the present application can effectively improve the transportation efficiency and flexibility of the transportation system.

Description

Technical Field

[0001] The present application relates to the technical field of rail transportation, and in particular to a dynamic task scheduling method, product, medium and device for a circular rail transportation system. Background Art

[0002] A circular rail transportation system consists of a single closed rail and multiple transport vehicles. As the main transportation equipment suitable for hilly and mountainous areas, improving its operation efficiency is the key to the development of hilly and mountainous agriculture. Due to the characteristics of the one-way operation and unique path of the transport vehicles, the subjectivity of manual decision-making easily causes problems such as empty running and blockage of the transport vehicles during the operation process, resulting in low efficiency of the transportation system.

[0003] To address this problem, in the prior art, in a paper titled "Static Scheduling Optimization Method for a Hilly and Mountainous Circular Monorail Transportation System", a static scheduling method is used to schedule the mountainous circular monorail transportation system. This method is only applicable to the one-time task allocation and scheduling of all transport vehicles and all goods at specific locations. Static task scheduling can only perform one-time task combination and allocation when all transportation tasks and transport vehicles are in the initial state at the loading point, so as to realize the scheduling of all tasks and transport vehicles. However, during the actual transportation process, there will be situations of suddenly emerging urgent tasks that need to be dispatched and newly arriving random tasks. This method obviously cannot meet the actual transportation requirements and lacks flexibility. Patent CN202311275155.1, an intelligent mountain orchard transportation system, proposes to solve the "carpooling" combination problem with the goal of minimizing the number of times a transport vehicle loads goods based on the simulated annealing algorithm, and uses the ant colony algorithm to solve the problem of transport vehicle task allocation to achieve effective scheduling of static tasks. This method is also a static scheduling method and cannot meet the actual transportation requirements. Patent CN202311670175.9, a scheduling method for a monorail transportation system based on an improved chicken swarm algorithm, proposes a static scheduling method that combines the simulated annealing algorithm and the improved chicken swarm algorithm to achieve effective scheduling of static tasks, but it also cannot meet the actual transportation requirements. Dynamic task scheduling can realize dynamic task scheduling for the situation where random goods arrive at the loading point and urgent tasks are issued, and the transportation scheduling efficiency will be higher and more in line with the actual transportation scenario. Therefore, it is urgent to study a dynamic task scheduling method to meet the actual transportation requirements and improve the transportation efficiency. Summary of the Invention

[0004] The purpose of the present application is to provide a dynamic task scheduling method, product, medium and device for a circular rail transportation system, which can effectively improve the transportation efficiency and flexibility of the transportation system.

[0005] To achieve the above object, the present application provides the following solutions:

[0006] In a first aspect, the present application provides a dynamic task scheduling method for a ring - shaped track transportation system, including:

[0007] Obtain the basic data during the operation of the ring - shaped track transportation system; the ring - shaped track transportation system is a transportation system with multiple loading points and a single unloading point as the task form; the basic data includes the status data of the transport vehicle and the status data of the loading point;

[0008] Construct a task scheduling framework according to the basic data by using a method based on periodic rescheduling;

[0009] Determine the dynamic task scheduling strategy of the ring - shaped track transportation system based on the task scheduling framework and a dynamic scheduling algorithm; the dynamic scheduling algorithm includes a reinforcement learning algorithm and a nearest - neighbor strategy;

[0010] Operate the transport vehicles in the ring - shaped track transportation system according to the dynamic task scheduling strategy.

[0011] In a second aspect, the present application provides a computer program product, including a computer program which, when executed by a processor, implements the dynamic task scheduling method of the ring - shaped track transportation system.

[0012] In a third aspect, the present application provides a computer - readable storage medium, on which a computer program is stored, and the computer program, when executed by a processor, implements the dynamic task scheduling method of the ring - shaped track transportation system.

[0013] In a fourth aspect, the present application provides a computer device, including: a memory, a processor, and a computer program stored on the memory and executable on the processor, and the processor executes the computer program to implement the dynamic task scheduling method of the ring - shaped track transportation system.

[0014] According to the specific embodiments provided by the present application, the following technical effects are disclosed:

[0015] The present application provides a dynamic task scheduling method, product, medium, and device for a ring - shaped track transportation system. For the two dynamic events of random arrival of transportation tasks and arrival of emergency tasks, a task scheduling framework is designed based on the method of periodic rescheduling, and a reinforcement learning algorithm and a nearest - neighbor strategy are respectively adopted on the basis of the designed task scheduling framework to determine the dynamic task scheduling strategy. Finally, the transport vehicles in the ring - shaped track transportation system are operated according to the dynamic task scheduling strategy, meeting the transportation requirements of random arrival of transportation tasks and arrival of emergency tasks in actual transportation, and improving the efficiency and flexibility of the dynamic task scheduling of the ring - shaped track transportation system. Description of the Drawings

[0016] To more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the accompanying drawings required for use in the embodiments. Obviously, the accompanying drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can be obtained based on these drawings.

[0017] Figure 1 Schematic structural diagram of the ring track transportation system provided by the present application;

[0018] Figure 2 Schematic flow diagram of the dynamic task scheduling method for the ring track transportation system provided by the present application;

[0019] Figure 3 Schematic diagram of the rescheduling method selection provided by the present application;

[0020] Figure 4 Schematic flow diagram of Framework 1 provided by the present application;

[0021] Figure 5 Schematic flow diagram of Framework 2 provided by the present application;

[0022] Figure 6 Schematic flow diagram of Framework 3 provided by the present application;

[0023] Figure 7 Schematic flow diagram of Framework 4 provided by the present application;

[0024] Figure 8 Trajectory diagram of the transport vehicle completing an emergency task provided by the present application;

[0025] Figure 9 Schematic flow diagram of Strategy 1 provided by the present application;

[0026] Figure 10 Schematic flow diagram of Strategy 2 provided by the present application;

[0027] Figure 11 Structure diagram of DQN provided by the present application;

[0028] Figure 12 Graph of the actual number of dispatched transport vehicles of different frameworks varying with the test number provided by the present application. Detailed implementation manners

[0029] The following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, rather than all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present application.

[0030] To make the above objects, features, and advantages of the present application more obvious and understandable, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0031] Aiming at the problem that the existing scheduling means cannot meet the actual transportation requirements, the present application takes the specific task form of multiple loading points and a single unloading point as the research object, and aims at the maximum full load rate, the shortest total task completion time, and the least number of times the transport vehicle visits empty task points. Considering the characteristics of dynamic events of two types of task disturbances, namely random task arrival and emergency task arrival, a mathematical model is constructed and a task scheduling framework is designed respectively.

[0032] As Figure 1 shown, the present application provides a circular track transportation system, which is mainly applied to the cargo transportation in mountain orchards. In the figure, M 1 、M 2 、M 3 、....、M n are the numbers of the loading points, and E 1 、E 2 、E 3 、....、E n are the numbers of the transport vehicles (vehicles). The circular track transportation system needs to complete the transportation task in the form of multiple loading points and a single unloading point. Since the transport vehicle transports goods in a clockwise direction and can only move in one direction, the transport vehicle only needs to transport goods from multiple loading points to a single unloading point.

[0033] The above circular track transportation system further includes a central control system, which is used to realize functions such as data acquisition and storage, algorithm operation, and data calculation. Specifically, it includes:

[0034] Server: Responsible for data storage and the execution of calculation tasks. The server stores the real-time data and historical data of the transport vehicle and the loading point, and performs complex scheduling and optimization calculations.

[0035] Processor: Executes various algorithms and logics in the central control system, including carpooling combination algorithms, reinforcement learning algorithms, path planning, load calculation, etc.

[0036] Storage device: Used to store the operation data, task data, and historical records of the system. The storage device can be a physical hard disk, a solid-state drive, or cloud storage.

[0037] Wireless communication module: Ensures real-time data transmission between the central control system and each transport vehicle and loading point. The wireless communication network uses a cellular communication network (such as 4G, 5G) or other wireless technologies (Wi-Fi).

[0038] Software components: including operating systems, database management systems, middleware, and various application software. The software components are responsible for data collection, processing, analysis, and display.

[0039] Based on the above-mentioned circular track transportation system and central control system, as Figure 2 shown, this application discloses a dynamic task scheduling method for a circular track transportation system, including:

[0040] Step 1: Obtain the basic data during the operation of the circular track transportation system.

[0041] Number the multiple transport vehicles and multiple loading points of the mountain circular track transportation system. Set an RFID reader, a GNSS system, and a load sensor on each transport vehicle, and obtain the status data of each transport vehicle through the RFID reader, the GNSS system, and the load sensor. The status data of the transport vehicle includes the cargo load of the transport vehicle, the cargo loading status of the transport vehicle, and the real-time position of the transport vehicle. Set a load sensor and an RFID tag at each loading point, and obtain the status data of each loading point through the load sensor and the RFID tag. Among them, the load sensor is used to monitor and record the cargo status data of the loading point in real time, and the RFID tag is used to identify the unique identity and the precise position on the track of each loading point. The status data of the loading point includes the precise position on the track of the loading point, the situation of the loading and unloading cargo volume at the loading point, and the current cargo volume status at the loading point. Obtain the status data of the transport vehicle and the status data of the loading point during the operation of the mountain circular track transportation system as the basic data.

[0042] After obtaining the above status data, transmit these status data to the central control system in real time through a wireless communication network (Wi-Fi, cellular network). The central control system records and processes these status data, and at the same time sends the arrival time at the loading point, the task completion status, and the task arrangement status to the mobile terminal.

[0043] Specifically, the RFID reader of each transport vehicle reads the RFID tags arranged along the track, and combines with the fusion positioning system of the GNSS device to obtain the real-time position of the transport vehicle, and the central control system updates the position of the transport vehicle in real time. The load sensor equipped on the transport vehicle monitors the real-time cargo load. RFID tags are arranged at each loading point and the initial position. After the transportation task is planned, the task plan of each loading point is input into the central control system.

[0044] In the central control system, the real-time position of each transport vehicle can be viewed through the integrated positioning system of RFID and GNSS equipment, and the cargo volume and cargo status of the transport vehicle can be obtained through the load sensor. The central control system will calculate and transmit the dispatch instructions to each transport vehicle. These dispatch instructions include the route planning of the transport vehicle, the loading point number, the loading volume, the estimated arrival time, the time to arrive at the loading point, the task completion status and the task arrangement status.

[0045] Step 2: Build a task scheduling framework based on the basic data using a periodic rescheduling method.

[0046] The scheduling framework with random arrival of tasks as dynamic events is composed of a rescheduling method and a carpooling combination algorithm. The rescheduling method is used to determine the specific content of the rescheduling operation and the execution location and time point. This application selects different rescheduling methods to construct 4 scheduling frameworks, with the maximum full load rate, the shortest total task completion time, and the minimum number of transport vehicles visiting empty task points as the objective function. Simulation tests are carried out based on different task arrival rates (the number of new tasks generated per unit time is different). Based on 6 groups of simulation tests with different task arrival rates and the same carpooling combination algorithm, the performance of the 4 scheduling frameworks constructed is compared. The comparison shows that Framework 3, which pre-allocates global loading points and loading quantities based on the periodicity at the initial position and actually allocates the loading quantities at the loading points, has the best performance and is determined as the task scheduling framework. The specific framework structure of Frameworks 1, 2, and 4 and their performance verification process are detailed in the simulation experiment and result analysis section. I will not go into details here, and only describe the specific content of Framework 3, which is determined as the task scheduling framework.

[0047] Frame 3

[0048] Re-dispatch location: initial location and target loading point.

[0049] Rescheduling strategy (trigger condition): At the initial position, it is triggered based on the period, that is, rescheduling is performed at regular intervals, which means setting a fixed time interval and re-evaluating and adjusting all transportation tasks at each interval; at the target loading point, it is triggered when the transport vehicle arrives at the predetermined loading point.

[0050] Rescheduling parameters: At the initial position, determine the departure time, global loading point and pre-allocated loading quantity; at the target loading point, re-plan the actual loading quantity based on the current task volume.

[0051] Rescheduling content: including re-determining the loading point number and loading quantity of each transport vehicle. At the end of each cycle, all transport vehicles at the initial position will receive new scheduling instructions.

[0052] Determination of global loading points and loading quantities: In each cycle, the central control system calculates the globally optimal allocation of loading points and loading quantities based on the quantity of goods at all current task points, the positions and loads of the transport vehicles, so as to ensure that the transport vehicles are as fully loaded as possible.

[0053] Frame 3 performs rescheduling based on the cycle at the task docking points. Rescheduling can be carried out at the initial position and the target loading points, and there are two rescheduling positions. The rescheduling parameters corresponding to the two positions are different. At the initial position, the rescheduling parameters are the global loading points and loading quantities. The scheduling plan given at the initial position is only a preliminary allocation result. When the transport vehicle docks at the target loading point, the loading quantity of this loading point can be re-planned to maximize the transportation task of the same task point according to the existing task volume. The re-planned loading quantity is the actual allocation result. The arrival of new tasks will make the task volume at the current loading point more than the preliminary allocation, which will cause the transport vehicle to be fully loaded before docking at all the preliminary allocated loading points. Therefore, after the transport vehicle is fully loaded, it will no longer dock at the subsequent target loading points. When multiple transport vehicles are assigned to the same loading point and the arrival rate of goods is slow, the task volume at the current loading point will be less than the preliminary allocation, which will cause the transport vehicle not to be fully loaded after docking at all the preliminary allocated loading points. In this case, if there are still task points after the last preliminary allocated target loading point, the loading point number can be rescheduled at this loading point, and all subsequent loading points can be added to the task list of this trip to ensure that the transport vehicle is fully loaded to the maximum extent; if there are no task points, the transport vehicle has to waste its capacity and return to the initial position to unload the goods; therefore, when generating the global scheduling plan using the scheduling algorithm at the initial position, adjacent departing transport vehicles should be avoided from being assigned to the same task point as much as possible. The algorithm flow chart is as Figure 6 shown.

[0054] The specific implementation steps of Frame 3 are as follows:

[0055] 1) Deposit all transport vehicles at the initial moment into the queue q, and all transport vehicles are in an available state.

[0056] Specifically, at the initial moment, that is, at the initial position, there are multiple transport vehicles, and the vehicle numbers are stored in the queue q, and the time point t is set to 0.

[0057] 2) Calculate the maximum interval time t for the transport vehicle to run gap .

[0058]

[0059] where L is the total length of the circular track, V is the speed of a single transport vehicle, and c is the number of all transport vehicles performing tasks on the track at this time.

[0060] 3) Determine whether there are available transport vehicles at the initial position.

[0061] At this time, new task volumes are added to each loading point in the task set irregularly.

[0062] When the central control system of the transport aircraft obtains the information that a transport aircraft has completed its mission and returned to the initial position, and there are available vehicles at the initial position, it then determines whether the number of available transport vehicles is greater than 1. If there are available transport vehicles at the initial position at the current time t and the number of available transport vehicles is greater than 1, then according to the maximum interval time t gap Determine the departure time t of each transport vehicle start , remove all the transport vehicle numbers from the queue q and temporarily store them in the queue q' in sequence. When the central control system of the transport aircraft reads a transport aircraft that has not completed its mission and has not returned to the initial position, there are no available vehicles, and it directly jumps to step 4).

[0063] Calculate the departure time t start

[0064] For transport vehicle E, according to t gap and its position in the queue, calculate the departure time t of transport vehicle E start,E :

[0065] t start,E = t 0 + E · t gap (2)

[0066] Among them, t 0 is the initial time (usually 0).

[0067] Remove the vehicle numbers from the queue q

[0068] 1. Queue management: In the central control system, maintain a queue q of vehicle numbers to record all the current transport vehicles to be scheduled. When scheduling is required, remove the vehicle numbers from the queue q in sequence. The removal order is in accordance with the order in which the vehicles arrive at the initial position.

[0069] 2. Specific steps: The central control system removes the vehicle numbers from the queue q in sequence according to the calculated departure time t start , ensuring that the removal order matches the calculated departure time to avoid scheduling conflicts.

[0070] Temporarily store them in the queue q'

[0071] 1. Temporary storage queue: Temporarily store the removed vehicle numbers in another queue q' in sequence, which is used to store the vehicle numbers of the vehicles about to depart, ensuring that each vehicle departs according to the calculated t start departure.

[0072] 2. Scheduling Instruction Transmission: The central control system transmits the calculated scheduling instructions (including departure time, loading point number, and loading quantity) to each transport vehicle via a wireless communication network (Wi-Fi, cellular network). After receiving the instructions, the transport vehicles depart in sequence according to the calculated time and head to the designated loading points for loading.

[0073] In summary, this step has the following advantages: 1. Ensure orderly scheduling: By calculating the departure time t of each vehicle, ensure that each vehicle can depart as planned, avoiding congestion at the initial position. Orderly scheduling helps optimize the resource utilization rate of the loading points and improve the transportation efficiency. 2. Avoid conflicts: By removing the vehicle numbers in queue q and temporarily storing them in queue q', the scheduling order of the transport vehicles can be effectively managed, avoiding path conflicts caused by multiple vehicles departing simultaneously. This can ensure that each vehicle departs smoothly within its designated time and does not waste time waiting. 3. Increase the full-load rate: Calculate and allocate the loading points and loading quantities for each vehicle to ensure that each vehicle is as fully loaded as possible, improving the overall efficiency of the transportation system. Effectively managing the scheduling and path planning of the transport vehicles can reduce the phenomenon of running empty and further enhance the economic benefits of the transportation system. start Through the above steps, the dynamic scheduling of the transport vehicles in the circular track transportation system can be achieved, ensuring that each vehicle can complete the transportation task efficiently and orderly, and improving the overall operation efficiency of the system.

[0074] 4) Determine whether the current time t is equal to the departure time t of each transport vehicle.

[0075] If the current time point t = t, use the carpooling combination algorithm to obtain the global loading points and loading quantities of the transport vehicles, pre-allocate multiple transport vehicles at the initial position, and calculate the time points t when the transport vehicles reach each task point in the pre-allocation in sequence. Update the vehicle status, and remove the vehicle numbers from the temporary storage queue q' or queue q. If t is not equal to any t, proceed to step 5). start .

[0076] The dynamic scheduling problem with randomly arriving tasks as dynamic events has the same task form as the static scheduling problem. Therefore, the scheduling algorithm also includes two parts: the "carpooling" combination algorithm and the transport vehicle task allocation algorithm. The framework 3 of this application adopts a rescheduling strategy of simultaneously departing one transport vehicle per cycle, and only the carpooling combination algorithm is required. Therefore, in dynamic scheduling, only the carpooling combination algorithm for general goods needs to be designed. The carpooling combination algorithm of this application calculates the global loading points and loading quantities of the transport vehicles based on heuristic rules. start Since the static scheduling parameters are fixed, the task quantities of all task points are known and will not change, and the task set S arrival start start , proceed to step 5).

[0077]

[0078]

[0078] ​task The minimum task set is obtained through the carpooling combination algorithm. Both within the minimum tasks and in the minimum task sequence, they conform to the selection order of priorities. However, the final execution order of the minimum tasks is the result after global optimization by the transport vehicle task allocation algorithm. It can be seen that there is no constraint on the execution order among the minimum tasks obtained by static scheduling. But in dynamic scheduling, global optimization cannot be performed, and the minimum tasks with higher priorities should be completed earliest. Therefore, according to the initial generation order of the minimum tasks, select the minimum tasks to be completed based on the number of transport vehicles departing on this trip. The dynamic scheduling heuristic rules are as follows:

[0079] Rule 1: Taking adjacent positions as priorities, the main step is the task set S task Arrange in descending order according to the task point numbers.

[0080] Rule 2: Taking adjacent positions as priorities. The difference in the specific operation from Rule 1 is that the task set S task Is arranged in ascending order according to the task point numbers, and the rest of the operations are the same.

[0081] Rule 3: Taking the task volume as the priority.

[0082] Rule 4: Taking the task volume as the priority, with the largest task volume first.

[0083] Rule 5: Taking the number of times a task point is visited as the priority. During the operation of the framework, record the task point numbers that the transport vehicle has stopped at up to the current time point to obtain the set S of the number of times a task point is visited v ={M j :v j |M 1 :v 1 ,M 2 :v 2 ,...,M n :v n}, where v j Represents the number of times the task point with the number M j Is visited. The fewer times a task point is visited, the higher the priority. The specific operation is to arrange the set S of the number of times visited v In ascending order according to the task point numbers, select the combination of numbers whose sum of task volumes is not less than the load of a single transport vehicle in order, arrange the numbers within the combination in ascending order according to the task point numbers, update the task volumes of the remaining task points, and repeat the process of selection and update until the total task volume is less than the load of a single transport vehicle. The remaining task points are the final carpooling combination.

[0084] Rule 6: Taking the average generation rate of the task volume as the priority. During the operation of the framework, record the total task volume generated at each task point up to the current time point, use the total task volume of each task point to represent the average generation rate of the task volume at the task point, and obtain the set S r ={M j :rj |M 1 :r 1 ,M 2 :r 2 ,...,M n :r n},r j represents the average generation rate of the task volume for the task point numbered M. The faster the average generation rate of the task volume for the task point, the higher the priority. The specific operation is to sort the rate set S j in descending order according to the numbers of the task points, select the combination of numbers whose sum of task volumes is not less than the load of a single transport vehicle in sequence, arrange the combinations in ascending order according to the task point numbers, update the task volumes of the remaining task points, and repeat the processes of selection and update until the total task volume is less than the load of a single transport vehicle. The remaining task points are the final carpooling combinations. r

[0085] Through multiple experiments, it is determined that Rule 3 among the six rules has the best effect. Therefore, the global loading points and loading volumes of the transport vehicles are calculated according to Rule 3.

[0086] 5) Determine whether the current time t is equal to the time t of each loading point in the pre-allocation arrival 。

[0087] If the current time point t = t arrival , it means that the transport vehicle arrives at a certain loading point, reschedules the loading volume of this loading point, and removes the allocated task volume from the task set; if t is not equal to any t arrival , execute step 7).

[0088] Reschedule the loading volume

[0089] When the transport vehicle arrives at a certain loading point and needs to reschedule the loading volume, the following steps are required:

[0090] 1. Obtain the status of the current loading point: Obtain the location of the loading point through the fusion positioning of RFID and GNSS, and obtain the status of the cargo volume at the loading point through the load sensor at the loading point. Use the load sensor of the transport vehicle to obtain the status of the cargo load of the transport vehicle. These data are transmitted to the central control system through the wireless communication network. The central control system receives and processes these data.

[0091] 2. Calculate the new loading volume: The central control system calculates the cargo volume that the transport vehicle needs to load according to the remaining cargo capacity of the transport vehicle and the cargo volume at the loading point. The calculation result should ensure that the cargo load of the transport vehicle does not exceed its maximum cargo capacity.

[0092] ​3. Update the status of the transport vehicle and the loading point: The central control system sends the calculation result to the transport vehicle, and the transport vehicle loads goods according to the instruction. After the transport vehicle loads the goods, its load capacity and current position are updated. The quantity of goods at the loading point decreases by the quantity of goods loaded by the transport vehicle. The sensor updates the remaining quantity of goods at the loading point in real time and transmits the data back to the central control system.

[0093] Remove the assigned task quantity from the task set

[0094] After the transport vehicle completes the loading task, it is necessary to remove the assigned task quantity from the task set for the next task assignment and scheduling. The specific process of removing the task quantity is as follows:

[0095] 1. Task completion check: The central control system regularly checks the task completion status of the transport vehicle to determine whether the transport vehicle has completed all assigned tasks. If the transport vehicle has completed the task, the system records the task completion situation.

[0096] 2. Update the task set: The central control system removes the completed tasks from the task set and updates the status of the task set. Each task in the task set has a corresponding identifier, and the system removes the completed tasks according to the identifier. The updated task set is used for subsequent task assignment and scheduling.

[0097] 6) Determine whether the transport vehicle is fully loaded.

[0098] Determine whether the vehicle is fully loaded. If the transport vehicle is fully loaded at this time, this trip's rescheduling ends, and calculate the time t when the transport vehicle returns to the initial position renewal , record the scheduled transport vehicle numbers and the corresponding t in the order of rescheduling time at the initial position of the transport vehicle renewal After that, jump to step 7). If it is not fully loaded, if this loading point is the last pre-assigned loading point but not the last set loading point, add the subsequent task points to this trip's task list and calculate the time point t' to reach the added task point arrival After that, jump to step 7); if this loading point is the last pre-assigned loading point and is also the last set loading point, this trip's rescheduling ends, and calculate the time point t to reach the initial position renewal Jump to step 7); if this loading point is not the last pre-assigned loading point, then jump to step 7).

[0099] Implementation process of determining full load or not

[0100] 1. Obtain the load capacity of the current transport vehicle: Obtain the current load capacity status of the transport vehicle through the load sensor. The sensor will detect the total weight of the goods on the transport vehicle in real time and transmit the data to the central control system.

[0101] 2. Compare the load capacity and the maximum load-carrying capacity: The central control system compares the current load capacity of the transport vehicle with the maximum load-carrying capacity of the transport vehicle to determine whether the transport vehicle is fully loaded.

[0102] Specifically, a load sensor is used to detect the current load capacity of the transport vehicle. The detected load capacity is compared with the maximum load-carrying capacity of the transport vehicle. If the load capacity reaches the maximum load-carrying capacity, it is considered that the transport vehicle is fully loaded; otherwise, the transport vehicle is not fully loaded.

[0103] Implementation process of loading point judgment

[0104] 1. Determine whether the current loading point is the last pre-assigned loading point: The central control system records the task assignment order of each transport vehicle. By checking the position of the current loading point in the task list, it is determined whether it is the last pre-assigned loading point.

[0105] 2. Check the set last loading point: The last loading point set by the system is compared with the last loading point in the pre-assigned task to ensure that the transport vehicle executes the task according to the plan.

[0106] Specifically, the central control system records the order of the loading points in the task list. The current loading point is compared with the last loading point in the task list to determine whether it is the last pre-assigned loading point.

[0107] Implementation process of task addition and time point calculation

[0108] 1. Add a task to the current trip task list: If the transport vehicle is not fully loaded and the current loading point is not the last pre-assigned loading point, add the subsequent task points to the current trip task list.

[0109] 2. Calculate the time to reach the new task point: According to the current position and speed of the transport vehicle (the transport vehicle always runs at a constant speed), calculate the time to reach the new task point (calculated by combining the driving distance and speed of the transport vehicle).

[0110] Specifically, it is judged that the transport vehicle is not fully loaded and the current loading point is not the last pre-assigned loading point. Add the new task point to the task list. According to the current position of the transport vehicle, the position of the task point and the speed of the transport vehicle, calculate the time to reach the new task point.

[0111] In addition, the physical task scheduling process of the transport vehicle is as follows:

[0112] Parameter configuration: Set the track length, safety distance, task point position, and number of transport vehicles on the web side.

[0113] System debugging: Ensure that the functions of the transport vehicle are normal and control the transport vehicle to run to the corresponding initial position.

[0114] Task addition: The app mobile terminal adds tasks to the task list according to the quantity of goods at each task point, and selects the last task point as the unloading point.

[0115] Algorithm call: Call the algorithm to obtain the scheduling plan and return the result to the app terminal.

[0116] Goods transportation: The transport vehicle stops at the corresponding task point for loading according to the instruction. After loading is completed at each task point, scan the QR code on the vehicle body with the app, and enter the confirmation task completion instruction. Then the transport vehicle runs to the next loading point.

[0117] In summary, when performing this step, the following points should be specifically noted:

[0118] 1. The transport vehicle arrives at the loading point: At the current time point t = t arrival , the transport vehicle arrives at a certain loading point. The arrival time and position of the transport vehicle are recorded through sensors.

[0119] 2. Conduct re-scheduling of the loading quantity: Obtain the current quantity of goods at the loading point and the loading capacity of the transport vehicle. The central control system receives and processes this data. Calculate the loading quantity of the transport vehicle to ensure that it does not exceed its maximum loading capacity. The central control system sends the calculation result to the transport vehicle. The transport vehicle loads goods according to the instruction and updates its loading quantity and current position. The sensor updates the remaining quantity of goods at the loading point in real time and transmits the data back to the central control system.

[0120] 3. Determine whether the transport vehicle is fully loaded: Use the load sensor to obtain the current loading quantity of the transport vehicle. The central control system compares the current loading quantity with the maximum loading capacity of the transport vehicle to determine whether the transport vehicle is fully loaded.

[0121] 4. Loading point judgment: The central control system checks whether the current loading point is the last loading point in the pre-allocated tasks and whether it is the set last loading point.

[0122] 5. Task addition and time point calculation: If the transport vehicle is not fully loaded and the current loading point is not the last pre-allocated loading point, add the subsequent task points to the task list. Calculate the time to reach the new task point based on the current position and speed of the transport vehicle.

[0123] In this step, for each transport vehicle, according to the number of task points assigned, combined with the vehicle operation parameters, calculate the time point t renewal to reach the initial position, record the scheduled vehicle numbers and the corresponding t renewal in the chronological order of the re-scheduling time at the initial position of the transport vehicle, and ensure that the t renewal of the later-departing vehicle is not less than the t renewal of the earlier-departing vehicle. The detailed explanation is as follows:

[0124] Initialization and Parameter Acquisition

[0125] During system initialization, the current status of each transport vehicle, including its position and cargo capacity, is obtained through sensors and data acquisition devices. After manually determining the operating parameters, they are uploaded to the central control system via a mobile terminal. The operating parameters include the load capacity of the transport vehicle, the track length, the number and positions of task points, etc.

[0126] Calculate t renewal

[0127] For each transport vehicle, based on its current position and the positions of the task points, calculate the time required for the transport vehicle to complete the current task and return to the initial position.

[0128] t renewal = t current + H / V (3)

[0129] where t current is the current time, and H is the distance between the current position of the transport vehicle and the initial position.

[0130] Recording and Sorting

[0131] Record the serial numbers of the scheduled vehicles and their corresponding t in the order of the rescheduling time of the transport vehicles at the initial position renewal . Ensure that the t renewal of the vehicles that depart later is not less than the t renewal of the vehicles that depart earlier renewal . This process can be achieved by sorting and comparing the recorded t

[0132] 7) Determine whether the current time t is equal to the time t when the transport vehicle returns to the initial position renewal .

[0133] If t = t renewal , add the serial number of the transport vehicle that should reach the initial position at time t renewal to the queue q, and determine whether all tasks are completed. If the current time t is not equal to any t renewal , directly determine whether all tasks are completed

[0134] Determine whether all tasks are completed. If not, the scheduling continues, the time point t = t + 1 and jump to step 3); if so, the scheduling ends and output all scheduling schemes

[0135] Step 3: Determine the dynamic task scheduling strategy of the circular track transportation system based on the task scheduling framework and the dynamic scheduling algorithm

[0136] When the dynamic event is a task random arrival event, a dynamic task scheduling strategy for the circular orbit transportation system is determined based on the task scheduling framework and the reinforcement learning algorithm.

[0137] Reinforcement learning algorithm

[0138] Since the heuristic rule combines task points through a single rule and has poor flexibility, this application adopts a reinforcement learning algorithm to select the most feasible scheduling rule according to the current state, making up for the shortcoming that a single rule cannot be adjusted according to the environment.

[0139] Q_learning is a classic reinforcement learning algorithm based on the value function. It needs to maintain a Q-value table storing state-action pairs. The agent tries the selectable actions in each state and obtains the rewards feedback by the environment for the corresponding actions to update the Q-values. However, in practical problems, the state space is large and continuous, and the content stored in the Q-value table increases accordingly. It is difficult to traverse, resulting in Q_learning being difficult to obtain good results. DQN is a deep reinforcement learning model that combines deep learning and Q_learning. It uses state features as the input of the neural network and the Q-values of state-action pairs as the output, and can handle complex decision-making problems with a large and continuous state space. Its structure is as Figure 11 shown.

[0140] Therefore, this application adopts the DQN reinforcement learning algorithm to complete the carpool combination task. Its action space is the 6 heuristic rules. In addition, the state space and the reward and punishment function also need to be set.

[0141] State space design

[0142] The state space needs to summarize the most important features of the environment. The more features there are, the more comprehensively the state space represents the environment, and the more accurately the agent learns. In the carpool combination processing of dynamic scheduling, the task points that are preferentially combined will have their task volumes completed first. It can be seen that the relevant information of the task points needs to be reflected in the state space as much as possible, which is convenient for the agent to learn and determine the most effective scheduling rule in the current state during subsequent decision-making, and use this as the priority for carpool combination. The feature quantities related to the task points are: the current task volume task j , the average generation rate r j (expressed in the total task volume), the number of visits v to the task point j and the position O jThe fewer the number of task points where the transport vehicle has stopped when fully loaded, the better the effect of the scheduling rule. It can be seen that the task volume characteristic of the task point is directly proportional to the priority. The fewer the number of visits to the task point, the more the task volume accumulates. It can be seen that the number of visits to the task point characteristic is inversely proportional to the priority. In the case where the task point can be rescheduled in the dynamic scheduling framework, selecting a task point with a smaller number currently will increase the selection opportunities in subsequent rescheduling, and the probability of being fully loaded will also increase. It can be seen that the smaller the number of the task point location, the higher the priority. In order to reduce the dimension of the state space, the proportion of all characteristic quantities of the task point is superimposed, which is the relative urgency SU of the task point n , the current task volume and the average task volume generation rate are calculated according to formula (4), and the number of visits to the task point and the location are calculated according to formulas (4) to (6). The state space S is defined as formula (7).

[0143]

[0144]

[0145] S = [SU 1 , SU 2 ,..., SU n (7)

[0146] In the formula, j represents the index of the characteristic quantity. According to the context, the characteristic quantities related to the task point are: the current task volume, the average task volume generation rate (total task volume), the number of visits to the task point, the location, etc. are the original data of each characteristic quantity; is the proportion of the task volume characteristic of the current task volume and the average task volume generation rate; represents the supplementary part of the proportion of the number of visits to the current task point and the location characteristic quantity, which is used to balance the influence of the proportion of the task volume characteristic; is the proportion of the number of visits to the task point and the location characteristic quantity.

[0147] Design of the reward and punishment function

[0148] The design of the reward and punishment function is a key step in the reinforcement learning algorithm. The environmental feedback after the agent takes an action is calculated through the reward and punishment function. In dynamic scheduling, the carpool combination algorithm is called multiple times. The state space defines the state of each rescheduling point that needs to call the carpool combination algorithm. Therefore, the reward and punishment function evaluates the minimum task content finally executed by the transport vehicle at the corresponding rescheduling point. In order to guide the agent to optimize the objective function of the dynamic scheduling mathematical model for random task arrival events, the design of the reward and punishment function includes three-dimensional reward and punishment items, namely the full load rate rω, the number of loading times rn up , and the number of times the transport vehicle visits empty task points rn 0, its meaning is consistent with the objective function, but the calculation method is different. rω is the ratio of the actual transportation volume of the transport vehicle to the maximum transportation volume (load). In the carpooling combination algorithm, the number of loading times is used instead of the total task completion time as the evaluation criterion. The full load rate and the number of loading times and the number of times the transport vehicle visits empty task points have opposite optimization directions and different units. Therefore, the difference between 1 and rω is taken, and rn up and rω are weighted after being mapped to [0, 1], and weights are given according to the priority of the optimization objective. The reward and punishment function R is specifically defined as formula (8).

[0149]

[0150] In the formula, A, B, and C are the weights of the three reward and punishment items, and n is the number of task points.

[0151] When the dynamic event is an emergency task arrival event, the dynamic task scheduling strategy of the ring track transportation system is determined based on the task scheduling framework and the nearest strategy.

[0152] While ordinary goods arrive randomly, in order to efficiently handle emergency transportation tasks, two different emergency task response strategies (Strategy 1 and Strategy 2) are designed and compared. To ensure the fairness of the comparison, the same ordinary goods scheduling scheme is used as the basis, and the performance of each strategy in terms of the speed of responding to emergency tasks and the impact on the ordinary goods transportation plan is evaluated. Finally, Strategy 2, which can quickly respond to emergency tasks and minimize the interference to the ordinary goods transportation at the same time, is selected as the best strategy, so as to improve the overall efficiency of the ring track transportation system. The specific verification process can be seen in the simulation experiment and result analysis section.

[0153] The content of the emergency task is to transport agricultural machinery from one task point to another designated task point through a transport vehicle. The task priority is higher than that of ordinary goods in the task random arrival event and cannot be transported by carpooling with other tasks. The emergency task set is expressed as where t u is the time point when the u-th emergency task occurs, are the loading point and unloading point of this emergency task respectively and cannot be the same task point, and n u is the total number of emergency tasks. If the loading point number is less than the unloading point number the task flow direction is the same as the running direction of the transport vehicle, as shown by the Figure 8 red trajectory, and the emergency task can be completed without passing through the initial position; if the loading point number is greater than the unloading point number Since the transport vehicle can only travel in one direction and cannot reverse, the transport vehicle needs to be loaded as shown by the green trajectory and then reach the unloading point after passing through the initial position. When transporting ordinary goods, the main consideration is how to give a scheduling plan that allows the transport vehicle to be fully loaded with the fewest stops. When transporting urgent tasks, more attention is paid to the time when the task occurs, the starting and ending positions, and the status of each transport vehicle. The goal is to determine the transport vehicle number to complete the urgent task with the fastest speed and the least impact on the original transport plan of ordinary goods. The objective function f of the dynamic scheduling model with randomly arriving urgent tasks 3 is established as follows:

[0154]

[0155] G d = [(ω a - ω y ) (n y - n a )] (10)

[0156] min f 3 = [T d G d T (11)

[0157] In the formula, is the time for the system to respond to the u-th urgent task and determine the transport vehicle number. T d is the average difference between the system response time point and the urgent task occurrence time point, which is used to evaluate the response speed of the urgent task. The smaller the difference, the faster the system responds to the urgent task. A difference of 0 means immediate response; ω a , n a are the full load rate and the number of rescheduling times of the task with only randomly arriving ordinary goods without urgent tasks respectively. ω y , n y are the full load rate and the number of rescheduling times of the task with urgent tasks under the same randomly arriving ordinary goods task respectively. G d is used to evaluate the impact of the occurrence of urgent tasks on the original transport plan. The larger the two differences in G d , the greater the negative impact on the ordinary goods transportation. A difference of 0 means no impact.

[0158] The model assumptions and constraints are the same as those under the randomly arriving task event.

[0159] Before the occurrence of urgent tasks, the system only responds to the transportation requirements under the randomly arriving event of ordinary goods; after the occurrence of urgent tasks, the system gives priority to responding to urgent tasks and delays or re-plans the transport plan of ordinary goods; after the end of urgent tasks, the system resumes the priority response to ordinary goods. The urgent task scheduling framework is built on the basis of Framework 3, and two urgent task response strategies are proposed.​

[0160] Strategy 1

[0161] Strategy 1 responds to emergency tasks according to the scheduling rhythm of Framework 3. Framework 3 pre-assigns the global loading points and loading quantities of general goods to the transport vehicles at the initial position based on the cycle, adjusts the actual loading quantity or adds task points at the target loading point, and no rescheduling operations can be performed in other cases. It can be seen that Framework 3 can only determine the task type of the transport vehicle at the initial position, and the task of the transport vehicle has been assigned after it leaves. After an emergency task occurs, the first available transport vehicle that can be rescheduled at the initial position is used to give priority to completing the emergency task. After completing the emergency task, the transport vehicle needs to return to the initial position for the next rescheduling. When traveling empty through the task points located at and after the unloading point of the emergency task, it can transport general goods by the way. The algorithm flow chart is as Figure 9 shown.

[0162] The specific steps are briefly described as follows:

[0163] S1. Initialize the algorithm parameters.

[0164] S2. Update general tasks and emergency tasks irregularly.

[0165] S3. If there is a reschedulable vehicle at the initial position at the current time point t, transport general goods when there is no emergency task in the task set, and jump to S5; when there are both emergency tasks and general tasks, the transport vehicle gives priority to completing the emergency task. If the loading point of the emergency task is less than the unloading point, calculate the time to reach the loading point and unloading point of the emergency task and pre-assign the general goods at and after the unloading point, and jump to S5; if the loading point is greater than the unloading point, calculate the time to reach the loading point of the emergency task and the time to pass through the initial position with the emergency goods If there is no available transport vehicle, enter S4.

[0166] S4. If at the current time point the transport vehicle passes through the initial position, at this time, the time to reach the unloading point of the emergency task can be calculated and the general goods at and after the unloading point can be pre-assigned. If t is not equal to any enter S5.

[0167] S5. If at the current time point t the transport vehicle has completed this trip, it means that the transport vehicle returns to the initial position and can receive the next task; if not, enter S6.

[0168] S6. Judge whether all tasks are completed. If not, the scheduling continues, the time point t = t + 1 and jump to S2; if so, the scheduling ends and output all scheduling schemes.

[0169] Strategy 2

[0170] The aforementioned Strategy 2 is the nearest strategy. For a transport vehicle that can execute an emergency task after an emergency task appears, it only needs to meet the following requirements: the position status is less than the target loading point, and the general cargo load status is 0. To achieve the fastest response speed, the nearest strategy can be adopted to quickly determine the transport vehicle number and select the transport vehicle that meets the requirements and is closest to the target loading point to execute the current emergency task. There are three types of transport vehicles in Strategy 2 that can respond to emergency tasks: the same as Strategy 1, if there is a transport vehicle with reschedulable initial position, assign the current emergency task and the general cargo at the unloading point and subsequent task points to the transport vehicle; if the transport vehicle has been assigned general cargo, but the transport vehicle is in an empty load state at the current time point and is before the emergency task loading point, cancel the assigned general cargo, and the transport vehicle preferentially transports the current emergency task and the general cargo at the unloading point and subsequent task points; the transport vehicle has been assigned an emergency task. When the unloading point of the assigned emergency task meets the task point position requirement before the current emergency task loading point, and the loading point of the assigned emergency task is less than the unloading point and the transport vehicle position requirement before the current emergency task loading point, and the transport vehicle is not carrying general cargo, if the loading point of the assigned emergency task is greater than the current emergency task unloading point, the transport vehicle passes through the initial position and is before the current emergency task loading point and the transport vehicle is not carrying general cargo, then after completing the assigned emergency task, the transport vehicle continues to complete the current emergency task and the general cargo at the unloading point and subsequent task points; if there are multiple transport vehicles that meet the execution requirements, calculate the positions of each transport vehicle according to the operation parameters, and select the transport vehicle closest to the emergency task loading point to execute the current emergency task. The algorithm flowchart is as Figure 10 shown.

[0171] Step 4: Operate the transport vehicles in the mountain-ring rail transit system according to the dynamic task scheduling strategy.

[0172] When the transport task is manually input on the mobile device, it is processed and calculated by the server according to the appropriate task scheduling framework and algorithm, and then the task is released. After receiving the task, the transport vehicle executes the instruction to go to the loading point for loading, runs one week to complete the goods transportation at one or more task points, returns to the initial position to unload the goods into the storage point at the foot of the mountain, and repeats this process until the transportation tasks at each task point are completed.

[0173] In addition, the present application also provides a network layer structure based on the dynamic task scheduling method of the mountain-ring rail transit system of the present application, specifically as follows:

[0174] Data acquisition layer: Obtain the real-time position data of the transport vehicle and the loading point through the GNSS system and RFID, obtain the real-time load data and status data of the loading point through the load sensor at the loading point, and obtain the real-time load data and status data of the transport vehicle through the load sensor on the transport vehicle.

[0175] Data transmission layer: Uses a wireless communication network to transmit data and instructions.

[0176] Data processing layer: A central control system that runs scheduling algorithms for task allocation and status update.

[0177] Data storage layer: The database stores historical data and status information of the transport vehicles and loading points.

[0178] User interface layer: Provides a graphical interface to display real-time monitoring and scheduling results.

[0179] The implementation principles of each network layer structure are as follows:

[0180] The data acquisition layer obtains real-time data and transmits it to the central control system through the data transmission layer.

[0181] The data processing layer runs scheduling algorithms to perform task allocation and status update based on real-time data.

[0182] The task allocation results are sent to the transport vehicles through the data transmission layer, and the transport vehicles execute tasks according to the instructions.

[0183] The data storage layer saves historical data and status information for subsequent analysis and optimization.

[0184] The user interface layer displays real-time monitoring data and scheduling results to provide visual support.

[0185] Simulation experiment design and result analysis

[0186] The rescheduling method consists of three parts: rescheduling location, rescheduling strategy, and rescheduling parameters. Different operations are selected for each part, and different rescheduling methods correspond to different dynamic scheduling frameworks. According to the characteristics that tasks arrive randomly and the task volumes at each loading point are updated irregularly, four scheduling frameworks, referred to as frameworks for short, are proposed. The specific operation contents are as Figure 3 shown. Among them, the framework structures of each framework except Framework 3 are as follows:

[0187] Framework 1

[0188] Rescheduling location: Rescheduling is only performed at the initial location.

[0189] Rescheduling strategy (trigger condition): There are available vehicles at the initial location and there are tasks to be transported at the loading points.

[0190] Rescheduling parameters: Global loading points and loading volumes, the number of dispatched vehicles (global allocation).

[0191] The framework 1 is rescheduled based on events only at the initial position. The content of the rescheduling is to determine the number of dispatched vehicles, the loading point numbers where each transport vehicle stops, and the loading quantity. Since the transport vehicle can only be rescheduled at the initial position for each trip and there is no opportunity for adjustment during the operation, the loading point numbers and the loading quantity need to be determined in full at the initial position and cannot be changed subsequently. That is, the rescheduling parameters are the global loading points and the loading quantity, and the total loading quantity should make the transport vehicle fully loaded as much as possible. The algorithm flow chart is as shown in Figure 4 shown below.

[0192] The specific steps of the framework 1 are as follows:

[0193] S1. Initialize the algorithm parameters: The vehicle numbers are stored in the queue q, indicating that the vehicles are at the initial position and are in an available state. At the initial moment, all vehicles are in the queue q; the time point t is set to 0.

[0194] S2. New task quantities are added to each loading point in the task set at irregular intervals.

[0195] S3. If the current time point t meets the requirements for the occurrence of an event: there are available vehicles at the initial position and there are tasks to be transported at the loading point, then a rescheduling operation can be performed. Determine the number of dispatched vehicles based on the current total task quantity and the total number of available vehicles at the initial position, and use the scheduling algorithm to obtain the global loading points and the loading quantity of each dispatched transport vehicle; if the requirements for the occurrence of the event are not met, jump to S6.

[0196] S4. Each transport vehicle calculates the time point t to reach the initial position based on the number of task points assigned and in combination with the vehicle operation parameters renewal , and record the dispatched vehicle numbers and the corresponding t in the order of the rescheduling time of the transport vehicles at the initial position renewal .

[0197] S5. Update the vehicle status and the task set, remove the vehicle numbers from the queue q, and remove the assigned tasks from the task set; ensure that the t renewal of the later-departing vehicles is not less than the t renewal of the earlier-departing vehicles.

[0198] S6. If the current time point t = t renewal , it means that the transport vehicle returns to the initial position, and add the vehicle numbers that should reach the initial position at the t renewal time point to the queue q; if t is not equal to any t renewal , enter S7.

[0199] S7. Judge whether all tasks are completed. If not, the scheduling continues, the time point t = t + 1 and jump to S2; if so, the scheduling ends and output all scheduling plans.

[0200] Framework 2

[0201] Rescheduling Location: Rescheduling is only performed at the initial location.

[0202] Rescheduling Strategy (Trigger Condition): Triggered based on a cycle, i.e., rescheduling is performed at regular intervals.

[0203] Rescheduling Parameters: Global loading points and loading quantities, number of dispatched vehicles (same as in Framework 1).

[0204] Framework 2 performs rescheduling only at the initial location based on a cycle, and the rescheduling parameters are also the global loading points and loading quantities. The difference from Framework 1 is that the number of dispatched vehicles does not need to be considered during rescheduling. Only one transport vehicle can be dispatched from the initial location. When there are multiple transport vehicles at the initial location, the departure time of each transport vehicle needs to be determined. The departure time is the rescheduling time point using the scheduling algorithm. The longer the interval between the departures of adjacent transport vehicles, the more total tasks have been generated during rescheduling, and the easier it is for the transport vehicles to be fully loaded. Therefore, the maximum interval time is selected, and the interval time is evenly distributed among all vehicles on the track. When using the cycle-based rescheduling strategy, only one transport vehicle departs each time, and there is no part of the transport vehicle task allocation in the scheduling algorithm. Only "carpooling" combination processing needs to be performed, which will be described in detail below. The algorithm flowchart is as Figure 5 shown.

[0205] The specific steps of Framework 2 are as follows:

[0206] S1. Initialize the algorithm parameters: The vehicle numbers are stored in the queue q, indicating that the vehicles are at the initial location and are in an available state. At the initial moment, all vehicles are in the queue q; the time point t is set to 0.

[0207] S2. Determine the maximum interval time t between vehicle departures based on the track length L, the speed V of the transport vehicle, and the total number of transport vehicles c gap .

[0208] S3. New task quantities are added to each loading point in the task set at irregular intervals.

[0209] S4. If there are available transport vehicles at the initial location at the current time point t, rescheduling operations can be performed. When there are multiple transport vehicles, calculate the departure time t start of each transport vehicle. The departure time t start of the first transport vehicle is the current time point, and the departure time of the subsequent transport vehicles is separated from the previous one by t gap ; Remove all vehicle numbers from the queue q and temporarily store them in the queue q' in order, indicating that rescheduling times have been arranged for all transport vehicles at the initial location at the current time point, and there are no available vehicles in the queue q. When there is only one transport vehicle, the departure time t start is the current time point, and jump to S6. If there are no available transport vehicles, enter S5.

[0210] S5. If the current time point \(t = t start \), go to S6; if \(t\) is not equal to any \(t start \), jump to S9.

[0211] S6. Obtain the global loading point and loading quantity of the vehicle using the scheduling algorithm.

[0212] S7. Each transport vehicle calculates the time point \(t renewal \) to reach the initial position according to the number of task points assigned and the vehicle operation parameters, and records the scheduled vehicle numbers and the corresponding \(t renewal \) in the order of the rescheduling time of the transport vehicle at the initial position.

[0213] S8. Update the vehicle status and task set, remove the vehicle number from the temporary queue \(q'\) or queue \(q\), and remove the assigned tasks from the task set; ensure that the \(t renewal \) of the later-departing vehicle is not less than the \(t renewal \) of the earlier-departing vehicle.

[0214] S9. If the current time point \(t = t renewal \), it means that the transport vehicle returns to the initial position, and add the vehicle number that should reach the initial position at the \(t renewal \) time point to the queue \(q\); if \(t\) is not equal to any \(t renewal \), go to S10.

[0215] S10. Judge whether all tasks are completed. If not, the scheduling continues, the time point \(t = t + 1\) and jump to S3; if so, the scheduling ends and output all scheduling schemes.

[0216] Framework 4

[0217] Rescheduling positions: Initial position and target loading point.

[0218] Rescheduling strategy (trigger condition): At the initial position, triggered based on the cycle, the same as Framework 2; at the target loading point position, triggered when the transport vehicle reaches the predetermined loading point.

[0219] Rescheduling parameters: At the initial position, only determine the next target loading point and loading quantity, without global planning; at the target loading point position, re-plan the actual loading quantity of the current loading point and determine the next target loading point.

[0220] Framework 4 performs cycle-based rescheduling at the task stop point, and there is only a difference in rescheduling parameters from Framework 3. Framework 4 only determines the next target loading point and loading quantity at the initial position, and re-plans the loading quantity of the current loading point and the next target loading point after reaching this task point. The algorithm flow chart is as Figure 7 shown.

[0221] Framework 4 and Framework 3 differ only in the rescheduling parameters of the transporter in S4 and S6. The rest of the steps are the same, so only the specific operations of these two steps are given:

[0222] S4. Pre-allocate at the initial position. If the current time point t = t start , use the scheduling algorithm to obtain the next target loading point and loading quantity of the vehicle, and calculate the time point t when the transport vehicle arrives at each task point in the pre-allocation arrival , update the vehicle status and move the vehicle number out of the temporary queue q' or queue q; if t is not equal to any t start , enter S5.

[0223] S6: Determine whether the vehicle is fully loaded. If the vehicle is fully loaded, the rescheduling is completed and the process goes to S7. If the vehicle is not fully loaded, and the loading point is not the last loading point, rescheduling is performed, and the scheduling algorithm is called to determine the next target loading point for the vehicle, and the time point t to arrive at the added loading point is calculated. arrival Then jump to S8; if the loading point is the last loading point set, the current re-dispatching is completed and enters S7.

[0224] Experimental design

[0225] In the experiment under the dynamic event of random arrival of tasks, 6 groups of experiments were designed according to the different rates of ordinary goods arriving at the task point within a fixed time. Based on the same scheduling algorithm, the dynamic scheduling schemes obtained by 4 scheduling frameworks were compared with the full load rate as the main evaluation criterion, and the best scheduling framework was obtained; based on the optimal scheduling framework, the solution performance of 6 heuristic rules and reinforcement learning algorithms with heuristic rules as actions were compared. In the experiment under the dynamic event of emergency tasks, 5 groups of experiments were designed according to the time point and task type of the emergency tasks. The allocation of transport vehicles for emergency tasks was completed based on two response strategies, and the solution performance of the two strategies was compared and analyzed with the average response speed and the impact on the original plan as the objective function.

[0226] Parameter settings

[0227] The parameters of the transport vehicle in all the tests of this application are based on the remote-controlled electric sliding contact transport vehicle. The accurate position of the transport vehicle cannot be directly obtained in the simulation test, and the operation trajectory of the transport vehicle is indirectly simulated by time and speed. To simplify the calculation, the transport vehicle is assumed to be in a uniform speed state throughout the whole process, and the speed parameter is a constant value, denoted as V = 0.6m / s; in order to make the number of the transport vehicle's stops at the loading points proportional to the time used, the time the transport vehicle stops at the task point is set to a constant value, which is independent of the number of cargo boxes required for loading / unloading, and is denoted as t up_d =t down_d = 10s; response time T from task issuance to start of operation of transport vehicle response =1s; single transport vehicle load nload =10 boxes; the number of on-track transport vehicles c = 4, the number of task points on the mountain n = 8; the time point parameter t used to determine the position of the transport vehicle is in seconds as the minimum unit, and t cannot have a decimal part, so the track length L and the task point position should be an integer multiple of the speed, L = 1020m, the task point number and the corresponding position are shown in Table 1, the initial position of the first transport vehicle is at the track scale 0, and the initial position interval distance of adjacent transport vehicles is the sum of the vehicle length and the shortest safety distance, recorded as L gap_d =3.6m; the arrival rate of ordinary goods at the task point is related to the picking efficiency of the orchard. The simulation test is designed from the smallest unit of goods. The total amount of goods arriving at the task point in a single test is fixed. The total amount of goods arriving at the task point in different tests increases with the test number. The number of cargo arrivals in all tests n arrive_d and the cargo arrival interval t gap_d Fixed, n arrive_d = 8 times, t gap_d =300s, the 6 sets of test tasks for random arrival events are shown in Table 2; there are two types of emergency tasks, type 0 is that the loading point number is less than the unloading point, type 1 is that the loading point number is greater than the unloading point, and multiple emergency tasks with different occurrence times or different task types are combined to design a total of 5 sets of emergency task sets, each set of tasks contains 2 separate emergency tasks, and the specific task sets are shown in Table 3. The networks in the DQN algorithm are all BP neural networks, the number of hidden layer neurons is set to 8, the learning rate α=0.001, and the discount factor γ=1; the training data is generated by randomly selected rules, the number of iterations is 200 times, and the parameters of the current value network are copied every 5 times by the target value network; in the reward and punishment function R, A=0.7, B=0.2, and C=0.1 are set according to the priority of the objective function. Each experiment is repeated 10 times when the reinforcement learning algorithm is applied, and the objective function value of each run is recorded.

[0228] Table 1 Mission point numbers and corresponding locations

[0229] Number 1 2 3 4 5 6 7 8 Position / m 63 180 342 405 534 690 810 900

[0230] Table 26 Task set of test task points

[0231]

[0232] Table 3 Emergency task set

[0233]

[0234]

[0235] Result analysis under random arrival events of tasks

[0236] Table 4 shows the comparison results of four dynamic scheduling frameworks under six sets of experiments with different arrival rates of goods at task points, based on the same scheduling algorithm (Rule 4), using the full load rate ω, the total task completion time T task , and the number of times the transport vehicle visits empty task points n 0 as evaluation indicators to analyze the performance of the proposed scheduling framework, with the importance of the indicators decreasing in turn. The full load rate is the most important evaluation indicator in dynamic scheduling. To more intuitively reflect the performance gap of different frameworks in terms of the full load rate indicator, the actual number of transport vehicles dispatched n c of the four frameworks is given as a function of the experiment number. n c is the original data for calculating ω, as shown in Figure 12 .

[0237] Generally speaking, the larger the experiment number, the greater the total task volume, and the corresponding number of transport vehicles to be dispatched is more. The scatter plot of the same framework shows an increasing trend, and the T task in Table 4 is also longer. The optimal frameworks for the six sets of experiments in Table 4 are in bold. The scheduling scheme obtained by Framework 3 has 3 sets of optimality, the scheduling scheme obtained by Framework 1 has 2 sets of optimality, the scheduling scheme obtained by Framework 2 has 1 set of optimality, and the scheme obtained by Framework 4 has the worst effect. In terms of the full load rate indicator analysis, the n c of all experiments of Framework 2 and Framework 3 are exactly the same and the number is the least, and the full load rate is 1, which has been optimized to the maximum value without wasting transport capacity; Framework 1 has 2 sets of experiments the same as Framework 2, and 1 more vehicle in the remaining 4 sets of experiments; n c of Framework 4 is the most in all experiments, and there is no case of full load, and the optimization effect is not significant. From the comparison results, the optimization performance of each framework from good to bad is Framework 3, Framework 2, Framework 1, Framework 4.

[0238] Framework 1 and Framework 2 execute the scheduling scheme given according to the initial position, and there is no opportunity for rescheduling during the transportation process. Therefore, n 0 are both 0, and the two only differ in the rescheduling strategy. Framework 2 can be fully loaded in all experiments; Framework 1 can only be fully loaded in 33% of the experiments. Since Framework 1 can adaptively dispatch the time and number of transport vehicles according to the task generation time and the existing task volume, the T task of the full load scheme is the shortest among all frameworks and is the optimal scheduling scheme. It can be seen that the scheduling framework with the cycle as the rescheduling strategy has better stability than the event-based one. Sending a transport vehicle at intervals based on the cycle scheduling reduces the requirements for the existing task volume at the task point in both space and time, making it easier to be fully loaded. Therefore, when the full load rate is the same, the T task of Framework 2 is longer.

[0239] Both Frame 2 and Frame 3 are based on cyclic scheduling, rescheduling the global loading points and loading quantities at the initial positions, and the full-load rates of the 6 groups of tests are all 1; since Frame 3 allows the transport vehicle to transport all the existing task quantities at the target loading point, if the existing task quantity is more than the pre-allocated quantity, it can be fully loaded by stopping at fewer task points, making T task shorter. Therefore, in 4 groups of tests, the T of Frame 3 task is shorter than that of Frame 2, and in 2 groups of tests, they are the same; the n of all tests 0 of Frame 3 is more. The time taken for the transport vehicle to stop at an empty task point is one-tenth of that to stop at a task point with a loading request. If the system can promptly feedback the situation where the task point is 0 to the transport vehicle, time can be saved by avoiding it in advance. It can be seen that rescheduling the task quantity at the target loading point may obtain a scheduling scheme with a shorter T task .

[0240] The only difference between Frame 3 and Frame 4 is the rescheduling parameter, but the scheduling scheme obtained by Frame 4 is far inferior to that of Frame 3 and is the worst in terms of performance among all frames. Frame 4 only determines the next target loading point at each rescheduling position, increasing the probability of multiple transport vehicles being assigned to the same task point. After the goods arrive at the end of the task, the transport vehicle has many empty runs, resulting in a low full-load rate. It can be seen that determining the global loading point number at the initial position results in a higher full-load rate of the scheduling scheme than determining them one by one.

[0241] Based on the above analysis, it can be known that Frame 3 has the best performance among the 4 proposed scheduling frames for dealing with the dynamic event of random task arrival.

[0242] Table 4 Comparison results of different frames

[0243]

[0244] Table 5 is the comparison result of Frame 3 using different scheduling algorithms, and the evaluation indexes are still the full-load rate ω, the total task completion time T task , and the number of times n 0 that the transport vehicle visits empty task points.

[0245] Table 5 Comparison results of different algorithms

[0246]

[0247]

[0248] Note: x max represents the maximum value of the objective function x in the 10 running results of the reinforcement learning algorithm, x min represents the minimum value of the objective function x in the 10 running results, and

[0249] The full load rate of the schemes obtained by using any scheduling algorithm in the six groups of experiments is 1, except for T task 、n 0 There is a difference between the two indicators. If the vehicle stops at one more loading point, it will take 10 seconds longer. Therefore, the T of different scheduling algorithms is task The difference is an integer multiple of 10. In terms of the comparative results of the six rule algorithms, the optimal rule algorithms corresponding to the six groups of experiments in Table 5 are bolded. The optimal solution of experiment 1 is generated by rules 2 and 6, the optimal solutions of experiments 2, 4, and 6 are generated by rule 3, the optimal solution of experiment 3 is generated by rule 2, and the optimal solution of experiment 5 is generated by rule 5. Rule 3 can obtain the optimal scheduling solution in 50% of the experiments, accounting for the largest proportion. It can be seen that the multi-step heuristic rule algorithm with task volume as priority is still the best in dynamic scheduling. However, the combination of a single rule is fixed and the optimal solution cannot be obtained in all experiments. The reinforcement learning algorithm regards all single rules as a set of actions, and the agent is trained to decide the current best rule to use based on the state during rescheduling. As for the comparison results of the optimal solution obtained by the rule algorithm and the solution obtained by the reinforcement learning algorithm, in 6 groups of experiments, the optimal solution obtained by the reinforcement learning algorithm after 10 runs was higher in 5 groups of experiments, and the same in 1 group. It can be seen that the reinforcement learning algorithm can obtain the optimal scheduling solution in different experiments using hybrid rules; the average value of the objective function of the 10 running solutions is not much different from the optimal value. The average values ​​of 3 groups of experiments are lower than the optimal value of the rule algorithm, and 3 groups are slightly higher. It can be seen that the reinforcement learning algorithm has better stability.

[0250] Result analysis under emergency mission events

[0251] In order to ensure the consistency of the conditions for the two response strategies, Experiment 3 was selected to generate a scheduling plan for ordinary cargo based on Framework 3 and Rule 2. Table 6 is the comparison result of using different emergency task response strategies based on the same ordinary cargo scheduling plan. d , Impact on the original plan d as evaluation indicators. is the response speed of the uth emergency task in the emergency task experiment; G d It is a matrix. The first value is the difference between the full load rate when there is no urgent task and when there is an urgent task. The second value is the difference between the number of rescheduling when there is an urgent task and when there is no urgent task. The larger the difference, the greater the negative impact.

[0252] From the final results of the comparison of the two strategies, it can be seen that the emergency tasks in the first 3 groups of experiments are all of type 0. The transport vehicles assigned to complete the emergency tasks can transport ordinary goods to the unloading point by the way. The full load rate of using any emergency task response strategy has not changed. The last 2 groups of experiments include an emergency task of type 1. The transport vehicle for transporting type 1 needs to carry the emergency goods through the initial position and cannot transport ordinary goods on this trip. Therefore, the full load rate has decreased to the same extent. Strategy 1 responds according to the scheduling rhythm of Framework 3, and the emergency task can only be assigned to the available transport vehicles at the initial position. Therefore, the number of rescheduling does not increase, and the second value of G in all experiments d is 0, but the response speed is slow and it is necessary to wait for an available transport vehicle at the initial position. Strategy 2 adopts the nearest strategy. The transport vehicles that meet the response requirements can re-plan the transport plan even if they have already been assigned tasks. The increase in the number of rescheduling speeds up the response speed of emergency tasks, and the fastest can respond immediately. It can be seen that the two strategies perform the same in terms of the full load rate index, but Strategy 2 is more flexible in responding to emergency tasks than Strategy 1, and emergency tasks can be processed faster.

[0253] Regarding the comparative analysis between emergency task experiments, the response rhythm of Strategy 1 is fixed. The response speed of any emergency task only depends on the time point when the task appears. The closer the time is to the appearance of an available transport vehicle at the initial position, the shorter it is. The of Strategy 2 is also related to the loading and unloading positions between emergency tasks. When the first emergency task appears in all experiments, the system has just dispatched the first transport vehicle for transporting ordinary goods and has not yet reached the first target loading point, which meets the requirements for executing emergency tasks, both are 0. The two emergency tasks included in Experiments 1, 2, and 3 are of type 0. The second emergency task in Experiments 1 and 2 can be relayed by the transport vehicle that completed the first emergency task in terms of load and the current position of the transport vehicle. However, the loading point of the second emergency task in Experiment 1 is before the unloading point of the first emergency task, and the task point position constraint is not met, so it cannot respond immediately and needs to wait for an available transport vehicle at the initial position, is not 0; the task point position constraint in Experiment 2 is met, and the transport vehicle that responds to the first emergency task can continue to respond to the second emergency task, is 0; when the second emergency task appears in Experiment 3, the load states of all transport vehicles do not meet the execution requirements and need to wait, is not 0. The first emergency task in Experiments 4 and 5 is of type 1. When the second emergency task occurs in Experiment 4, the transport vehicle that completed the first emergency task has not yet passed through the initial position, and the current position constraint of the transport vehicle is not met and it needs to wait, is not zero; In Experiment 5, the time interval between the occurrences of the two emergency tasks is relatively large. The transport vehicle that has completed the first emergency task has passed the initial position, satisfying the current position constraint of the transport vehicle. At the same time, it also satisfies the load and task point position constraints, and can continue to respond to the second emergency task. is zero.

[0254] Based on the comprehensive analysis of all indicators, it can be seen that Strategy 2 better meets the transportation requirements of emergency tasks, with more diverse response methods and faster speed without affecting the full load rate.

[0255] Table 6 Comparison results of different response strategies

[0256]

[0257] For the problem of solving the dynamic scheduling model with randomly arriving tasks as dynamic events, four frameworks with different rescheduling methods are proposed. Based on six groups of simulation experiments with different task arrival rates and the same scheduling algorithm, it is concluded that Framework 3, which pre-allocates the global loading point and loading quantity based on the cycle at the initial position and actually allocates the loading quantity at the loading point, has the best performance. In all experiments, the transport vehicles can be fully loaded, and in 50% of the experiments, the obtained scheduling scheme is optimal; based on Framework 3, the solving performances of six heuristic rule algorithms and reinforcement learning algorithms are compared. The solving effect of the rule algorithm is related to the simulation experiment, and the optimal solutions of different experiments are not obtained by the same scheduling rule. The reinforcement learning algorithm can obtain the optimal solutions of all experiments, is better than the optimal solutions obtained by the rule algorithm in 83% of the experiments, and has a stable solving effect, making the transport vehicle stop at fewer loading points and saving transportation time. For the dynamic scheduling problem with ordinary tasks and emergency tasks occurring simultaneously, two emergency task response strategies are designed based on Framework 3. Under the same ordinary cargo scheduling scheme, a comparison is made based on five groups of emergency task experiments. The two strategies perform the same in terms of the full load rate index. The near strategy has more diverse response methods and is faster than responding according to the framework rhythm, and better meets the response requirements of emergency tasks.

[0258] In summary, the dynamic task scheduling method for the ring track transportation system provided by this application has the following advantages:

[0259] (1) Most existing studies adopt a single static scheduling method, lacking flexibility. This application proposes a variety of dynamic scheduling frameworks, such as scheduling strategies based on event rescheduling and cycle rescheduling; and scheduling parameters including global loading quantity, current loading quantity, next loading point and loading quantity, loading quantity of the current loading point and the next loading point. Appropriate scheduling strategies can be selected according to different scenarios and requirements. Finally, it is concluded that Framework 3 combines the advantages of global loading point planning and actual loading quantity adjustment, which not only ensures the overall rationality of the transportation strategy but also can dynamically adjust the loading quantity according to the actual situation, improving the transportation efficiency.

[0260] (2) The types of scheduling algorithms used in existing research are limited. For example, simulated annealing algorithms and ant colony algorithms. This application uses a variety of heuristic rules and intelligent algorithms, such as heuristic rules based on the amount of tasks and reinforcement learning algorithms. Among them, the reinforcement learning algorithm is more suitable for dealing with the situation where tasks are randomly generated in dynamic scheduling, and can more effectively solve the scheduling problem and improve transportation efficiency. Specifically, this application proposes six heuristic rules based on different priorities to guide the carpooling combination algorithm, which can select appropriate rules according to different scenarios to achieve goals such as full-load transportation and reducing the number of empty runs. Subsequently, the heuristic rules are used as the action space, and the DQN algorithm is used to learn the optimal carpooling combination strategy, which can dynamically adjust the strategy according to environmental changes and achieve better scheduling results.

[0261] (3) This application proposes a frame rhythm strategy and an emergency task nearest response strategy to meet the transportation requirements of emergency tasks. Through simulation experiments, the performance of the two strategies is compared, and the conclusion is drawn that the nearest response strategy is superior to the frame rhythm strategy in terms of response speed.

[0262] (4) This application designs a state space that can effectively describe the relative urgency of each task point, providing an important basis for the DQN algorithm to select the optimal carpooling combination strategy. In addition, a reward and punishment function consistent with the dynamic scheduling objective function is designed, which can effectively guide the intelligent agent to select a carpooling combination strategy that can improve the full-load rate, reduce the number of loading times, and reduce the number of empty runs.

[0263] (5) Aiming at the problem that current research mainly focuses on static task scheduling and cannot handle sudden emergency tasks and newly arrived tasks that randomly occur in the actual transportation process, this application focuses on dynamic task scheduling, considers two dynamic events of randomly arriving tasks and emergency tasks, and proposes different scheduling frameworks and algorithms, which can more effectively meet the actual transportation requirements.

[0264] (6) Existing research mostly takes minimizing the number of loading times of transport vehicles or the shortest task completion time as a single goal. In the dynamic scheduling problem of randomly arriving tasks, this application comprehensively considers three optimization goals: the maximum full-load rate, the shortest total task completion time, and the minimum number of times transport vehicles visit empty task points to establish a scheduling model, which can more comprehensively evaluate the advantages and disadvantages of scheduling schemes. In the dynamic scheduling problem of randomly arriving tasks with the occurrence of emergency tasks, two optimization goals of the fastest average response speed and the least negative impact on the original task transportation are comprehensively considered to design a scheduling strategy.

[0265] In summary, by comprehensively considering various factors such as task arrival rate, emergency tasks, and transporter status, the present application constructs an efficient dynamic scheduling method for the circular track transportation system, which can effectively improve transportation efficiency and flexibility, and is superior to existing scheduling methods. It has made an important breakthrough in the scheduling of the circular track transportation system, solved the limitations of existing research, and proposed a more flexible and efficient scheduling framework and algorithm, which can better meet the actual transportation needs and has important theoretical significance and application value.

[0266] In some embodiments, the present application also provides a computer program product, including a computer program, which when executed by a processor implements the dynamic task scheduling method for the circular track transportation system described above.

[0267] In some embodiments, the present application also provides a computer-readable storage medium, on which a computer program is stored, which when executed by a processor implements the dynamic task scheduling method for the circular track transportation system described above.

[0268] In some embodiments, the present application also provides a computer device, including a processor, a memory, an input / output interface (Input / Output, abbreviated as I / O), a communication interface, and a computer program stored on the memory and executable on the processor, and the processor executes the computer program to implement the dynamic task scheduling method for the circular track transportation system described above.

[0269] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope recorded in this specification.

[0270] Specific examples are used in this article to elaborate on the principles and implementation manners of the present application. The description of the above embodiments is only used to help understand the method and its core idea of the present application; at the same time, for those of ordinary skill in the art, according to the idea of the present application, there will be changes in the specific implementation manners and application scopes. In summary, the content of this specification should not be construed as a limitation to the present application.

Claims

1. A dynamic task scheduling method for a circular rail transportation system, characterized in that: include: Acquire basic data of the circular rail transport system during operation; the circular rail transport system is a transport system with multiple loading points and a single unloading point as the task form; the basic data includes the status data of the transport vehicle and the status data of the loading point; According to the basic data, a task scheduling framework is constructed by using a periodic rescheduling method; The task scheduling framework specifically includes: At the initial moment, all transport vehicles are stored in queue q, and all transport vehicles are available; Calculate the maximum interval time t of the transport vehicle operation gap ; Determine whether there is an available transport vehicle at the initial location; If there is an available transport vehicle at the initial position at the current time t, then according to the maximum interval time t gap Determine the departure time t of each transport vehicle start , move all transport vehicle numbers out of queue q and store them in queue q' in order; If the number of available transport vehicles at the initial position at the current time t is zero, jump directly to step "Determine whether the current time t is equal to the departure time t of each transport vehicle start ”; Determine whether the current time t is equal to the departure time t of each transport vehicle start ; If t = t start , the carpooling combination algorithm is used to obtain the global loading point and loading quantity of the transport vehicle, multiple transport vehicles are pre-allocated at the initial position, and the time t at which the transport vehicle arrives at each loading point in the pre-allocation is calculated arrival , update the status of the transport vehicle and move the transport vehicle number out of the temporary storage queue q' or queue q; If the current time t is not equal to any t start , jump directly to step "determine whether the current time t is equal to the time t of each loading point in the pre-allocation arrival ”; Determine whether the current time t is equal to the time t of each loading point in the pre-allocation arrival ; If t = t arrival , reschedule the loading quantity of the loading point and remove the assigned task quantity from the task set; If the current time t is not equal to any t arrival , jump to step "Judge whether the current time t is equal to the time t when the transport vehicle returns to the initial position renewal ”; Determine whether the transport vehicle is fully loaded; If the transport vehicle is fully loaded, calculate the time t when the transport vehicle returns to the initial position renewal , record the dispatched transport vehicle numbers and corresponding t in the order of the transport vehicle's re-dispatch at the initial position renewal Then jump to step "Judge whether the current time t is equal to the time t when the transport vehicle returns to the initial position renewal ”; If the transport vehicle is not fully loaded, the loading point is the last pre-assigned loading point but not the last loading point of the circular track system. Add the subsequent loading point to the task list of this trip and jump to step "Determine whether the current time t is equal to the time t when the transport vehicle returns to the initial position renewal ”; If the transport vehicle is not fully loaded, the loading point is the last loading point pre-assigned and the last loading point of the circular track system, the rescheduling is completed, and the time point t at which the vehicle arrives at the initial position is calculated. renewal Jump to step "Determine whether the current time t is equal to the time t when the transport vehicle returns to the initial position renewal ”; If the transport vehicle is not fully loaded and the loading point is not the last pre-assigned loading point, jump to step "Determine whether the current time t is equal to the time t when the transport vehicle returns to the initial position renewal ”; Determine whether the current time t is equal to the time t when the transport vehicle returns to the initial position renewal ; If t = t renewal , then t renewal The transport vehicle number that should arrive at the initial position at the moment is added to the queue q, and it is determined whether the task is completed; If the current time t is not equal to any t renewal , directly determine whether all tasks are completed; if not, the scheduling continues, set t+1, and jump to the step "determine whether there is an available transport vehicle at the initial location"; If all tasks are completed, the scheduling ends and all scheduling plans are output; Determine a dynamic task scheduling strategy for the circular rail transportation system based on a task scheduling framework and a dynamic scheduling algorithm; the dynamic scheduling algorithm includes a reinforcement learning algorithm and a proximity strategy; The transport vehicles in the circular rail transport system are operated according to the dynamic task scheduling strategy.

2. The dynamic task scheduling method for the circular track transportation system according to claim 1, characterized in that: The obtaining of basic data during the operation of the circular track transportation system specifically includes: Numbering of multiple transport vehicles and multiple loading points of mountain circular rail transport system; An RFID reader, a GNSS system and a load sensor are set on each transport vehicle, and the status data of each transport vehicle is obtained through the RFID reader, the GNSS system and the load sensor; the status data of the transport vehicle includes the cargo volume of the transport vehicle, the cargo status of the transport vehicle and the real-time position of the transport vehicle; A load sensor and an RFID tag are set at each loading point, and the status data of each loading point is obtained through the load sensor and the RFID tag; the status data of the loading point includes the precise position of the loading point on the track, the loading and unloading quantity of the loading point, and the current quantity status of the loading point; The status data of the transport vehicle and the status data of the loading point during the operation of the mountain circular track transport system are obtained as basic data.

3. The dynamic task scheduling method for the circular track transportation system according to claim 1, characterized in that: The carpooling combination algorithm calculates the global loading points and loading quantities of the transport vehicles based on heuristic rules.

4. The dynamic task scheduling method for the circular track transportation system according to claim 3 is characterized in that: The dynamic task scheduling strategy of the circular rail transportation system is determined based on the task scheduling framework and the dynamic scheduling algorithm, specifically including: When the dynamic event is a random arrival event of a task, the dynamic task scheduling strategy of the circular rail transportation system is determined based on the task scheduling framework and reinforcement learning algorithm; When the dynamic event is an urgent task arrival event, the dynamic task scheduling strategy of the circular rail transportation system is determined based on the task scheduling framework and the proximity strategy.

5. The method for dynamic task scheduling of a circular rail transportation system according to claim 4, characterized in that: The reinforcement learning algorithm is the DQN algorithm.

6. The method for dynamic task scheduling of a circular rail transportation system according to claim 4, characterized in that: The proximity strategy specifically includes: If there is a re-dispatched transport vehicle at the initial location, the current emergency task and the ordinary cargo after the current emergency task is unloaded will be assigned to the transport vehicle; If the transport vehicle has been assigned ordinary cargo, but is currently empty and is before the loading point of the emergency task, the assigned ordinary cargo will be cancelled and the transport vehicle will give priority to transporting the current emergency task and ordinary cargo after the current emergency task is unloaded; The transport vehicle has been assigned an emergency task. If the assigned emergency task unloading point meets the task point location requirements of being before the current emergency task loading point and the assigned emergency task loading point is smaller than the unloading point and the transport vehicle is before the current emergency task loading point, and the transport vehicle is not carrying ordinary goods, if the assigned emergency task loading point is larger than the current emergency task unloading point, the transport vehicle has passed the initial position and is before the current emergency task loading point, and the transport vehicle is not carrying ordinary goods, then after completing the assigned emergency task, the transport vehicle continues to complete the current emergency task and ordinary goods at the unloading point and subsequent task points; If there are multiple transport vehicles that meet the execution requirements, calculate the positions of each transport vehicle and select the transport vehicle closest to the emergency task loading point to execute the current emergency task.

7. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method for dynamic task scheduling of a circular track transportation system according to any one of claims 1 to 6 is implemented.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for dynamic task scheduling of a circular track transportation system according to any one of claims 1 to 6 is implemented.

9. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method for dynamic task scheduling of a circular track transportation system according to any one of claims 1 to 6.

Images