A method for quickly generating AGV yard scheduling plans

Abstract

The present invention belongs to the technical field of automated terminals, and specifically is a method for quickly generating an AGV yard scheduling plan. The present invention generates an AGV scheduling plan for the yard based on the time when the AGV arrives at the yard and the expected processing time, so as to shorten the total waiting time of the AGV in the yard. The specific method is based on a depth-first search strategy, and is combined with task selection to generate a plan. Each trial in the depth-first search process is divided into two situations: there are still tasks that have not arrived; all tasks have arrived; when all trials are completed, the final plan is retained. The method of the present invention has a fast calculation speed, and the generated plan can effectively reduce the waiting time of the AGV in the yard. Experiments show that compared with existing genetic algorithms, the method of the present invention has significant advantages in running speed and saving the total waiting time of AGVs in the yard.

Description

Technical Field

[0001] The present invention belongs to the technical field of automated terminals, and in particular relates to a method for generating an AGV dispatching plan for a storage yard. Background Art

[0002] With the development of economic globalization, communication and cooperation between countries around the world are becoming increasingly frequent, and the scale of port logistics has grown significantly. Reasonable planning of the operation of port production equipment will greatly promote energy conservation and efficiency in daily port operations. AGVs (Automated Guided Vehicles) are one of the core equipment in port production. Improving AGV scheduling is crucial to improving overall port operational efficiency.

[0003] Most existing literature on optimization methods for AGV scheduling in automated terminals uses genetic algorithms to solve mixed-integer programming models. Genetic algorithms are characterized by strong search capabilities, but limited solution accuracy and are not easily combined with other search methods and selection rules. Furthermore, parameters such as the crossover rate and mutation rate contained in the algorithm often require repeated experimental debugging to set, and any slight change in the problem requires readjustment. Summary of the Invention

[0004] The purpose of the present invention is to provide a method for generating an AGV yard scheduling plan with high precision and high speed.

[0005] The proposed method for generating an AGV yard scheduling plan is based on the arrival time of the AGV and the estimated processing time, thereby shortening the total waiting time of the AGV in the yard. Specifically, the method uses a depth-first search strategy as the basis, combined with task selection, to generate a plan. Each trial is divided into the following two cases:

[0006] (1) If there are still tasks that have not arrived, then the task with the shortest expected processing time among the tasks that have arrived is selected as the candidate task that has arrived and added to the candidate task set (the candidate task set is initially empty). Based on the expected processing time of the currently arrived candidate task plus the current time, tasks that are expected to arrive before the benchmark time but have not arrived are searched and added to the candidate task set as the candidate tasks that have not arrived. The depth-first exploration is then performed on each task in the candidate task set in turn.

[0007] (2) When all tasks have arrived, the depth-first search of the current branch is stopped, and the subsequent part of the scheduling plan generated by the branch is constructed in ascending order of the estimated processing time of the remaining tasks. If the formed plan is better than the existing optimal plan, the formed plan will be used as the new optimal plan, otherwise the plan will be discarded; when all trials are completed, the final plan is retained.

[0008] The method of the present invention has a fast calculation speed, and the generated plan can effectively reduce the waiting time of AGV in the yard.

[0009] The specific steps for situation (1) are:

[0010] Step (1), select the task with the shortest expected processing time among the tasks that have arrived as the candidate task that has arrived and add it to the candidate task set (the candidate task set is initially an empty set);

[0011] Step (2), taking the estimated processing time of the currently arrived candidate task plus the current time as the base time, searching for unreachable tasks that are expected to arrive before the base time and adding them to the candidate task set as unreachable candidate tasks;

[0012] Step (3) performs depth-first exploration on each task in the candidate task set in turn.

[0013] The specific steps for situation (2) are:

[0014] Step (1), stop the depth-first search of the current branch, and instead construct the subsequent part of the scheduling plan generated by the branch in ascending order of the estimated processing time of the remaining tasks;

[0015] In step (2), if the plan formed is better than the existing optimal plan, the formed plan is used as the new optimal plan; otherwise, the plan is discarded. When all trials are completed, the plan that is retained is the final plan.

[0016] Experiments show that compared with existing genetic algorithms, the method of the present invention has significant advantages in operating speed and saving the total waiting time of AGV in the yard. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 Schematic diagram of the method for quickly generating an AGV yard scheduling plan according to the present invention. DETAILED DESCRIPTION

[0018] The present invention is further described below with reference to examples and drawings.

[0019] 1. Scenario (1) The specific steps are:

[0020] (1) Select the task with the shortest expected processing time among the tasks that have arrived as the candidate task that has arrived and add it to the candidate task set (the candidate task set is initially an empty set);

[0021] (2) Taking the estimated processing time of the candidate tasks that have arrived plus the current time as the benchmark time, find the unreachable tasks that are expected to arrive before the benchmark time and add them to the candidate task set as unreachable candidate tasks;

[0022] (3) Perform depth-first exploration of this step on each task in the candidate task set.

[0023] For example, consider 10 containers arriving sequentially. Their estimated processing times at the yard are [58, 67, 47, 26, 68, 137, 152, 103, 120, 78], expressed in seconds. For simplicity, assume that these containers arrive at a fixed 40-second interval. At time 0, only the first container arrives. This task is the one with the shortest processing time, with a processing time of 58 seconds. Adding this to the current time of 0 gives a baseline time of 58. Of the unreached tasks before 58 seconds, only task 2 can arrive. Tasks 1 and 2 are added to the candidate set and are then tested one by one using a depth-first approach.

[0024] 2. The specific steps for situation (2) are:

[0025] (1) Stop the depth-first search of the current branch and instead construct the subsequent part of the scheduling plan generated by the branch in ascending order of the estimated processing time of the remaining tasks;

[0026] (2) If the formed plan is better than the existing optimal plan, the formed plan will be used as the new optimal plan; otherwise, the plan will be discarded. When all trials are completed, the plan that is retained is the final plan.

[0027] Using the aforementioned 10 containers arriving in sequence as an example, suppose that after gradual exploration, the current exploration branch has formed a partial plan [1, 2, 4, 3, 5, 6]. At this point, the remaining tasks [7, 8, 9, 10] have all arrived. The remaining schedule generated by this branch is constructed in ascending order of their estimated processing times, forming the plan [1, 2, 4, 3, 5, 6, 10, 8, 9, 7]. The AVG total waiting time at the yard for this plan is 1173 seconds. If this plan is shorter than the total waiting time of the optimal plan obtained from the previous exploration, it is used as the new optimal plan; otherwise, it is discarded.

[0028] Currently, genetic algorithms are commonly used in this work. For comparison, the present invention uses simulated data generated based on actual data characteristics to conduct tests. The running speeds of the genetic algorithm and the method of the present invention and the total waiting time of the generated plans of AGVs are compared. From the experimental results, it can be seen that the method of the present invention is significantly better than the genetic algorithm in both indicators. The specific test results are shown in Table 1 below:

[0029] surface

[0030] .

Claims

1. A method for generating an AGV yard scheduling plan, characterized in that: The AGV scheduling plan for the yard is generated based on the arrival time of the AGV and the estimated processing time, so as to shorten the total waiting time of the AGV in the yard. The specific method is based on the depth-first search strategy, combined with task selection, to generate the plan. Each trial is divided into the following two cases: (1) There are still tasks that have not been completed; At this time, the task with the shortest expected processing time among the tasks that have arrived is selected as the candidate task that has arrived and added to the candidate task set. Based on the expected processing time of the current candidate task plus the current time, the tasks that are expected to arrive before the benchmark time but have not arrived are searched and added to the candidate task set as the candidate tasks that have not arrived. The depth-first exploration is then performed on each task in the candidate task set in turn. (2) All tasks have been completed; Then stop the depth-first search of the current branch and construct the subsequent part of the scheduling plan generated by the branch in ascending order of the estimated processing time of the remaining tasks. If the formed plan is better than the existing optimal plan, the formed plan will be used as the new optimal plan; otherwise, the plan will be discarded; When all the trials are completed, what remains is the final plan.

Images