The Research On The Coordinated Scheduling Problems Of Batch Machine Production And Transportation Based On Reinforcement Learning Algorithm

Under the intelligent industrial system,the coordination of production and transportation is one of the most important links in the production system.The production process is characterized by complicated processes,high energy consumption and strict requirements.This makes it important to arrange transportation between processes reasonably.In order to shorten the whole process time and reduce energy consumption,this paper makes an in depth study on the coordination od production and transportation.In this paper,we consider the characteristics of high-temperature operation and transportation in steel enterprises of the production process.Two coordinated scheduling problems of high-temperature transportation and batch production are extracted.We consider that the coordinated scheduling problem of production and transportation is proved to be NP-hard,and there is a disaster of dimensionality in the solution process.Therefore,we use reinforcement learning algorithm to solve the two problems respectively.The main contents of this paper are as follows:(1)In the single batch machine environment,we considers vehicles and machines have capacity limitation when they are transported before production.Jobs arrives at the transportation area dynamically and waiting for transportation.Jobs are transported to the production area for batch processing.Batch machine can process multiple jobs as a batch at the same time.The processing time of each batch jobs is the maximum value of the processing time in the batch of jobs.Taking the minimum total completion time as the goal,a two-stage coordinated scheduling model is established.According to the busy and idle state of vehicles and machines in the problem and the real-time information of jobs,the state characteristics are defined.Schedule rules as actions.Define rewards according to problem objectives to achieve problem transformation.For the problem of continuous state and large space,the Q-learning algorithm is improved by using function approximation method,and the transformed problem is solved.Experimental results show that the algorithm has good stability and feasibility.(2)In a multi-batch machines environment,we consider that transport with multiple vehicles,set vehicle capacity as 1,with constraints such as machine capacity limit.And we consider machine setup time.Jobs arrival dynamically and waiting for transportation.After the completion of transportation,a batch jobs be allocated for processing.Aiming at the total completion time,a mixed integer programming model is established.Then define state features from the real-time changes of the state of the vehicle,machine and job,select the scheduling rule and regard it as action,the objective of problem as reward function,complete the problem conversion.In the case of continuous state features,a reinforcement learning algorithm with linear value function approximation is proposed to solve the problem.The experimental results show that the algorithm has good stability and effectiveness for solving the production and transportation coordination scheduling problem.