Research On Multi-task Multi-device Matching Algorithm Based On Machine Learning

There are various matching problems between multi-task and multi-device in various practical scenes.At present,in most cases,it's still manual matched and assigned according to personal experience.Thus,it is necessary to design an intelligent matching algorithm to improve efficiency and reduce labor costs.As one of the most popular subj ects in recent years,machine learning has achieved good results in solving problems in many fields because of its powerful induction and learning ability.With the container dispatching system of a large domestic port as background of this thesis,the scheduling problem is defined as multi-task multi-device matching problem.Different machine learning algorithms are designed according to the matching problems in different stages of the scheduling process.The past manual matched data is fully exploited to train algorithm models,which is verified by simulation.Firstly,this paper introduces the layout and process of the container dispatching system.The whole scheduling process is divided into two parts:task batch and truck fleet matching and specific task and vehicle matching.The algorithm model of the overall process is designed.Secondly,a decision tree-based matching algorithm is presented based on the matching problem between task batch and truck fleet in the current manual scheduling process.The decision tree model is trained to be close to manual decision rules by using the past data of dispatching vehicle.Then,a neural network-based optimization algorithm is designed to solve the matching problem between specific container transportation tasks and vehicles.Based on the definition of the cost coefficient between the task and the equipment,the optimization solution is obtained using the Hopfield neural network.Finally,the two algorithms are integrated to design a complete intelligent scheduling matching algorithm.The loading and unloading machinery,the vehicle and the task pool in the collection and dispatching process are simulated equivalently.The whole process of the collection and dispatching is simulated.The effectiveness of the algorithm is verified.