Design And Implementation Of A Markov Decision Process Based Reservation Model For Institute Large Instrument Sharing

In recent years,instrument resources are unevenly distributed across the colleges and institutes.Namely shortage of instrument resources and waste of instrument resources coexist in colleges and institutes.Thus,constructing a mechanism of instrument resources sharing is an important research topic of the management of scientific instruments.To deal with this issue,this thesis unions the reality of my college,and attempts to propose a Markov decision process based reservation model for institute large instrument sharing.Specifically,the model determines whether to accept the reservation request in real time among users from the institute or other institutes or other universities,according to the priorities of these three type users.Based on a given service capacity,reservation requests with high-priority users could be accepted more likely.This ensures that the instrument can meet the needs of the institute users in the largest extent,at the same time its utilization can be effectively increased by the way of opening and sharing.Furthermore,in this thesis,a reservation platform for institute large instrument sharing is designed for opening and sharing instrument resources,by promoting the exchange of scientific research between colleges.First of all,from the perspective of real situation,this thesis introduces domestic and foreign research works of service capacity allocation and Markov decision process model.Furthermore,we review the related work with domestic reservation system for large instrument sharing,then we propose a Markov decision process based reservation model for sharing institute large instrument.In particular,we propose a Markov decision process based reservation model by defining the priority-based target gain function from large instrument's recording data and parameter analysis.In addition,considering the case of a cancellation,this thesis extends the model to cancellation in reservation stage with a dynamic cancellation probability.Then we prove that the model has an optimal policy,and utilize dynamic programming to produce the corresponding optimal policy.Furthermore,we design and realize a reservation platform for institute large instrument sharing,which consists of two parts: a reservation platform on WeChat and a monitoring system of large instrument.In particular,one important part of the reservation platform is the decision-making module.When the estimated number of the reservation requests exceeds the limit of the reservation,the decision-making module will be triggered to decide to accept or reject the reservation requests of users among different priorities.At last,a simulation test is given to validate the efficiency of the model.First this thesis sets several test scenarios by choosing different parameter values.Then by testing these scenarios,the impacts imposed by different parameter values are analyzed and summarized with two important conclusions:(1)Under MDP reservation model,inappropriate settings of priorities may lead to the idle time of the instrument.(2)MDP reservation model outperforms other two traditional methods(First Come First Service Strategy and Segment Reservation Strategy),under the condition of different arrival rates.In other words,when the number of the reservation requests is fixed,the optimal strategy provided by MDP reservation model can accept more reservation requests from high-priority users,so as to improve the efficiency.