Study On Operational Policies Optimzization In Vehicle-based Storage And Retrieval Systems

With the rapid development of information technology and e-commerce,logistic industry has become the third profit source of enterprises and obtained unprecedented development force.Warehouse is the material and information transition center of logistic system,which has critical effects on the operational efficiency and cost of enterprises'logistic system.Traditional warehouses have relatively low utilization of floor space,high operational cost and poor throughput capacity flexibility.To overcome these disadvantages,autonomous vehicle has been applied in warehousing system,forming vehicle-based storage and retrieval systems.These systems store loads in high-density storage units,use autonomous vehicles for transportation and order picking.With high throughput capacity and flexibility,they have been used in many online retailers.Currently,the studies on vehicle-based storage and retrieval systems mainly focus on system design.Operational policies,which may significantly affect the operational efficiency and cost of the system,have not drawn enough attention.Therefore,building accurate analytical models and studying operational policies will play an important role in the application of vehicle-based storage and retrieval systems.Based on this consideration,this thesis build queueing networks for vehicle-based storage and retrieval systems,and conduct study on several key operational policies.First,this thesis studies the vehicle-based storage and retrieval system that uses fixed storage units.While previous studies assume parallel operational policy of vehicles and the lift.This thesis proposes parallel operational policy for vehicles and the lift,in order to improve the throughput capacity.A fork-join queueing network is formulated to estimate the system performance and phase-type based approximation method is designed to solve the queueing network.Simulations are used to validate the analytical model.The comparison between these two operational policies show that the parallel operational policy outperforms the sequential operational policy when the customer demand is small.Moreover,the parallel policy is investigated in a real case,which shows that the sytem throughput time can be decreased by 2.11%.Then,this dissertation studies two vehicle-based storage and retrieval systems that both use movable storage units,including robotic-compact storage and retrieval system that uses standard tote and robotic mobile fulfillment system that uses movable shelf.Robotic-compact storage and retrieval system has drawn extensive attendion in recent five years,while has not been studied.This thesis build semi-open queueing networks to estimate system performance,considering both dedicated and shared storage policies and immediated and delayed reshuffling policies.Simulations are used to validate the analytical models.Moreover,cost minimization model is formulated to optimize the system size,the number of robots and operational policies.The comparison results show that dedicated storage policy benefits the system throughput time,while increasing the investment.The immediate reshuffling policy benefits the dual-command cycle time,while sacrificing the retrieval throughput time.Third,this thesis studies battery recovery problem in robotic mobile fulfillment systems,considering plug-in charging,inductive charging and battery swapping policies.Semi-open queueing networks are built to estimate the system performance under these three battery recovery strategies and simulation models are used to validate the analytical models.Comparisons are carried out in terms of system throughput time and investment,the results show that inductive charging strategy performs the best in terms of system throughput time,while it is expensive.Plug-in charging is cheapest,while the system throughput time will be sacrificed.The performance of battery swapping strategy depends on the price of spare batteries.At last,this thesis studies the assignment problem of workstations to robots in robotic mobile fulfillment systems.Previous studies on RMFS assume that workstations are randomly assigned to robots with shelves.To improve the system performance,this thesis proposes handling-speeds based assignment rule and designs neighborhood search approach to find a near optimal assignment rule.Queueing networks are formulated to estimate the system performance under these rules.Comparisons are conducted between random,handling-speeds based,near optimal and optimal assignment rules obtained through enumerate algorithm.The results show that the neighborhood seach approach can provide a near-optimal assignment rule that is very close to optimal one,while uses much less time.The advantage of working-speed assignment rule depends on the location of workstations.