Research On Conflict Prediction-Driven Robotic Mobile Fulfillment System Scheduling

Robotic mobile fulfillment system(RMFS)is a type of automatic pickup and delivery system.In recent years,RMFS has been widely used in logistics.Conflict between autonomous vehicles(AV)is the main reason that affects operation efficiency in the system.This topic focuses on the conflict problem between AV in RMFS.The scheduling in RMFS to reduce conflicts based on conflict prediction is explored.The conflict-free scheduling problem and the dynamic scheduling problem in RMFS are studied respectively.The scheduling method in RMFS based on conflict prediction and the test set of evaluation scheduling method are proposed.Finally,the prototype system is designed and developed.The main research work as follows.1)Contributing factors of conflicts between AVs in RMFS are analyzed,such as operating time,working state and pathfinding.And conflicts are classified as stationary obstacles’conflicts,moving obstacles’conflicts,and dynamic events’conflicts.Conflict prediction method is proposed and scheduling before conflicts may replace scheduling after conflicts.A mathematical model of AV scheduling in RMFS is presented.2)To free conflicts,the conflict prediction-driven conflict-free scheduling method of RMFS is proposed.The modified A*algorithm for path-set planning is proposed to resolve the fixed obstacles’conflicts.A simulated annealing(SA)algorithm is employed for scheduling tasks,AVs and their paths.Conflict prediction method is adopted to completely avoiding moving obstacles’ conflicts and passive AV stop.AV conflicts are predicted using the proposed method through experiments.Conflict-free scheduling are achieved.The scheduling solutions with faster complete times,fewer stop-start numbers of AVs,and shorter paths lengths than the scheduling after conflicts method are obtained.3)To free dynamic events’ conflicts,the conflict prediction-driven predictive-reactive dynamic scheduling method of RMFS is proposed.Characteristics of dynamic events in RMFS are analyzed.The consequences of different dynamic events and the effects of scheduling are determined.A pre-scheduling method based on rolling scheduling is proposed.The rescheduling rules when dynamic events occur are established.The tasks arriving in real time using the proposed method are completed faster through experiments.The scheduling solutions with faster complete times and shorter paths lengths than the scheduling after conflicts method are obtained quickly when dynamic events such as urgent order arrival and robot failure occur.4)To verify the above method,the conflict prediction-driven benchmark of RMFS is proposed.The 12800 instances with 256 different combinations are designed based on 8 parameters that affect the number of conflicts such as layout scale,pillar,cross-aisle,direction,storage area strategy,the number of robots,the number of tasks in one batch,and the number of batches.The performance of above method is confirmed by experiments.The scheduling solutions with shorter completion times are obtained in various parameter combinations.Finally,2560 instances are selected as benchmark.The lack of the benchmark for RMFS scheduling is improved.5)To verify the application of the above method,the prototype RMFS scheduling system with physical system,virtual system,data center,and support services for system operation is established.Instruction acceptance and data acquisition mechanism for physical system,scheme generation and simulation mechanism for virtual system,and data processing and comparison mechanism for data center are established.Support services for system operation is used to ensure the smooth operation of the three mechanisms.Future conflicts can be predicted in real time and avoided in advance by the proposed method through the data interaction between simulation model and virtual system.The scheduling solutions with faster complete times,fewer stop-start numbers of AVs,and shorter paths lengths than the scheduling after conflicts method are obtained in the application.