A Control Method Of Automated Guided Vehicle System In Intelligent Warehouse Based On Collision-Avoidance And Deadlock-Control

With the development of robot and Internet of Things(IoT),applying Automated Guided Vehicle(AGV)in logistic warehouse management systems is becoming a promising alternative in enterprises,for its high efficiency and flexible deployment.With the increase of warehouse scales and more AGVs adopted,the control and management of the Automated Guided Vehicle System(AGVS)is becoming complex and difficult,especially facing collision and deadlock problems.Pervious works mainly use a layout with unidirectional lanes to solve these problems.However,this kind of layout increases the total travel distance of AGV and reduces the system efficiency,since each AGV needs to travel a longer distance for its own tasks.This paper focuses on designing effective traffic control policies to ensure that there is no collision or deadlock in the system,so that AGVS can achieve its maximum throughout.A control mode based on zone control is proposed to avoid the collision.Besides,two traffic control policies: Detection and Recovery(DR)algorithm and Deadlock Avoidance(DA)algorithm are proposed.The DR algorithm solves the deadlock by rerouting the blocked AGVs,while the DA algorithm predicts the existence of the deadlock and reroutes the potential blocked AGVs.Based on previous studies,this paper first introduces the definitions of multi-level deadlock and general deadlock to describe the deadlock caused by using the DA algorithm.Traditional deadlock detection method searches the whole layout to detect the deadlock.Different from that,this paper proposes a deadlock detection method beginning from the current AGV,and the complexity of this algorithm is linear to the square of the number of AGVs in the system,which is suitable for the intelligent warehouse with a large scale.After identifying the deadlock in the system,we propose a rerouting policy,which can recovery the deadlock even in a complex layout.Moreover,this paper proposes a Future Path Planning(FPP)algorithm based on the feature of AGVS.The main idea of this algorithm is to evaluate the distance value of each path based on the sampling of future traffic condition.Paths travelled by more AGVs are given a larger path distance to make the routes for multiple AGVs as scattered as possible,so that deadlock and collision will reduce significantly in the system.Based on this idea,another traffic control policy named History-Based Path Planning(HBPP)is proposed which is based on the sampling of history traffic condition.We design two different kinds of automated warehouse layouts for simulation,which is suitable for the parcel handling and good-to-human transportation respectively.To evaluate the performances of different algorithms,a performance metric waste ratio is designed to represent the time wasted on waiting and rerouting.Extensive experiments are conducted to justify the efficiency and validity of these algorithms,and analysis are given to the performances of these algorithms in two different layouts.