1 Conflict-free AGV Routing in a Bi-directional Path Layout

• Routing Algorithms for Specific Path Topologies Different path topologies have been proposed together with the routing algorithms based on the path layouts. Here specific path topologies refer to that path networks are regular graphs, e.g., loop/loops, mesh, etc. Tanchoco and Sinriech suggested an optimal single closed loop path layout for an AGV system [11]. The proposed routing algorithm is very simple because there is only one single closed loop path that serves all the pickup/drop-off (or, P/D for short) stations along the loop. A non-overlapping multi-loops path layout is proposed together with algorithms for routing control of a tandem AGVS [8,9]. Only a single dedicated vehicle is employed to serve P/D stations in every loop. Task-list and time-window are used in the algorithm to find a shortest travel time route for an AGV based on the current system status. Hsu and Huang gave time and space complexities for some basic AGV routing operations in specific bi-directional path topologies, which include linear array, ring, binary-tree and H-tree, star, 2D-mesh, n-cube and complete graph [3,4]. Algorithms for general path topology emphasize on finding the optimal routes without considering the characteristics of path. Therefore, the amount of computation could be very large with respect to the size of path networks. On the other hand, other existing works reviewed focus on the design of specific path layouts. As a result, the routing algorithms trend be very simple, e.g., routing one AGV running in a single loop. However, the throughput of the overall system could be very low. To overcome some of the drawbacks described above, we present a bi-directional path on which an algorithm is given to route AGVs to complete their P/D jobs without conflicts upon the shortest possible distance within a short time. The bi-directional path is comprised of two parallel lanes and bridges that connect lanes at P/D stations. All of the P/D jobs are classified into two non-overlapping groups based on the positions of their pickup and drop-off stations. Correspondingly, AGVs are divided into two arrays to serve each job group. A routing algorithm is given to route these two arrays of AGVs to run on both lanes simultaneously in opposite directions. The algorithm is a parallel procedure which allows vehicles to run without conflicts. A batch of P/D jobs scheduled by the procedure could be completed within the time requirement of the most time-consuming job among all jobs. When the central controller assigns a batch of P/D jobs to a set of AGVs, the route for every AGV could be determined before it sets out. This is possible because in our scheme, routes for AGVs are mutually independent. Hence, both centralized and decentralized control mechanisms could be adopted. For the centralized control mechanism, what the central controller should do is to assign jobs and send routing instructions to AGVs. The central controller could make all routing decisions within Θ ( N ) time complexity, where N is the number of jobs rather than that of vehicles or nodes of the path network. While for the decentralized control mechanism, the central controller only need assign jobs to AGVs. The local controller of every AGV then could compute the route for the vehicle itself within Θ ( 1 ) time complexity. 2
1. Introduction
Automated Guided Vehicle (AGV) systems nowadays are becoming popular in automatic material handling systems and shipping applications [1,2,5,7,8,11,12]. In order to improve the system efficiency in terms of throughput and response time, good routing algorithms are always needed to meet the requirements of applications. Previous work on routing algorithms can be classified into the following two classes. • Routing Algorithms for General Path Topology These algorithms aim to find universal optimal routes for general path topology. Here general path means that path networks can be any irregular graphs. Broadbent et al. first gave the concept of conflict-free shortest-time AGV routing and proposed a routing procedure based on Dijsktra’s shortest path algorithm [1]. Daniels first introduced an algorithm to route vehicles in a bi-directional path network based on the branch-and-bound method [2]. The algorithm can detect and find a conflict-free shortest-time route for a newly added AGV without changing the existing routes of others. Another conflict-free shortest-time routing algorithm is presented, in which the concept of time window graph is introduced [5,6]. An algorithm, which could give an optimal integrated solution for planning the dispatching, conflict-free routing, and scheduling of AGVs, is presented based on dynamic programming [7]. 1