Hierarchical Traffic Control for Partially Decentralized Coordination of Multi AGV Systems in Industrial Environments Valerio Digani, Lorenzo Sabattini, Cristian Secchi and Cesare Fantuzzi Abstract— This paper deals with decentralized coordination of Automated Guided Vehicles (AGVs) used for logistics oper- ations in industrial environments. We propose a hierarchical traffic control algorithm, that implements path planning on a two layer architecture. The high-level layer describes the topo- logical relationships among different areas of the environment. In the low-level layer, each area includes a set of fixed routes, along which the AGVs have to move. In the proposed control architecture, each AGV autonomously computes its path, on both layers. The coordination among the AGVs is obtained exploiting shared resources (i.e. centralized information) and local negotiation (i.e. decentralized coordination). The proposed strategy is validated by means of simulations. This work is developed within the PAN-Robots European project. I. I NTRODUCTION This paper deals with the path planning and coordination of multiple Automated Guided Vehicles (AGVs) in an auto- mated warehouse. The standard approach to coordinate a fleet of AGVs lies in a centralized supervisor (the control center) which manages all the information coming form the Warehouse Management System (WMS) and from the environment. The control center handles the coordination of the fleet, solving a multi-robot path planning problem. Several works can be found in the literature that face this kind of problem. Generally speaking, multi-robot path planning can be solved exploiting centralized or decentralized strategies. With centralized strategies, a single decision maker de- termines the entire path plan for all the robots. These approaches can theoretically find optimal solutions for multi- robot path planning problems [1], but they are restrictive in the number of robots for which they can plan, as the complexity of planning grows exponentially with the number of robots. Thus, while they provide the highest-quality solu- tions overall, they are generally intractable for large teams. Several centralized strategies can be found in the literature. For instance [1], [2] solve the problem of coordinating a multi-robot system using a coordination space representation of the robot motions. The basic idea is to reduce the size of the problem (exponential with the number of robots involved) exploiting a path decomposition method, which decomposes it into its elementary pieces consisting of either straight line segments or arcs of a circle. Authors are with the Department of Science and Methods for Engineering (DISMI), University of Modena and Reggio Emilia, Italy {valerio.digani,lorenzo.sabattini, cristian.secchi, cesare.fantuzzi}@unimore.it This paper is written within PAN-Robots project. The research leading to these results has received funding from the European Union Seventh Framework Programme (FP7/2007-2013) under grant agreement n. 314193. Another method to reduce the search space is to weakly constrain the allowable paths that robots can follow by limiting the motion of the robots to lie on route maps in the environment. Intuitively, route maps are akin to automotive highways, where robots move from their starting position to a route map, move along the route map to the proximity of the goal, and then move off the route map to the specific goal location. Several strategies can be found in the literature for the coordination of multi-robot systems on a route map, based on different optimality principles [3]–[7]. In order to solve local conflicts, traffic rules may be defined [8]. If robots are allowed to locally exchange information, several strategies can be found in the literature that consider the segments of the route map as resources to be allocated [9], and solve the allocation problem by means of negotiation [10]. The dimension of the multi-robot space may be reduced using of a multi-layer structure to represent the world. As explained in [11], [12], the approach is to construct a hierarchical route map which can abstract the traversable areas using the adequate number of nodes and edges of a graph. The path is searched using the graphs of the several layers. Conversely, completely decentralized approaches are very attractive. In these approaches, each robot autonomously determines its routes, dissolving the conflicts and collect- ing information from other robots. Decentralized techniques are generally faster than centralized ones, but they present several drawbacks, such as failing in finding valid paths for all robots due to deadlocks [13], [14]. In this paper we present a partially decentralized control strategy for the coordination of multi AGV systems. Specifi- cally, our idea is based on a hierarchical control architecture [13]. In detail, two layers are used in order to reduce the total complexity and to simplify the control. The first layer is a topological graph of the plant. The global map of the plant is divided into several macro-areas, called sectors. Each sector corresponds to a node of the graph. Its main purpose is to permit a dynamic re-planning of the paths in case of dynamic events. The second layer is the real route map on which the AGVs move. The coordination on the route map is limited only to a single sector of the first layer. In other words, in each sector, the traffic is managed in a decentralized manner on a local route map. Preliminary results on these topics were introduced in [15].