Global localization of automated guided vehicles in wireless networks

Flexible warehouse systems gain in importance as the demand for seamless transportation of products increases. Conventional consignment systems rely on conveyors and human operators who assemble the goods. In a new approach, a swarm of small automated guided vehicles (AGVs) is employed to collect items from storage shelves and take them to a picking station. A challenging issue connected with this approach is to provide efficient localization and tracking techniques in order to obtain the AGV's position at any point in time. In this paper the localization is realized by sensor fusion of wheel encoder data and range measurements obtained from an IEEE 802.15.4a CSS network. The choice of the filter algorithm is often crucial for the performance of the localization. For this reason this paper provides a comparison of three implementations of a Bayes Filter: the Extended Kalman Filter, Unscented Kalman Filter and the Particle Filter. Experimental results show how accurate position estimation can be performed by each of the filter algorithms.