Cooperative minimum expected length planning for robot formations in stochastic maps

This paper addresses a tightly integrated multi-robot planning, localization and navigation system in stochastic scenarios. We present a novel motion planning technique for robot formations in such kinds of environments, which computes the most likely global path in terms of a defined minimum expected length (EL). EL evaluates the expected cost of a path considering the probability of finding a non traversable zone and the cost of using an alternative traversable path. A local real time re-planning technique based on the probabilistic model is also developed for the formation when the scenario changes. The formation adapts its configuration to the shape of the free room. The partial views of all the robots are integrated to update the multi-robot localization using a modified EKF based on the measurement differencing technique which improves estimation consistency. As a result, a lower uncertainty map of the local navigation area is obtained for re-planning purposes. Experimental results, both in simulation and in real office-like settings, illustrate the performance of the described approach where a hybrid, centralized-distributed, architecture with wireless communication capabilities is used.