Multi-robot localization in a landmark-free environment using binaural sonar

It's interesting for a collaborative group of mobile robots to be able to identify and localize each other without referring to external beacons. In this context, we are able to develop a localizer model by combining an extended Kalman filter with an ultrasonic sensing model and synchronize them by a communication protocol among mobile agents. The approach we present in this paper consists of embedding on each agent traditional encoders in order to estimate their motion. Then, the EKF is used to correct the position estimate by a collective data fusion technique. The information required here consists of the relative pose between a pair of vehicles. It's an exteroceptive sensor, called binaural sonar, that allows to perform the observation. All along the localization process, these mobile agents share collected data and notify each other when new events procure. This paper also presents simulation results for this approach when applied to non-holonomic constraints mobile robots without using external tools or centralized computational units.