Autonomous Single Camera Exploration

In this paper we present an active exploration strategy for a mobile robot navigating in 3D. The aim is to control a moving robot that autonomously builds a visual feature map while at the same time optimises its localisation in this map. The technique chooses the most appropriate commands maximising the information gain between prior states and measurements, while performing 6DOF bearingonly SLAM at video rate. Maximising the mutual information helps the vehicle avoid ill-conditioned measurements appropriate to bearing-only SLAM. To validate the approach, extensive simulations over rugged terrain have been performed. Moreover, experimental results are shown for the technique being tested with a synchro-drive mobile robot platform.