Simultaneous 2D Localization and 3D Mapping on a mobile Robot with Time-of-Flight Sensors

Abstract

The problem of building consistent maps of unknown environments is one greatest importance within the mobile robot community. Since the first successful attempts, the variety of solutions has grown larger. One of the most famous approaches, namely the use of a Rao-Blackwellized Particle Filter(RBPF), was introduced by Murphy et al. It relies on sampling from the distribution over robot poses and map parameters. Amongst the large number of succeeding publications, a couple of them teamed the RBPF with some scan matching procedure. Acting on that idea, this thesis describes an algorithm, which is based upon the combination of the RBPF and a form of the Iterative Closest Point(ICP) algorithm. In different way from most established methods, this procedure manages with a much smaller number of samples. It aims to calculate a 3D grid-based map of environments with planar floors, using Time-of-Flight cameras. This kind of sensors allows to extract 3 dimensional information of the environment efficiently, measuring ranges to obstacles. The robustness of the resulting algorithm was proved by virtual experimental mapping of a laboratory, using an “omniRob” platform.