Welcome

Teleoperating a robot in rough terrain is difficult because cameras shake heavily. The project aims at aggregating spatial information from such cameras into a consistent 3-D representation in real-time. This representation is used to render a virtual overview ("guardian angel") perspective. The challenge of ambiguous environmental features is addressed by a Markov Chain Monte Carlo (MCMC) framework for searching through the space of data-association hypotheses. Each hypothesis is evaluated by an efficient Simultaneous Localization and Mapping (SLAM) algorithm.