Simultaneous localization and map building of skid-steered robots - Ensuring safe movement in structured outdoor environments

In the simultaneous localization and map-building (SLAM) problem, the number of states to be estimated is proportional to the number of objects included in the estimated map plus the states of the robot pose. The known approach of Kalman filter offers an optimal linear recursive solution to the SLAM problem when conditions such as perfect data association, linear motion model, linear measurement expressions and Gaussian error model are in place. However, every update/observation has a computational cost of n3. To reduce this cost, a suboptimal Kalman-based approach based on the covariance intersection filter is proposed. This approach is intended to be used during the medium-term intervals, where global positioning system (GPS) readings are not meaningful or unavailable, to provide high accuracy estimates for both localization and map-building, that assures the safe movement of an autonomous vehicle of skid-steering type, in outdoor semi-structured environments.