Experimental comparison of Bounded-Error State Estimation and Constraints Propagation

The vehicle´s localization is classically achieved by Bayesian methods like Extended Kalman Filtering. Such methods provide an estimated position with its associated uncertainty. Bounded-error approaches (Bounded-Error State Estimation and Constraints Propagation) use interval analysis and work in a different way as they provide a possible set of positions. An advantage of bounded-error approaches over Bayesian methods is that their results are guaranteed (whereas the results of Bayesian methods are probabilistically defined). This paper compares both Bounded-Error State Estimation and Constraints Propagation using the same experimental data. The results obtained aim to rank these approaches in terms of computing time, consistency and imprecision.