This paper presents a new approach for estimating the state of a timevariant linear dynamic system when the system input and the measurements suffer from two different types of uncertainties simultaneously. The first type of uncertainty is a stochastic process with given characteristics. The second type of uncertainty is only known to be bounded, the underlying distribution is unknown. The new estimator generalizes Kalman filtering and set theoretic filtering and contains these filters as special cases. Two applications in robotics are discussed.
Print ISSN: 0178-2312
Volume: 48, 06/2000
Pages: 265