Fuzzy approaches to driven Kalman filtering for small robot localization

Abstract

Recent robotics research has focused heavily on small robots - those that can be hand-carried or palmtop models. However, most recent research in robot localization has focused on highly computationally expensive algorithms, like various particle filter based approaches, that are inappropriate for these small platforms. This paper discusses a set of fuzzy controlled Kalman filters - fuzzy extended Kalman filter (FEKF) and a double fuzzy sigma-point Kalman filter (DFSPKF) - and compares them to traditional EKF and SPKF filters in a simulated environment. All of these filters are especially appropriate for small robots with uncertain sensors and limited computation capacity. Given a fast fuzzy logic controller, the Fuzzy EKF performs almost as well as the SPKF in the simulated environment, and the DFSPKF shows promise to create a robust and low-complexity localization scheme.