Filtering Strategies for State Estimation of Omniwheel Robots

Abstract

Various state estimation strategies are investigated using a kinematic model of a four-wheel holonomic robot with Swedish wheels. A multi-tiered filtering strategy is implemented using Kalman Filters (KF) developed to estimate wheel velocity with an Extended Kalman Filter (EKF) and a Smooth Variable Structure Filter (SVSF) developed for state estimation of the robot. The use of only KFs on the wheels, only EKF or SVSF on the robot, and KFs on the wheels with either an EKF or SVSF on the robot is tested. Simulation results show that using a KF on each wheel in conjunction with either a SVSF or EKF on the robot yields an order of magnitude better state estimation compared to other configurations allowing for increased control of the robot.