Constrained Kalman Filter for Mobile Robot Localization with Gyroscope

Abstract

The odometry information used in localization can be quite erroneous when the robot follows the curved path or suffers from slippage. Thus the use of the low-cost gyroscope to compensate for an angular error is considered by many researchers. Conventional Kalman filtering methods that fuse the odometry with the gyroscope may produce infeasible solution because the robot parameters are estimated regardless of their physical constraints. In this paper, we propose a constrained Kalman filtering method that applies general constrained optimization technique to the estimation of the robot parameters. The state observability is improved by the additional state variables and the accuracy is also improved by the nonapproximated Kalman filter design. Experimental results show the proposed method effectively compensates for the odometry error and yields feasible parameter estimation at the same time