Constrained Kialman Filter for Mobile RobotLocalization withGyroscope
Hyoung-Ki Lee, Seokwon Bang, Yong-Beom Lee, Sang-Ryoung Kim · 2006
Theodometry information usedinlocalization can bequite erroneous whentherobotfollows thecurvedpathor suffers fromslippage. Thustheuseofthelow-cost gyroscope to compensate foran angularerrorisconsidered by many researchers. Conventional Kalmanfiltering methods thatfusethe odometry withthegyroscope may produce infeasible solution because therobotparameters areestimated regardless oftheir physical constraints. Inthispaper,we propose a constrained Kalman filtering methodthatapplies generalconstrained optimization technique totheestimation oftherobotparameters. The stateobservability isimprovedby theadditional state variables and theaccuracy isalsoimproved by thenon- approximated Kalmanfilter design. Experimental results show theproposed methodeffectively compensates fortheodometry errorandyields feasible parameter estimation atthesametime. IndexTerms- Localization, KalmanFilter, Gyroscope, Constraints, Observability.