Sensor fusion in a SMA-based hexapod bio-mimetic robot

This paper implement three different algorithms for sensor fusion in a shape memory alloy (SMA) based hexapod bio-mimetic robot; a Kalman filter minimizes the estimation error variance, an H
∞
filter minimizes the worst-case estimation error, and a robust mixed Kalman/H
∞
filter that takes care of the system uncertainty. The study installed a calibrated inertial measurement unit and an electric compass sensor on the robot for inertial guidance as well as step touch sensors to estimate legged odometry. The performances of the two-dimensional inertial navigation are then compared for evaluation.