High Bandwidth Orientation Measurement and Control Based on Complementary Filtering

Abstract When sufficiently small and cheap enough, sensor systems for the fast and accurate, direct measurement of the orientation of a robot relative to its environment offer a large potential for enhancing the safety and efficiency in various robotic applications. Examples include precision assembly, safety supervision and run-away detection, or attitude control of mobile systems. This paper presents an estimation system capable of angle measurements of up to tens of Hertz while consisting of relatively small, cheap and lightweight sensors: an allowably slow inclinometer and two accelerometers. Based on careful analysis and simulations of the estimation system, design criteria are proposed for the tuning of the third-order complementary filters as a function of the characteristic sensor parameters. The performance of the analog and the digital filter design is then experimentally verified for two different types of micromechanical accelerometers: a high performance capacitive accelerometer and a piezoresistive accelerometer. The overall high-bandwidth estimation system exhibits small phase loss over a wide range of frequencies with good robustness against noise and external impacts. The successful application of this estimation system is finally demonstrated in the closed-loop control of an inverted pendulum.