Algorithm for Velocity Estimation in a Multivariable Motion Sensor

This paper proposes an algorithm for velocity estimation using the position and acceleration signals obtained respectively from a resistive potentiometric displacement sensor and a MEMS accelerometer. The algorithm is composed of two processing chains that independently estimate velocity starting from position and acceleration signals. Velocity estimation from position is obtained through an adaptive windowing differentiator while the estimation from acceleration is based on a leaky integrator low-pass filter. Such two estimations are fused together by means of a tailored weighted average. The proposed algorithm is first simulated in MATLAB and then experimentally implemented and tested. Both simulations and experimental results show that velocity estimation given by the fusion of the outputs of the two processing chains has a lower estimation error compared to the output of each single chain.