Comparative Analysis and New Proposal of Velocity Estimation Methods for Multivariable Motion Sensors

This paper presents a comparative analysis of three velocity estimation methods, namely the finite difference method, an adaptive differentiator, and a proposed method based on the combination of velocity estimations derived from the integration of acceleration and from the linear least-square fitting of position. The method performances have been compared through simulations and experimental tests, involving the implementation of the methods within an industrial multivariable motion sensor, which is composed of a linear displacement sensor paired with an inertial measurement unit (IMU). The proposed algorithm features the lowest estimation error among the compared methods. In particular, the maximum velocity estimation relative error is 2% for a reference velocity of 50 mm/s.