Adaptive parameter based Particle Swarm Optimisation for accelerometer calibration

TLDR: A particle swarm optimization scheme and few of its variants are used for estimating bias, scale and non-orthogonality parameter for an uncalibrated accelerometer and an improved version of PSO has been shown to provide better calibration results as compared to other variants of particle swarm algorithms.

Abstract: Motion estimation from inertial sensors is an art of computing translation and rotation about each axis of the vehicle to which it is strapped to. These inertial sensors, however, suffer from errors and noise which need to be calibrated and compensated, respectively. With growing use of low cost MEMS based inertial sensors in autonomous vehicles, it is desired to calibrate these systems online and infield without requiring any external equipment. The in-field calibration schemes are based on the simple principle that the norms of the accelerometer triads equals the magnitude of the Earth's gravity. In this paper, a particle swarm optimization scheme and few of its variants are used for estimating bias, scale and non-orthogonality parameter for an uncalibrated accelerometer. The proposed scheme provides a matrix of unknown parameters which can be fed to obtain calibrated sensor readings. And finally, an improved version of PSO has been shown to provide better calibration results as compared to other variants of particle swarm algorithms.