Particle swarm optimization aided calibration of sensor installation errors for MEMS accelerometers

Abstract

Accelerometers, gyroscopes, and magnetometers, constituting the Inertial Measurement Unit (IMU), are among the most fundamental sensors utilized in localization problems. However, their utilization is a challenging task since the reliability of measurements depends on the precision of the executed calibration method. This paper proposes an industrial robot-based infield calibration process in a numerical optimization aided framework, which does not need turntable nor dedicated expensive laboratory setup. The proposed method utilizes a calibrated industrial robotic arm, which is programmed to execute wide variety of motions with comprehensive dynamical ranges. The end effector of the robotic arm is equipped with IMUs with predefined installation errors. The proposed calibration process records the ground truth accelerations and raw IMU data in a database in Robot Operating System (ROS) framework. Then, realistic sensor models are established, and the Particle Swarm optimization (PSO) algorithm is executed with a custom fitness function to obtain the between sensor misalignment and bias errors. The comprehensive analysis highlights that the proposed method effectively determines the calibration parameters. Mean squared error (MSE) performance metrics shows that the precision of these parameters mainly depends on the complexity of the executed dynamical motions.