Método automático de calibração de IMU baseado no filtro de Kalman

Abstract

IMUs are widely used in applications that require position and attitude measurements. And with the production of low-cost IMUs, it has been possible to expand this use beyond military and high-cost applications, reaching consumer devices such as cell phones, wearable technologies, physical exercise and rehabilitation equipment. But every IMU needs to be calibrated for use and therefore, in addition to the sensor, there is the cost of calibration. Existing calibration methods are either proprietary to the manufacturers, or they are expensive and require specific equipment and laboratories, or they use non-optimal methods of parameter estimation. This work proposes the systematization of low cost calibration methods, with simple execution performed by the user and that provides optimized results closer to those achieved by high cost calibration methods. The data capture process is based on a multi-position method that does not require extra equipment, besides the sensor and computer itself. The experiments were executed with three commercial IMUs, one high-cost calibrated by the manufacturer, used as a reference, and two low-cost, uncalibrated. The raw accelerometer reading of each IMU was calibrated using the four forms of the Kalman filter developed in this work and the errors obtained were evaluated by comparing the high cost IMU and the calibration with the estimated parameters, showing the good performance of the methods.