Dynamic Adjustment of Measurement Noise Covariance Matrix in an Infrared-based Positioning and Tracking System

The accuracy of optical positioning systems can be compromised by multiple factors (reflections, calibration, etc.). As an alternative to the triangulation, multilateration, or fingerprinting techniques typically used in these systems, stochastic estimation techniques can be used, such as Kalman Filters (KF) in its different variants. They estimate the receiver position based on the acquired measurements and the estimated positions in previous iterations. This work presents the evaluation of a 3D optical positioning system, based on four LED beacons and a quadrant receiver, using an Extended Kalman Filter (EKF). The implementation of a measurement noise covariance matrix that is adjusted depending on the angle and distance between transmitters and receiver, obtained in the previous iteration, is analysed. The receiver position estimation using both a dynamic and a static measurement noise covariance matrix is evaluated and compared with simulations and experimental tests. In simulations, the achieved errors are below 6 cm and 12 cm in 75% of the cases when using a dynamic and a static noise matrix, respectively. In the experimental tests, the obtained errors in 75% of the cases for the position in plane XY and the rotation angle γ are smaller than 7.44cm and 1.06° with a dynamic noise matrix; and below 7.59 cm and 1.62° for a static one.