Tuning of Extended Kalman filter using Human Opinion Dynamics based optimization

Kalman filter is a well-known technique for optimal state estimation and is widely used for its applicability in different fields. Different derivatives of Kalman filter have been proposed in the past to consider the non-linear aspects of system and measurement model. However, these estimation techniques require precise tuning of process and measurement noise covariance matrices for a given system. This tuning is not only a non-trivial process, but also requires engineering intuition and huge number of Monte Carlo Simulations of the system noise, which at times takes days to freeze. In this paper, a Human Opinion Dynamics (HOD) based optimization of Extended Kalman Filter (EKF) has been proposed for obtaining the tuning parameters. Using these tuning parameters, EKF simulations are carried out for a permanent magnet synchronous motor system model, and thus obtained state estimates are compared with the state estimates obtained from Particle Swarm Optimization (PSO) based tuning method. The simulation results depicts HOD based technique being comparable with the PSO based technique on accuracy grounds and out performs in terms of convergence and ease of implementation.