A New Nonlinear Real-time Filter Based on Free Energy

This study introduces a novel non-linear real-time filter based on the concept of Free Energy, which is applied to estimate the state of mobile robots in dynamic environments. By leveraging the principles of Bayesian inference, the proposed Free Energy Filter aims to minimize prediction errors between the observed and estimated trajectories. The approach extends existing non-linear filtering techniques, incorporating elements such as the Fokker-Planck equation. The simulation results demonstrate the effectiveness of this method in improving the control accuracy of unicycle robots equipped with the Bessel controller. The study also discusses comparisons with traditional Kalman filtering, highlighting the advantages of the proposed approach in adapting to noisy and uncertain conditions.