Sensor Fusion via Brain Emotional Learning for Ground Vehicle

Abstract

In this work, the analysis of a filter consisting of the Brain Emotional Learning (BEL) algorithm is presented. The inner workings of the BEL filter are based on emotional learning model in mammalians brain. The BEL filter is implemented in simulation for the purpose of sensor fusion in a ground vehicle. In simulation, the signals from a Global Positioning System (GPS) and an Inertial Navigation System (INS) are integrated, in order to accurately track the trajectory of a ground vehicle around a track. The BEL filter is provided with some sensory signal and reward signal, subsequently the filter seeks to diminish noise from both sensing units, thus eliminating tracking error. A performance comparison between the BEL filter, and the more commonly utilized Kalman filter is presented. The BEL filter demonstrated robustness to uncertainties from the sensing units, it adapts quickly with dynamical change in the plant, and has small computational cost. The BEL filter demonstrated to be effective in sensor fusion.