Dynamic State Space Modeling with Bond Graph for Improved EKF Performance in Differential Drive Robots

This study explores the use of bond graph modeling to derive dynamic state space equations for a differential drive robot, integrating them into an Extended Kalman Filter (EKF). Standard EKF implementations often rely on kinematic models, leading to inaccuracies due to simplified dynamics. Bond graphs provide a systematic method for generating dynamic equations that capture complex interactions. However, a limitation was found in this method that bond graph software is producing linear Single Input Single Output (SISO) systems, whereas our model requires a Multiple Input Multiple Output (MIMO) approach to be applied to the EKF. Despite this limitation, our results indicate that dynamic state space models significantly improve state estimation accuracy compared to kinematic models and highlight a simpler approach to deriving dynamic models. This work highlights both the potential and limitations of using bond graph modeling to enhance EKF performance in robotics applications.