Digital twin-enabled real-time parameter estimation of a wheeled mobile robot

This paper presents a real-time parameter estimation framework for wheeled mobile robots (WMR) using a digital twin architecture based on the Wide-Array of Nonlinear Dynamics Approximation (WyNDA). A WyNDA-based adaptive observer is applied to perform static and time-varying parameter estimation through simulations and experiments, with comparative evaluation against SINDy. Simulation results show that WyNDA consistently estimates both static and time-varying parameters with a MAPE below 6.1%, while experimental validation yields a MAPE of approximately 10%, outperforming SINDy in both cases. Furthermore, the experimental results are implemented in a digital twin, which achieves accurate position tracking during data loss, with errors of 3% and 6% in the x and y estimates, respectively. These results demonstrate WyNDA's effectiveness for real-time parameter estimation and its suitability for digital twin applications under incomplete sensing conditions.