An unscented Kalman filter method for real time input-parameter-state estimation
This work addresses the joint online estimation of unknown inputs, time-varying parameters, and dynamic states for linear and nonlinear systems with output-only measurements. We propose a unified recursive framework based on the Unscented Kalman Filter (UKF), which augments the state space to jointly incorporate inputs, parameters, and states. By extending the unscented transform to this augmented space, our method avoids Jacobian computation, enabling robust and computationally efficient handling of strong nonlinearities and unknown input disturbances. A real-time data-driven update mechanism, combined with nonlinear function approximation, ensures rapid convergence and high estimation accuracy. In both simulations and physical experiments, the approach achieves millisecond-level response times and superior estimation precision, consistently outperforming conventional Extended Kalman Filters (EKF) and particle filters.