Twin-Models-Based Controller Output Observer for Vehicle Dynamics Estimation: Concept, Design and Validation

This paper presents a new concept for vehicle dynamics estimation based on the twin-models-based controller output observer . The objective is to estimate lateral speed, considered as an unmeasured state, as well as longitudinal wheel torque and steering torque, which are treated as vehicle unknown inputs (UIs). The central idea is to recast the estimation problem as a reference tracking control scheme composed of two model-based control loops. In the first loop, Takagi-Sugeno (TS) modeling is used to capture time-varying vehicle dynamics, and a state feedback controller is designed for tracking. This loop provides initial estimates of vehicle states and UIs. To address the mismatch between the TS model and the actual vehicle system, we introduce a second control loop featuring a static output feedback (SOF) compensator that offers a correction term based on a nominal vehicle model. We demonstrate that the control inputs and states of the nominal vehicle model in this correction loop can effectively serve as estimates of the vehicle UIs and states. The control design in both loops is formulated as convex optimization problems under linear matrix inequality (LMI) constraints. The proposed vehicle estimation method is validated using experimental data from the INSA autonomous vehicle on a real test track.