A novel observer concept for embedded drivetrain applications

This article describes a concept for dynamic state estimation of an automotive drivetrain comprising multiple friction elements (like clutches and brakes) within gear transmissions. Adaptive time-step methods for solving model equations are unfeasible for embedded control, therefore concepts using fixed time-step methods are required. However, standard fixed step approaches may induce unwanted oscillations at locking events of friction elements. The proposed approach overcomes this limitation and permits embedded execution at rates common to automotive control units. Although locking friction elements influence the mechanical degrees of freedom, which leads to a switched system, one single observer gain matrix suffices for the novel observer concept. Asymptotic stability of the estimation error is shown based on Lyapunov theory. Finally, the proposed observer is validated using vehicle measurement data.