Optimal input shape terminal iterative learning control for robust backlash mitigation in EV powertrains

초록

Backlash in electric vehicle (EV) powertrains causes torque transmission delays and impact-induced jerk during re-engagement, degrading drivability and ride comfort. This paper proposes an input-shaping Terminal Iterative Learning Control (TILC) approach that optimizes backlash traversal without requiring explicit estimation of backlash size or position. The control input is designed as a slope-limited bang-bang profile, and the switching time is iteratively learned from terminal measurements, enabling direct optimization of backlash duration while preserving smooth torque response. By leveraging the repetitive nature of backlash events, the proposed method eliminates computationally intensive backlash estimation and achieves either reduced backlash duration under the same jerk level or lower jerk for a comparable transition time. A sampling-time compensation scheme is introduced to mitigate discretizationinduced errors, and a model-based compensation mechanism ensures robustness under iteration-varying driving conditions. Simulation and experimental results demonstrate faster learning convergence and improved robustness compared with conventional backlash compensation strategies.

키워드