A fast unknown-input reconstruction solution framework for desired steady-state steering of distributed drive vehicles.

This paper develops a solution framework for fast reconstruction of unknown wheel-end inputs for four-wheel steering distributed drive vehicles, enabling steady-state steering with the desired small turning radius while maintaining the desired vehicle attitude and driving speed. The core contributions of this paper include a novel system constraint analysis method to determine the valid known-variable combinations, a simplified nonlinear tire model for fast computation without accuracy loss, and differentiated iterative solving procedures adapted to different reconstruction problems. Simulations validate rotational speed reconstruction and steering angle reconstruction with output errors of only 1.78% and 3.07%, respectively, with robustness under varying road adhesion coefficients and driving speeds. Compared with model predictive control and deep reinforcement learning, this framework outperforms mainstream path-tracking methods in both accuracy and efficiency.