A novel data-driven input shaping method using residual impulse vector via unscented Kalman filter

Driven by escalating demands for precision and speed in modern industrial applications, residual vibrations in flexible structures and underactuated systems have emerged as a critical technical challenge, particularly during high-speed emergency braking scenarios. Input shaping has proven to be an effective technique for vibration control. However, existing input shapers commonly encounter challenges with time delay and inaccurate parameters, leading to suboptimal control performance. To address these critical issues, this paper proposes an Unscented Kalman filter-based Residual negative equal-magnitude Shaping (URS) model with two-fold ideas: a) reducing the time delay and compensating the modeling error via the consideration of negative and residual impulse vector; and b) identifying system parameters using a data-driven unscented Kalman filter to enhance control effectiveness. To validate its performance, four experimental datasets from laboratory systems have been established and publicly released. Empirical studies demonstrate that the proposed URS model has achieved a significant vibration suppression effect over several state-of-the-art methods.