Indirect Adaptive Predictor Preview Control with Unknown Time Varying Input Delay and Parameter
HyunBin Kwon, HAK YI
AI summary
Problem
Unknown time-varying input delays and slowly varying parameters rapidly degrade phase margin and tracking performance in networked or digital control loops, yet existing designs treat delay estimation, preview tracking, and model adaptation separately without unified stability guarantees.
Approach
The method couples an adaptive super-twisting algorithm for real-time delay estimation with an indirect recursive least-squares module for parameter tracking, feeding both into a frozen-parameter predictor and a preview-based feedforward controller.
Key results
- Derives a practical input-to-state stability bound accounting for estimation errors, disturbances, and numerical approximations
- Validates the controller on a DC motor speed servo benchmark under four challenging scenarios
- Reduces steady-state RMSE and peak error to 0.046/0.074 and 0.062/0.099, outperforming three recent baselines
- Eliminates phase lag under time-varying delays while maintaining prediction accuracy during parameter drift
Why it matters
Provides a unified, theoretically grounded framework for maintaining precise tracking in real-world systems like robotics and industrial automation where delays and model parameters unpredictably change.
Abstract
This paper presents an indirect adaptive predictor–preview control architecture for continuous-time sys- tems with unknown time-varying input delays and unknown (slowly varying) parameters. An adaptive super-twisting al- gorithm (STA) estimates the unknown delay online using a monotone ramp probe, and an indirect recursive least-squares (RLS) module tracks slow parameter variations; both feed a frozen-parameter predictor and a preview feedforward based on r(t + ˆh(t)). Nominal exponential tracking is shown under exact prediction, and a practical input-to-state stability (ISS) bound is derived that accounts for delay/parameter estimation errors, disturbances, and numerical approximation. On the DC motor speed servo benchmark, the controller reduces steady- state RMSE/peak error to 0.046/0.074 (S1) and 0.062/0.099 (S4), below all compared baselines.