Statistical Learning of Contractive Dynamical Representations for Composite Adaptive Control
A hard-EM method learns contractive disturbance models for predictive composite adaptive tracking control.
The paper connects classical disturbance-accommodating control to last-layer adaptive disturbance rejection by learning dynamical representations whose latent evolution is uniformly contractive. A hard expectation-maximization procedure uses a Kalman smoother to identify the representation from measured plant features and control inputs. Combined with Bayesian filtering, the composite adaptive controller predicts disturbances and has provable exponential convergence to a bounded neighborhood. On a slippery vehicle carrying a sloshing tank and pendulum, and on coupled Duffing oscillators, it improves disturbance prediction and tracking versus fixed-decay, linear time-invariant, and model-based PD baselines.
- Hard-EM with a Kalman smoother learns contractive disturbance dynamics.
- The method extends fixed-decay adaptation to predictive disturbance control.
- The controller converges exponentially to a bounded tracking neighborhood.
- Tests cover a sloshing-load vehicle and coupled Duffing oscillators.
From the original article
1 min read at arxiv.orgWe present a representation-learning framework for composite adaptive tracking control under dynamically coupled disturbances. The framework connects classical disturbance-accommodating control (DAC) to recent last-layer adaptive disturbance-rejection methods.
By Min Kim · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.