Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper introduces PIECE-CD, a novel self-tuning regulation algorithm for piecewise-stationary autoregressive systems with exogenous inputs, achieving logarithmic regret \(O((C+1)\log((T+1)/δ))\) under bounded disturbances and sufficient time between changes. The method leverages passive change detection by analyzing output energy under the current controller to identify coefficient shifts, distinguishing it from switching bandit approaches where unobserved changes can occur. The core innovation lies in an energy test comparing windowed output power to a noise floor, extended to detect destabilizing changes by monitoring reference controller input proposals, thereby providing a robust control strategy for dynamic environments.
Logarithmic regret is achieved in adaptive control of unknown, changing systems by exploiting the unique energy signatures of admissible plant changes during exploitation.
We study minimum-variance control of an unknown autoregressive system with exogenous inputs and coefficients that change at unknown times. Under bounded independent disturbances, fixed detection gaps, stability and feasibility conditions, and sufficient time between changes, we prove \(O((C+1)\log((T+1)/δ))\) regret with probability at least \(1-δ\), where \(T\) is the horizon and \(C\) the number of changes. Unlike switching bandits, where unselected arms can change unobserved, admissible plant changes provide information during exploitation: the correct feasible controller leaves only the disturbance in the output, whereas a detectable change raises output energy under the old controller. PIECE-CD explores initially and after alarms, then uses gated recursive least squares for control. Its energy test compares windowed output power with a threshold above the noise floor; the extension to unstable controller mismatches also monitors the reference controller's input proposal. We control false alarms across the horizon and prove logarithmic detection delay. Inputs are clipped to prescribed bounds. Logarithmic regret also holds under an explicit condition ensuring that clipping becomes inactive after a finite burn-in. Under the stated feasibility conditions, the extended detector covers destabilizing changes with detectable excess energy over a fixed window.