Expected Minimax Regret Rates for Adaptive LQG Control on Fixed Parameter Sets
Abstract
We study how a fixed model family determines expected minimax regret in adaptive LQG control. The family is compact and affine, with partial observation and unknown noise covariances. We quantify how information from residual covariances and input responses can vanish relative to differences between optimal control laws. This relation yields matching bounds that classify regret as bounded, logarithmic, sublinear polynomial or linear. Polynomial exponents are rational. For real algebraic model and cost data, a terminating procedure determines the regret class and the exact exponent of polynomial growth before data collection. A causal policy attains these rates using model estimation and bounded probes, reselecting model sets from observations without a supplied common stabilizer. For costs with fixed positive definite output and input weights, polynomial exponents are at most one half.
Related
- Expected Utility Regret Rule: Minimax and Bayes Optimal Portfolio Choice
- Simple-regret rates and minimax optimality of fixed-prior expected improvement in Matérn and squared-exponential RKHSs
- Controller Identifiability: The Boundary Between Logarithmic and Square-Root Local Minimax Regret in Structured Adaptive LQR
- Sharp Critical Minimax Laws and No-Learning Thresholds in Continuous-Time Adaptive Control
- Nash Social Welfare for Multi Armed Bandits: Trajectory-wise Expected and High Probability Regret
BibTeX
@misc{liu2026expected,
title = {{Expected Minimax Regret Rates for Adaptive LQG Control on Fixed Parameter Sets}},
author = {Zhaobo Liu},
howpublished = {arXiv},
year = {2026},
month = nov,
doi = {10.48550/arxiv.2611.01739},
eprint = {2611.01739},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2611.01739},
}
RIS
TY - UNPB
TI - Expected Minimax Regret Rates for Adaptive LQG Control on Fixed Parameter Sets
AU - Zhaobo Liu
PY - 2026
DA - 2026/11/01
PB - arXiv
DO - 10.48550/arxiv.2611.01739
UR - https://arxiv.org/abs/2611.01739
AB - We study how a fixed model family determines expected minimax regret in adaptive LQG control. The family is compact and affine, with partial observation and unknown noise covariances. We quantify how information from residual covariances and input responses can vanish relative to differences between optimal control laws. This relation yields matching bounds that classify regret as bounded, logarithmic, sublinear polynomial or linear. Polynomial exponents are rational. For real algebraic model and cost data, a terminating procedure determines the regret class and the exact exponent of polynomial growth before data collection. A causal policy attains these rates using model estimation and bounded probes, reselecting model sets from observations without a supplied common stabilizer. For costs with fixed positive definite output and input weights, polynomial exponents are at most one half.
ER -