TY - JOUR
T1 - Can Learning Deteriorate Control? Analyzing Computational Delays in Gaussian Process-Based Event-Triggered Online Learning
AU - Dai, Xiaobing
AU - Lederer, Armin
AU - Yang, Zewen
AU - Hirche, Sandra
N1 - Publisher Copyright:
© 2023 X. Dai, A. Lederer, Z. Yang & S. Hirche.
PY - 2023
Y1 - 2023
N2 - When the dynamics of systems are unknown, supervised machine learning techniques are commonly employed to infer models from data. Gaussian process (GP) regression is a particularly popular learning method for this purpose due to the existence of prediction error bounds. Moreover, GP models can be efficiently updated online, such that event-triggered online learning strategies can be pursued to ensure specified tracking accuracies. However, existing trigger conditions must be able to be evaluated at arbitrary times, which cannot be achieved in practice due to non-negligible computation times. Therefore, we first derive a delay-aware tracking error bound, which reveals an accuracy-delay trade-off. Based on this result, we propose a novel event trigger for GP-based online learning with computational delays, which we show to offer advantages over offline trained GP models for sufficiently small computation times. Finally, we demonstrate the effectiveness of the proposed event trigger for online learning in simulations.
AB - When the dynamics of systems are unknown, supervised machine learning techniques are commonly employed to infer models from data. Gaussian process (GP) regression is a particularly popular learning method for this purpose due to the existence of prediction error bounds. Moreover, GP models can be efficiently updated online, such that event-triggered online learning strategies can be pursued to ensure specified tracking accuracies. However, existing trigger conditions must be able to be evaluated at arbitrary times, which cannot be achieved in practice due to non-negligible computation times. Therefore, we first derive a delay-aware tracking error bound, which reveals an accuracy-delay trade-off. Based on this result, we propose a novel event trigger for GP-based online learning with computational delays, which we show to offer advantages over offline trained GP models for sufficiently small computation times. Finally, we demonstrate the effectiveness of the proposed event trigger for online learning in simulations.
KW - Gaussian process regression
KW - computational delay
KW - event-triggered learning
KW - learning-based control
KW - online learning
UR - http://www.scopus.com/inward/record.url?scp=85166346683&partnerID=8YFLogxK
M3 - Conference article
AN - SCOPUS:85166346683
SN - 2640-3498
VL - 211
SP - 445
EP - 457
JO - Proceedings of Machine Learning Research
JF - Proceedings of Machine Learning Research
T2 - 5th Annual Conference on Learning for Dynamics and Control, L4DC 2023
Y2 - 15 June 2023 through 16 June 2023
ER -