Innovation-triggered Learning for Data-driven Predictive Control: Deterministic and Stochastic Formulations

29 Jan 2024  ·  Kaikai Zheng, Dawei Shi, Sandra Hirche, Yang Shi ·

Data-driven control has attracted lots of attention in recent years, especially for plants that are difficult to model based on first-principle. In particular, a key issue in data-driven approaches is how to make efficient use of data as the abundance of data becomes overwhelming. {To address this issue, this work proposes an innovation-triggered learning framework and a corresponding data-driven controller design approach with guaranteed stability. Specifically, we consider a linear time-invariant system with unknown dynamics subject to deterministic/stochastic disturbances, respectively. Two kinds of data selection mechanisms are proposed by online evaluating the innovation contained in the sampled data, wherein the innovation is quantified by its effect of shrinking the set of potential system dynamics that are compatible with the sampled data. Next, after introducing a stability criterion using the set-valued estimation of system dynamics, a robust data-driven predictive controller is designed by minimizing a worst-case cost function.} The closed-loop stability of the data-driven predictive controller equipped with the innovation-triggered learning protocol is proved with a high probability framework. Finally, numerical experiments are performed to verify the validity of the proposed approaches, and the characteristics and the selection principle of the learning hyper-parameter are also discussed.

PDF Abstract
No code implementations yet. Submit your code now

Tasks


Datasets


  Add Datasets introduced or used in this paper

Results from the Paper


  Submit results from this paper to get state-of-the-art GitHub badges and help the community compare results to other papers.

Methods


No methods listed for this paper. Add relevant methods here