Data-driven Controller Synthesis

Achieve formal safety guarantees from data

Ensuring the safety of control systems is becoming more and more important due to the increasing number of safety-critical real-life applications in the past few decades. However, obtaining accurate models for some of these applications may require a significant amount of effort, making it challenging to apply those model-based approaches to synthesize controllers and achieve safety guarantees.

Luckily, thanks to recent progress in sensors and data processing technologies, we are able to take advantage of a massive amount of data collected from physical systems. In this project, we focus on synthesizing correct-by-construction controllers leveraging data for black-box systems. In particular, we are interested in so-called direct data-driven approaches, with which controllers are synthesized directly based on data without any intermediate identification phase.

Highlights of the results:

  1. We proposed a direct data-driven approach to synthesize safety controllers for unknown linear systems affected by unknown-but-bounded disturbances. The data-driven computation is formulated as semidefinite programming problems with linear complexities with respect to the system dimension and the number of data.

  2. We proposed a direct data-driven approach for synthesizing safety controllers for continuous-time polynomial systems via data-driven computation of control barrier certificates.

Related papers

  1. B. Zhong, M. Zamani, and M. Caccamo, Synthesizing safety controllers for uncertain linear systems: A direct data-driven approach, In: Proceedings of IEEE Conference on Control Technology and Applications (CCTA), pp. 1278-1284, 2022. (Preprint)

  2. A. Nejati*, B. Zhong*, M. Caccamo and M. Zamani, Data-Driven Controller Synthesis of Unknown Nonlinear Polynomial Systems via Control Barrier Certificates, In: Learning for Dynamics and Control Conference (L4DC), PMLR 168, 2022.

  3. A. Nejati*, B. Zhong*, M. Caccamo, and M. Zamani. Data-Driven Controller Synthesis of Unknown Nonlinear Polynomial Systems via Control Barrier Certificates, In: Proceedings of the International Workshop on Computation-Aware Algorithmic Design for Cyber-Physical Systems, ACM, 2022

* Contribute equally

Related talks

  1. Presentation in IEEE Conference on Control Technology and Applications (CCTA) 2022 (Link: Bilibili / YouTube)

  2. Presentation in CAADCPS Workshop, CPS-IoT Week 2022 (Link: Bilibili / YouTube)