Data-Driven Control Based on Control Barrier Functions with Recursive Feasibility Guarantee
- Jiaping Liao,
- Weiming Xiang(corresponding author)
- Augusta University,
Open access
Abstract
We propose an identification-free, data-driven control framework that enforces hard state and input constraints at each step, ensuring recursive feasibility and robustness. From sampled data, we derive a tractable uncertainty description for the dynamics that captures modeling error and measurement noise. Building on this description, we develop a polyhedral algorithm that constructs a control-invariant set (CIS) within the safety set and, via a data-driven robustness verification theorem, yields a two stage strategy for constructing a robust control-invariant set (RCIS). We then formulate a planner with control Lya punov function and control barrier function (CLF–CBF) using the constructed CIS/RCIS. We prove that the resulting planner is recursively feasible and that, for all admissible uncertainty and measurement noise, closed-loop trajectories remain within the safety set. Together, these components form an identification-free pipeline from offline data to an online planner with certified safety, robustness, and recursive feasibility. Lane-keeping (LK) and adaptive cruise control (ACC) simulations for autonomous vehicles, with noisy measurements, demonstrate effectiveness.
Publication Information
Output type
Original language
English (US)Journal (Volume, Issue Number)
IEEE Transactions on Automatic ControlPublication milestones
- Accepted/In press - 2026
Publication status
ISSN
0018-9286Publication IDs
- Scopus: 105043739103
