Location
MC 6460
Speaker
Sze Zheng Yong, Northeastern University
Title
Interval Bounding via Mixed Monotonicity: Applications to Learning CBFs and Designing Set-Valued Observers
Abstract
Interval bounding of nonlinear functions and dynamical systems is essential for analysis, verification, and control of safety-critical systems under uncertainty. Mixed-monotonicity, together with Lipschitz and Jacobian interpolation, offers an effective framework for constructing tight interval enclosures. Leveraging these tools, we develop methods for semi-parametric learning of safe data-driven control barrier functions (CBFs) for unknown continuous-time systems from noisy data. In addition, we show that the same framework enables the design of interval and polytopic observers under less restrictive assumptions than existing approaches. These results highlight the versatility of mixed-monotonicity-based interval bounding, advancing both data-driven safety certification and robust observer design for uncertain systems.