A Multi-Fidelity Bayesian Approach to Safe Controller Design

Safely controlling unknown dynamical systems is one of the biggest challenges in the field of control systems. Oftentimes, an approximate model of a system's dynamics exists which provides beneficial information for control design. However, differences between the approximate and true systems present challenges as well as safety concerns. We propose an algorithm called SafeSlope to safely evaluate points from a Gaussian process model of a function when its Lipschitz constant is unknown. We establish theoretical guarantees for the performance of SafeSlope and quantify how multi-fidelity modeling improves the algorithm's performance. Finally, we present a case where SafeSlope achieves lower cumulative regret than a naive sampling method by applying it to find the control gains of a linear time-invariant system.