Computation of local ISS Lyapunov functions for discrete-time systems via linear programming

Abstract

This paper presents a numerical algorithm for computing ISS Lyapunov functions for discrete-time systems which are input-to-state stable (ISS) on compact subsets of the state space. The algorithm relies on solving a linear optimization problem and delivers a continuous and piecewise affine ISS Lyapunov function on a suitable triangulation covering the given compact set excluding a small neighbourhood of the origin. The objective of the linear optimization problem is to minimize the ISS gain. It is shown that for every ISS system there exist a suitable triangulation such that the proposed algorithm terminates successfully.

Safe Learning-based Observers for Unknown Nonlinear Systems using Bayesian Optimization

“…For fairness, we choose Legendre polynomials as basis functions and not monomials, which would perfectly fit the nonlinearity φpqq with very few coefficients. Our basis functions are ψpqq " 10 » ------p3q 2 1 ´1qp3q 2 2 ´1q p3q 2 1 ´1qq 2 q1 p3q 2 2 ´1q p5q 3 1 ´3q 1 q p5q in N " 200 training iterations, with 1000 uniformly random samples on r´p ‹ , p‹ s 5 to select p j at each j using (22). The surrogate GP model is constructed at each time using sklearn in Python or MATLAB's fitrgp function in the Statistics and Machine Learning Toolbox.…”

“…µ, σ Ð use GP to compute on samples drawn from P 9: p j`1 Ð obtain via expected improvement (22) 10:…”

Safe Model-based Reinforcement Learning with Stability Guarantees

“…However, for physical models, like the prior model h in (1), the energy of the system (e.g., kinetic and potential for mechanical systems) is a good candidate Lyapunov function. Moreover, it has recently been shown that it is possible to compute suitable Lyapunov functions [31,32]. In our experiments, we exploit the fact that value functions in RL are Lyapunov functions if the costs are strictly positive away from the origin.…”