ABSTRACT
This paper introduces a novel framework for learning optimal seminorms to automate contraction analysis of switched systems, addressing challenges in metric construction, mode-dependent behavior, and switching-induced transients.
PAPER · PDF
Loading PDF…
Key findings
A parameterized family of seminorm-based metrics generalizes weighted Euclidean norms for anisotropic contraction analysis.
A physics-informed neural network encodes contraction conditions as differentiable constraints.
A convex optimization layer provides formal verification of learned contraction certificates.
A multi-objective learning framework balances contraction strength, metric complexity, and robustness to mode transitions.
Limitations & open questions
The paper does not discuss the limitations of the proposed approach.