{"id":"arxiv/2609.38051","url":"https://arxiv.org/abs/2609.38051","tags":["math.OC","eess.SY"],"score":0,"title":"Non-Holonomic Gradient Play: Leafwise Nash Equilibria, Stability, and Deception","domain":"math.OC","source":"arxiv","authors":["Mahmoud Abdelgalil","Miroslav Krstic","Jorge I. Poveda"],"reviews":0,"subject":"math.OC","summary":"We study generalized learning dynamics in multi-agent systems whose joint state evolves on a manifold and whose agents act through state-dependent, potentially nonholonomic vector fields. Under a bundle-splitting condition, we show that these dynamics admit an intrinsic representation as projected Riemannian gradients, giving rise to a class of \\emph{nonholonomic gradient play} dynamics. We characterize the local stability of its equilibria through an intrinsic linearization that explicitly captures the effects of the Riemannian connection and the nonholonomy of the actuation frame. The…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-09-29","reactions":{}}