English

H\"older equivalence of the value function for control-affine systems

Optimization and Control 2019-02-20 v2

Abstract

We prove the continuity and the H\"older equivalence w.r.t.\ an Euclidean distance of the value function associated with the L1L^1 cost of the control-affine system q˙=\drift(q)+j=1mujfj(q)\dot q = \drift(q)+\sum_{j=1}^m u_j f_j(q), satisfying the strong H\"ormander condition. This is done by proving a result in the same spirit as the Ball-Box theorem for driftless (or sub-Riemannian) systems. The techniques used are based on a reduction of the control-affine system to a linear but time-dependent one, for which we are able to define a generalization of the nilpotent approximation and through which we derive estimates for the shape of the reachable sets. Finally, we also prove the continuity of the value function associated with the L1L^1 cost of time-dependent systems of the form q˙=j=1mujfjt(q)\dot q = \sum_{j=1}^m u_j f_j^t(q).

Cite

@article{arxiv.1304.6649,
  title  = {H\"older equivalence of the value function for control-affine systems},
  author = {Dario Prandi},
  journal= {arXiv preprint arXiv:1304.6649},
  year   = {2019}
}

Comments

25 pages, some corrections

R2 v1 2026-06-22T00:05:40.509Z