English

On Divergence-based Distance Functions for Multiply-connected Domains

Complex Variables 2018-01-23 v1 Analysis of PDEs

Abstract

Given a finitely-connected bounded planar domain Ω\Omega, it is possible to define a {\it divergence distance} D(x,y)D(x,y) from xΩx\in\Omega to yΩy\in\Omega, which takes into account the complex geometry of the domain. This distance function is based on the concept of ff-divergence, a distance measure traditionally used to measure the difference between two probability distributions. The relevant probability distributions in our case are the Poisson kernels of the domain at xx and at yy. We prove that for the χ2\chi^2-divergence distance, the gradient by xx of DD is opposite in direction to the gradient by xx of G(x,y)G(x,y), the Green's function with pole yy. Since GG is harmonic, this implies that DD, like GG, has a single extremum in Ω\Omega, namely at yy where DD vanishes. Thus DD can be used to trace a gradient-descent path within~Ω\Omega from xx to yy by following xD(x,y)\nabla_x D(x,y), which has significant computational advantages over tracing the gradient of GG. This result can be used for robotic path-planning in complex geometric environments.

Keywords

Cite

@article{arxiv.1801.07099,
  title  = {On Divergence-based Distance Functions for Multiply-connected Domains},
  author = {Renjie Chen and Craig Gotsman and Kai Hormann},
  journal= {arXiv preprint arXiv:1801.07099},
  year   = {2018}
}

Comments

7 pages

R2 v1 2026-06-22T23:51:54.165Z