Efficiency despite anisotropy: a case study.
Alex Vladimirsky (Math, Cornell)

Static Hamilton-Jacobi-Bellman PDEs arise in many applications including optimal control and in front propagation problems. The "isotropic" case reduces to an Eikonal equation with additional "causal" properties very useful in building fast (non-iterative) numerical methods. Two such Dijkstra-like methods were introduced by Tsitsiklis and Sethian in the mid-1990s.

In this tutorial-style talk, I will describe several competing approaches to causal discretizations of more general (anisotropic) problems. The focus of the presentation will be on the fundamental/simple geometric ideas rather than the technical details.