Some algorithms arising in the proof of the Kepler conjecture

Thomas C. Hales

Published 2002-05-19Version 1

By any account, the 1998 proof of the Kepler conjecture is complex. The thesis underlying this article is that the proof is complex because it is highly under-automated. Throughout that proof, manual procedures are used where automated ones would have been better suited. This article gives a series of nonlinear optimization algorithms and shows how a systematic application of these algorithms would bring substantial simplifications to the original proof. This article includes a discussion of quantifier elimination, linear assembly problems, automated inequality proving, and plane graph generation in the context of discrete geometry.