Solving 'Still life' with soft constraints and bucket elimination

Abstract

In this paper we study the aplicability of bucket elimination (BE) to the problem of finding stilllife patterns. Very recently, it has been tackled using integer programming and constraint programming, both of them being search-based methods. We show that BE, which is based on dynamic programming, provides an exponentially lower worstcase complexity than search methods. Then, we empirically show that BE is quite competitive with search-based approaches. Finally, we show how BE can be adapted to exploit the problem symmetries, with which in several cases we outperform previous results.