Further Explorations in State-Space Search for Optimal Task Scheduling

Abstract

The problem of task scheduling with communication delays is NP-hard. State-space search algorithms such as A* have been shown to be a promising approach to solving this problem optimally. A recently proposed state-space model for task scheduling, known as Allocation-Ordering (AO), allows state-space search methods to be applied to the problem of optimal task scheduling without the need for duplicate avoidance mechanisms. This paper examines the performance of two parallel search algorithms when applied to both the AO model and the older ELS state-space model. This suggests that its use may provide an advantage with many different variations on state-space search. This paper explores the application of AO to some of these variants, namely depth-first branch-and-bound (DFBnB) and parallel search. We also present an update to the formulation of AO that prevents invalid states from being considered during a search. An evaluation shows that AO gives a clear advantage to DFBnB and allows greater scalability for parallel search algorithms. The update to AO's formulation has no significant impact on performance either way.