Optimal schedules for parallelizing anytime algorithms: the case of independent processes

The performance of anytime algorithms having a nondeterministic nature can be improved by solving simultaneously several instances of the algorithm-problem pairs. These pairs may include different instances of a problem (like starting from a different initial state), different algorithms (if several alternatives exist), or several instances of the same algorithm (for nondeterministic algorithms).In this paper we present a general framework for optimal parallelization of independent processes. We show a mathematical model for this framework, present algorithms for optimal scheduling, and demonstrate its usefulness on a real problem.