An anytime algorithm for scheduling tasks for multiagent systems

Abstract

This research proposes an any time algorithm for a task scheduling problem among agents. The tasks are composed of atomic tasks and are to be distributed to coalitions of agents as subtasks for parallel execution. We model the problem and propose an algorithm for it. The algorithm calls other low level algorithms to recursively generate plans for agents. The results show satisfactory results that the convergent times are reasonably short and are close to termination time in many settings. We also found that the distribution of input values affect the performance of the algorithm similar to the optimal coalition structure problem.