Multi-Agent Coordination in Dynamic Task Environments

This paper considers a dynamic variant of a coordination problem that was studied in, e.g., [9, 7, 8]. The original problem is concerned with a static task environment where all composite tasks consisting of elementary tasks and their interdependencies are given in advance. We introduce a dynamic variant of this problem where multiple composite tasks arrive dynamically in the system. We present a framework for dynamic task environments and explain the coordination problem that arises if multiple autonomous agents have to work together on parts of a composite task. From this explanation of the coordination problem we derive a formal definition. As an approach to solve the coordination problem, we suggest the idea of task partitioning. We conclude the paper with a simple solution method: the partitioning algorithm. This algorithm was originally designed to work in static environments, but turned out to be applicable in a dynamic context as well.