Modular Cooperative Tasking for Multi-agent Systems

Abstract

This paper develops a modular framework for cooperative tasking for multi-agent systems. The paper focuses on logical behaviors of multi-agent systems that can be represented by automata. The main challenge here is that adding a new global task and designing a new set of local controllers for cooperative set of agents, whether the composition of the new and old tasks can be modularly achieved by the composition of new and old local controllers. The result proves that the decomposability of two successive global task automata and the existence of corresponding local controllers for each global task is a sufficient condition for modular cooperative tasking to guarantee the achievement of the modular task (the composition of the old and new global tasks). The decomposability of the modular task, on the other hand, turns to be not required for the modular cooperative tasking; however, it is shown to be a necessary condition for the decomposability of the old and new task automata. The concept of modular cooperative tasking is illustrated through a cooperative robotic example.