A MULTI-AGENT ARCHITECTURE FOR ROBOTIC SYSTEMS IN REAL-TIME ENVIRONMENTS

Abstract

In this paper we explore the topic of designing an architecture and processing algorithms for a multi-agent system where agents need to address unexpected events in the environment while operating under soft real-time constraints. We examine the subtopics of (1) determining the appropriate level of autonomy for the agents in the community, (2) managing the communication required to support the allocation of new tasks (including ones to address possible obstacles), and (3) being sensitive to soft real-time constraints. More specifically, we allow for certain agents to operate autonomously and also provide for a coordinator agent, to avoid some overhead in negotiation among agents. Using examples, we illustrate the usefulness of the framework for various robotics applications. In particular, we discuss the potential of the framework for simulated robotics environments such as the RoboCup Search and Rescue, which are well suited to supporting a combination of agents that are fully autonomous and agents that are controlled by a supervisor. We briefly discuss an implementation to demonstrate the effectiveness of our model and emphasize the need to reduce communication and expedite processing when operating under real-time constraints for robotics applications.