A role-based cognitive architecture for multi-agent teaming

Agent teaming is a subfield of multi-agent systems that is mainly composed of artificial intelligence and distributed computing techniques. Autonomous agents are required to be able to adapt and learn in uncertain environments via communication and collaboration in both competitive and cooperative situations. The joint intension and sharedPlan are two most popular theories for the teamwork of multi-agent systems. However, there is no clear guideline for designing and implementing agents' teaming. As a popular cognitive architecture, the BDI (Belief, Desire, and Intension) architecture has been widely used to design multi-agent systems. In this aspect, flexible multi-agent decision making requires effective reactions and adaptation to dynamic environment under time pressure, especially in real-time and dynamic systems. Due to the inherent complexity of real-time, stochastic, and dynamic environments, it is often extremely complex and difficult to formally verify their properties a priori. For real-time, non-deterministic and dynamic systems, it is often difficult to generate enough episodes via real applications for training the goal-oriented agent's individual and cooperative learning abilities. In this article, a role-based BDI framework is presented to facilitate optimization problems at the team level such as competitive, cooperation, and coordination problems. This BDI framework is extended on the commercial agent software development environment known as JACK Teams. The layered architecture has been used to group the agents' competitive and cooperative behaviors. In addition, we present the use of reinforcement learning techniques to learn different behaviors through experience. These issues have been investigated and analyzed using a real-time 2D simulation environment known as SoccerBots. © 2010 Springer-Verlag Berlin Heidelberg.