A multi-agent system which involves several agents that collaborate towards the achievement of a joint objective is viewed as a team of agents. Most proposed teamwork structures (e.g. joint intentions [7], shared plans [13]) rely on agents in a multi-agent system to negotiate and/or contract with each other in order to initiate team plans. However, in dynamic, real-time domains with limited communication, complex negotiation protocols may take too much time and/or be infeasible due to communication restrictions.
Our work has been focused in time-critical environments in which agents in a team alternate between periods of limited and unlimited communication. This focus motivates the introduction of Periodic Team Synchronization (PTS) domains. In PTS domains, during the limited communication periods, agents need to act autonomously, while still working towards a common team goal. Time-critical environments require real-time response and therefore eliminate the possibility of heavy communication among team agents. However, in PTS domains, agents can periodically synchronize in a safe, full-communication setting.
Our work is situated in an example of a PTS domain, simulated robotic soccer [16]. In robotic soccer, teams can plan strategies before the game, at halftime, or at other breakpoints, but during the course of the game, communication is limited. The soccer server [25] is a widely used robotic soccer simulator with a single, low-bandwidth, unreliable communication channel for all agents in the environment [1].
In this article, we describe our general team member agent architecture suitable for creating teams of agents in PTS domains. This architecture includes a mechanism for defining pre-determined multi-agent protocols accessible to the entire team, called locker-room agreements. Within this team member agent architecture and using the locker-room agreement, we define our flexible teamwork structure that allows for task decomposition and dynamic role assignment in PTS domains. In addition, we define a communication protocol in service of the locker-room agreement that is suitable for use during low-communication periods in a class of PTS domains: domains with low-bandwidth, single channel communication environments.
The team member agent architecture described in this article defines a complete agent, including perception, cognition, and action. It is fully implemented as a simulated robotic soccer team.
This article is organized as follows. Section 2 expands upon the introduction of PTS domains and the two main contributions of this article: a flexible teamwork structure and a low-bandwidth communication paradigm. Section 3 presents the general agent architecture within which both the flexible teamwork structure and the low-bandwidth communication paradigm are situated. Sections 4 and 5 formally present the teamwork structure and the communication paradigm respectively. Section 6 gives details of our full implementations of both main contributions of this article within the simulated robotic soccer domain. Section 7 presents extensive empirical results testing the effectiveness of these implementations. Section 8 is devoted to discussion and related work and Section 9 concludes.