MARC - Applying Multi-Agent Systems to Service Robot Control

The development of mobile robots, capable of interacting with other agents in real-world environments, provides researchers in both Science and Engineering with a large body of theoretical and implementation problems that need to be overcome. One of these is the question of how to construct a complete robot control architecture, which allows the robot to perform different tasks, while processing numerous sources of information from the outside world. This topic of robot control architectures has been widely studied for a number of years. This thesis proposes the adoption of the Multi-Agent System (MAS) metaphor as the main modelling and design abstraction in the construction of software systems for service robot control. Based on this proposal, a MAS Architecture for Robot Control (MARC) was developed. This architecture directly aims to bridge the limitations of robustness, openness, and scalability currently seen in many service robot applications. MARC was realised via an implemented framework, which provides for the creation of robot control agents in a complete agent community. With an agent infrastructure in place, a scenario aimed at demonstrating the usefulness of MARC was conceived, and coded for with the creation of appropriate agents. Although the capabilities of the robot used were limited, it was sufficient to demonstrate how the MARC architecture gave the robot an amount of intrinsic social and task performance intelligence.