: The Social Robot Architecture

A doctoral dissertation (Duffy, 2000, University College Dublin) and developed also by Mauro Dragone, University College Dublin

This work is a demonstration of the necessity of both physical and social embodiment in designing a robot with coherent intelligent social behaviour. It develops the current thinking on embodiment beyond the physical by demonstrating the inherent importance of social embodiment and a corresponding social framework to maximise the fundamental social attributes found when more than one robot co-inhabit a physical space.

A synthesis of existing artificial intelligence approaches, incorporating reactive and deliberative methodologies in conjunction with multi-agent systems research, is developed further into a coherent control architecture for socially-competent collectives of autonomous mobile robots.

The Social Robotic Architecture (Duffy, 2000) is a design methodology based upon the combination of BDI (Belief, Desire and Intention) agents, a reactive behavioural system, and an explicit social infrastructure. Key components of the architecture are a hardware abstraction layer for heterogeneous robot platform applications, a coherent reactive-deliberative control synthesis, a Belief-Desire-Intention Multi Agent deliberative level developed through Agent Factory (Collier, 2001), and a FIPA compliant social level to support explicit social interaction (http://www.fipa.org).

The control architecture offers basic reflexes guiding robot responses to unexpected or dangerous events (e.g. halting in case of emergency or recover from collision) together with more sophisticated behaviours like obstacle avoidance, wall following, and for the robotic soccer scenario, dribbling, tracking and passing the ball.

For more detailed information, see the publications listing.

: Artificial Social Identity

These ideas pivot around the notion of identities and characters for each robot functioning in a social environment. The foundations for the development of a control architecture with sufficient functionality and performance are outlined in the Social Robot Architecture with a detailed explanation of its structure and functionality.

For more detailed information, see the publications listing.

: The Virtual Robotic Workbench

The Virtual Robotic Workbench provides a medium for experimentation in interoperability and cooperation between heterogeneous robots (real, simulated and virtual) and humans. As robots become more and more diffuse and employed in every-day tasks, some of the important issues will be the mediation between different forms of intelligence and autonomy and the assimilation of autonomous robotic entities within ubiquitous computing infrastructures. For these reasons the Virtual Robotic Workbench (Duffy, 1999; Mauro, 2005) was developed to ease configuration and networking for large scale heterogeneous real and virtual robotic teams, thus enabling dynamic composition of sensors and actuators and supporting dynamic discovery of resources and peers. In addition, the VRW offers a framework to support social integration, implemented in a re-usable and standardised form. The core features of the Virtual Robotic Workbench are the immersion of robots in a shared collaborative environment and the adoption of mature Multi Agent Systems technology in order to enable robot-robot and human-robot interaction. This includes the Virtual Robotic Workbench and its synthesis into a coherent whole within the social robot metaphor.

Although this work deals with the explicit social interaction of two or more socially capable robots, the framework is designed and has been implemented for human-robot social interaction.

For more detailed information, see the publications listing.