Padgham & Winikoff, 2004. Developing intelligent agents systems: A practical guide. John Wiley and Sons.
In recent years, there has been an increased interest in the development and deployment of surveillance systems. Such systems pose a number of challenges, including the need to integrate disparate information sources into a cohesive "picture" of the tactical situation. Software agents have proven to be very effective in this regard because they are autonomous, reactive, proactive and inherently "social", that is, they can communicate with one another to yield a high-level appraisal of the situation being surveilled.
The development of a mobile person-following robot is an interesting application area that combines multiple autonomous agents that need to be goal-directed without losing the ability to react to the target's movements. This capability is particularly challenging for robotic systems, and so requires careful design to obtain an effective decomposition of functions into agents. Such functions include compiling sensor data from multiple modalities and performing actions via various effectors. In a recent paper, Gascueña and Fernández-Caballero argue that these complexities are best handled through the adoption of an agent-oriented design and programming methodology, and in particular, one that supports the modelling of perception and action.
The Prometheus methodology (Padgham & Winikoff, 2004) is a complete approach to specifying and designing agent-based software systems. The Prometheus Design Tool can transform the design into JACK definitions that form the basis for implementing a running system, providing a significant boost in productivity and reliability. This approach also ensures that the design and implementation do not diverge, a classic problem in software engineering. If the design is allowed to become out of date, it becomes more of a hindrance than a help when working on the software application.
This work is still in progress, and the authors report that they will now tackle a multi-robot scenario using the same approach.
Antonio Fernández-Caballero pictured top right
José Manuel Gascueña pictured bottom left