Using Jason and MOISE+ to Develop a Team of Cowboys

This paper gives an overview of a multi-agent system forming a team of “cowboys” to compete in the Multi-Agent Programming Contest 2008 (the ‘Cows and Herders’ scenario). In the two previous contests, we tested and improved Jason [2], an agent platform based on an extension of an agent-oriented programming language called AgentSpeak(L) [5]. The language is inspired by the BDI architecture, thus based on notions such as goals, plans, beliefs, intentions, etc. The participation in previous contests also increased our experience both in using BDI concepts as well as in programming agents with Jason specifically. In the 2006 contest, the focus was on the definition of agent’s plans [1], leading to rather reactive agents. In the 2007 contest, the focus was on (declarative) goals [3], leading to more pro-active, goal-directed agents. For the 2008 contest, we were motivated to continue improving the multi-agent programming abstractions, now towards social or organisational agents, using the concepts such as roles and groups. The system is therefore developed in two dimensions: agents (using declarative goals) and organisation (using groups, roles, and shared goals). Among several organisational models available, we will use theMOISE model because it is well integrated with Jason [4]. Our objective in participating in this contest was thus twofold: (i) to continue to test and improve Jason and its integration withMOISE; (ii) evaluate the use of organisational constructs in the development of the team.