Raise BDI Agents, First Steps

Recent developments in game engines have opened the door to three-dimensionalworldsforagents. VEsNA(VirtualEnvironments via Natural language Agents) is a framework for building embodied BDI agents extending the Jason framework. It allows agents to have a situated body, communicate with other agents and users, interact with objects, and reason logically about space. VEsNA agents also interact with humans in natural language through ChatBDI and act according to a simulated personality and mood. The framework has been validated through several experiments: an evaluation of the nl2kqml natural language pipeline, an assessment of temperdriven plan selection, a complex multi-agent scenario developed in collaboration with the game company Untold Games, and a series of theoretical contributions extending the syntax and semantics of BDI agents. In this paper, I summarise the work carried out during the first two years of my PhD, outline the trajectory for the final year, and critically assess the strengths, limitations, and practical potential of VEsNA.

Authors

• Andrea Gatti Università degli Studi di Genova

Keywords

• BDI Agents
• Virtual Environments
• NLP
• Logics