Applying BDI To Serious Games: The PRESTO Experience

This article summarizes the main results of PRESTO, a 4-year industrial research project ending in 2016. Its main objective was the development of an environment for the authoring and control of training sessions in 3D-based serious games, especially in the domain of emergency management. PRESTO adopts artificial intelligence and reusable components (models of behavior of NPCs and game-level scripts) and offers end-user tools for their development, in addition to programming frameworks and APIs. The starting point was the authors’ experiences with the BDI (Belief, Desire, Intention) agent architecture and its cognitive extensions. Given the relative immaturity of the selected game environments (Unity and XVR) from a behavioral AI perspective, a significant number of challenges concerning semantics, perception and control had to be tackled. Results presented here include game-agnostic semantization, composition of tactical agents from reusable behavioral models, agent-level scripting for game-specific behaviors, game-level scripting to represent agent- and game strategies, and graphical development environments directed at game and domain specialists rather than software engineers. An analysis methodology and a few pilot studies are introduced. The PRESTO suite, used for the development of commercial training services, is currently available only on request, for both commercial and research uses.