UTBot: A Virtual Agent Platform for Teaching Agent System Design

Abstract

We introduce UTBot, a virtual agent platform for teaching agent system design. UTBot implements a client for the Unreal Tournament game server and Gamebots system. It provides students with the basic functionality required to start developing their own intelligent virtual agents to play autonomously UT games. UTBot includes a generic agent architecture, CAA (Context-sensitive Agent Architecture), a domain-specific world model, a visualization tool, several basic strategies (represented by internal modes and internal behaviors), and skills (represented by external behaviors). The CAA architecture can support complex long-term behaviors as well as reactive short-term behaviors. It also realizes high context-sensitivity of behaviors. We also discuss our experience using UTBot as a pedagogical tool for teaching agent system design in undergraduate Artificial Intelligence course.