A task definition language for virtual agents

Abstract

The use of Virtual Environments as a user interface can be important for certain types of applications, especially in the fields of education and entertainment. These synthetic worlds are even more attractive for the user when they exhibit dynamic characteristics and are populated by virtual agents. There is, however, a lack of generalpurpose tools for designing and implementing intelligent virtual environments, and especially in the case of defining virtual agents’ tasks, where there is a strong dependence between the task execution and the context. In this paper, we present our approach towards a context-independent definition of tasks using a high-level language. With the proposed task definition language, one can combine numerous built-in functions and commands to describe complex tasks as a combination of parallel, sequential and conditional execution of actions. It can be used to program complicated virtual agent interactions with the environment without going into much detail on how these tasks are implemented and how parallelism is achieved. The main advantage of the proposed language is that it enables tasks to be easily constructed and reused by different agents and in different environments. Our approach has been based on SimHuman, a platform for rendering and animating Virtual Agents in real-time.