A Multi-Agent Approach to Simulate Collaborative Classroom Activities Using Digital Humans

This paper describes a system that employs digital humans to simulate human activities in a classroom setting. A key contribution of our approach is support for collaborative multi-agent activities, such as group work. Existing works largely lack this feature, which limits the behaviors they can simulate. Our approach also introduces audio to support applications that leverage audio features. Diversity of simulated class meetings is produced by randomizing elements such as the number of agents, activity sequencing, and action motions. We introduce our system design and report initial results.