Grounding Agents in EMud Artificial Worlds

This paper suggests that in the context of autonomous agents and generation of intelligent behavior for such agents, a more important focus should be held on the symbolic context that forms the basis of computer programs. Basically, software agents are symbolic entities living in a symbolic world and this has an eeect on how one should think about designing frameworks for their evolution or learning. We will relate the problem of symbol grounding to that of sensory information available to agents. We will then introduce an experimental environment based on virtual worlds called EMuds, where both human and artiicial agents can interact. Next, we show how it can be applied in the framework of multi-agent systems to address emergence based problems and report preliminary results. We then conclude with some ongoing and future work.