Abstract
Atmosphere representation/understanding for humans-robots interaction in multi-agent society is investigated by applying the concept of atmosfield, support vector regression, and a new visualization method using fuzzy sets. It provides the necessary information for each agent to communicate with each other smoothly. Experiments by four agents (humans) are done using 13 scenarios that cover most part of the atmosfield, and it is confirmed that the necessary atmosphere information for each agent is obtained with the average accuracy 90%. It is aimed to open a new way for the realization of a human friendly distance education system as a potential application of the proposals.
