Intention-response model based on mirror neuron and theory of mind

An agent should exhibit proper behaviour depending on user's intention to provide convenience to the user. In this regard, even though there are a lot of researches dealing with generation of responses according to user's intention, creating robust reactions for the agent in a diverse environment is still a critical problem. Also, there are only a few studies utilizing methods pertaining to the human brain process. To be able to respond to the user's intention efficiently, we imiate the mirror neuron system, which has the ability to react rapidly to simple intentions, and the theory of mind system, which is activated by complex intentions. A behavior selection network(BSN) system selects actions according to external stimuli and achieves a subgoal, which is similar to the features of the mirror neuron. However, it cannot solve complex problems; thus, we control modules of the BSN to make behavioral sequences and accomplish long-term goals. We confirm the usability of the proposed method by performing several test scenarios using the NAO robot. Experiments show that the proposed model is able to make behavioral sequences that are able to respond to simple intentions as well as complex intentions.