Autonomous pet-type robot with emotion behaviour-learning system based on neuromodulators
Shunsuke Akiguchi, Y. Maeda · Proceedings of the Institution of Mechanical Engineers Part I Journal of Systems and Control Engineering · 2006
In recent years, research on embedding human emotions in autonomous mobile robots has been carried out. Research based not only on psychological but also on physiological knowledge of emotions has also been performed, and research in neuroscience is gradually making clear the role of neuromodulators. Meanwhile, other researchers have examined the influence between neuromodulators and meta-parameters of the reinforcement system for learning. In this research, attempts are made to develop an emotion generation system for a pet-type robot based on the relationship between the human learning system's meta-parameters, emotions, and neuromodulators, such as adrenaline, noradrenaline, serotonin, and dopamine. The proposed architecture for the emotion generation system uses a Q-learning method with meta-parameters controlled by neuromodulators. Experimental results confirming the efficiency of the method are reported.