A Neurorobotic Approach of Emotion: Implemented Neurodynamics Mediate a Coupling Between Top-Down Abductive Inference and Bottom-Up Sensations

Abstract

Event Abstract Back to Event A Neurorobotic Approach of Emotion: Implemented Neurodynamics Mediate a Coupling Between Top-Down Abductive Inference and Bottom-Up Sensations Moeka Saito1, Gyanendra N. Tripathi1 and Hiroaki Wagatsuma1, 2* 1 Kyushu Institute of Technology, Graduate School of Life Science and Systems Engineering, Japan 2 RIKEN BSI, Japan A simple extension of robotics hardly accomplishes the level of what a robot knows like an educated person and becomes skeptical about the own ability and counterpart’s true intentions. It is designed in the range of interpreter or translator without any feeling. Neither they are disgusted with someone’s attitude nor indignation swells inside. A lack of feeling or sensation. An abductive inference is considered to be necessary to be conscious what happens externally and internally [1, 2]. Computational models with oscillator synchronization mechanisms have proposed to explain the process phenomenologically as a potential neurodynamics [3, 4]. A synchronization mediates a coupling between top-down abduction and bottom-up possible interpretations based on sensory signals. Recently, neurorobotic approaches to understand what the brain works are highlighted such as testable platforms for motor control and locomotion, reward systems and action selections, hippocampus and memory systems, and even for medical cares of autistic symptoms and Parkinson’s disease [5, 6, 7]. We have proposed a phenomenological model of emotion [8] and extend it to a neural dynamics toward robotic implementations [9]. A non-linear oscillator dynamics can be applied to a top-down abduction by reading information of firing phases which differentiate distributed phases and a concentrated phase depending on the spatio-temporal context, called phase coding (Figure 1). In verification of a hypothesis of Damasio [1], which focuses on the difference between feeling and emotion, i.e. changes in physical and chemical state and its awareness, computational models with neurobiological background [2] remain within the framework to test an isolated system from environmental change and influences by the presence of others. Neurorobotic approaches offer a hybrid platform to take into consideration of combinatory effects involving real-time interactions with humans, and it may develop into an effective tool to investigate how our emotion comes from. Figure 1 Acknowledgements This work is partly supported by Neuroinformatics Japan Center (NIJC), RIKEN BSI.