Autonomous Robotic Dialogue System with Reinforcement Learning for Elderlies with Dementia

Abstract

The growing number of people living with dementia puts a heavy load on elderly care nurses. The use of humanoid robots as communication partners can help alleviate this load with the goal to assist seniors in maintaining their cognitive skills through conversation. In this paper we present an autonomous robotic dialogue system with the humanoid robot Bonoid that is capable of learning a conversation strategy from interactions with humans. The system was tested and evaluated with first results showing natural and engaging interactions with seniors, and the robotic system acquiring a conversation strategy through the use of reinforcement learning.