Design of elderly-assistant mobile servant robot

Abstract

Recently, elderly population increasing worldwide has put higher pressure on health-care providers and their families. The advent of elderly care robots will reduce that pressure. In this paper, a design of mobile servant robot with integrated tracking algorithm in order to assist the elderly by companionship is proposed not only to help families take care of their elderly at home but also reduce the pressure on health-care providers. The proposed robot is based on humanoid structure and AI-embedded-GPU controller. The design allows the robot to follow the elderly and accompany them in real-time. In addition, the video streaming algorithm with the pipeline mechanism is integrated on robot controllers so that the owner interacts with the elderly through the internet. The robot controller is embedded into hardware of 128 graphics processing unit cores and 4 ARM Cortex-A9 cores in order to execute convolutional neural network (NCNN) algorithms for elderly recognition and body tracking. The processing speed at 14 fps of video stream in real-time. The proposed robot can move on uneven surfaces with a speed at 0.21 m/s and an accuracy over 90%. However, the video stream processing speed is able to be reduced at 15 fps and latency less than 415 ms when four users appear concurrently.