Robotic Assistance for Skill Requiring Tasks – Perspectives with Applications in Industry and Medical Robotics

In this talk I will introduce my research vision on developing closed loop robotic assistance to humans for fine manipulation tasks. This vision is composed of understanding human behavior, identifying aspects for assistance, and finally implementation of robotic assistance. In my recent research, I focus on manual welding, a highly skill requiring industrial task, and aim at understanding the differences between professional and novice manual welders by measuring and comparing their hand impedance during welding. My research results indicate inferior impedance levels and larger position variations with the novice welders. In the assistance phase, I provide impedance compensation by means of a virtual dynamics implemented with an interactive robot. I estimate the intended path of human manipulation by a smooth Kalman filter and suppress the hand tremor across this path by impedance compensation. The deviations from the estimated path of intention give way to generation of real-time feedback alarms for training of novice welders to notice them about the increase of their hand tremor. I provide visual (flashing light) and audio (beep sound) alarms for this purpose. My long term research goal is to further develop and apply this assistance/training scheme to other industrial tasks (such as painting, polishing, fine manufacturing) and to minimally invasive surgery (MIS) in medical robotics. My analysis will expand towards identifying muscle and neural correlates of skilled manipulation, by means of recordings of arm muscle activation (EMG recordings) and monitoring the cortical brain activity (NIRS brain monitoring).