Human-Robot Collaboration via Deep Reinforcement Learning of Real-World Interactions

Abstract

We present a robotic setup for real-world testing and evaluation of human-robot and human-human collaborative learning. Leveraging the sample-efficiency of the Soft Actor-Critic algorithm, we have implemented a robotic platform able to learn a non-trivial collaborative task with a human partner, without pre-training in simulation, and using only 30 minutes of real-world interactions. This enables us to study Human-Robot and Human-Human collaborative learning through realworld interactions. We present preliminary results, showing that state-of-the-art deep learning methods can take human-robot collaborative learning a step closer to that of humans interacting with each other.

Error-related potential-based shared autonomy via deep recurrent reinforcement learning

“…The shared autonomy in human-robot interaction leverage the strengths of both human and robots, where robots can no longer act solitarily, but must share part of their autonomy space with human. In most traditional shared control tasks, the user needs to provide explicit input, such as keyboard or mouse commands [9][10][11], during interactions. BCI systems offer new channels that allow shared autonomy by integrating user intent directly according to the ongoing brain activity, thus eliminating the need to exploit muscular control [12,13].…”