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

摘要

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