VRGym: A Virtual Testbed for Physical and Interactive AI
This provides a tool for researchers in robotics, machine learning, and cognitive science to advance general-purpose agents and human studies, though it is incremental as it builds on existing virtual reality and simulation concepts.
The authors tackled the problem of realistic human-robot interaction by proposing VRGym, a virtual reality testbed that generates diverse 3D scenes with high realism through physics-based simulation, enabling the collection of human interactions and support for various robots and machine learning algorithms.
We propose VRGym, a virtual reality testbed for realistic human-robot interaction. Different from existing toolkits and virtual reality environments, the VRGym emphasizes on building and training both physical and interactive agents for robotics, machine learning, and cognitive science. VRGym leverages mechanisms that can generate diverse 3D scenes with high realism through physics-based simulation. We demonstrate that VRGym is able to (i) collect human interactions and fine manipulations, (ii) accommodate various robots with a ROS bridge, (iii) support experiments for human-robot interaction, and (iv) provide toolkits for training the state-of-the-art machine learning algorithms. We hope VRGym can help to advance general-purpose robotics and machine learning agents, as well as assisting human studies in the field of cognitive science.