VR-Goggles for Robots: Real-to-sim Domain Adaptation for Visual Control
This addresses the problem of domain adaptation for visual control in robotics, offering a decoupled and efficient method for transferring policies from simulation to reality, though it appears incremental as it builds on existing translation techniques.
The paper tackles the reality gap in visual control by translating real-world images to synthetic domains during deployment, enabling Deep Reinforcement Learning policies trained in simulation to be applied in real-world environments without extra training steps. The approach is validated in indoor and outdoor robotics experiments, showing it as a lightweight and flexible solution.
In this paper, we deal with the reality gap from a novel perspective, targeting transferring Deep Reinforcement Learning (DRL) policies learned in simulated environments to the real-world domain for visual control tasks. Instead of adopting the common solutions to the problem by increasing the visual fidelity of synthetic images output from simulators during the training phase, we seek to tackle the problem by translating the real-world image streams back to the synthetic domain during the deployment phase, to make the robot feel at home. We propose this as a lightweight, flexible, and efficient solution for visual control, as 1) no extra transfer steps are required during the expensive training of DRL agents in simulation; 2) the trained DRL agents will not be constrained to being deployable in only one specific real-world environment; 3) the policy training and the transfer operations are decoupled, and can be conducted in parallel. Besides this, we propose a simple yet effective shift loss that is agnostic to the downstream task, to constrain the consistency between subsequent frames which is important for consistent policy outputs. We validate the shift loss for artistic style transfer for videos and domain adaptation, and validate our visual control approach in indoor and outdoor robotics experiments.