Bridging the simulation-to-real gap of depth images for deep reinforcement learning

Yoonsu Jang, Jongchan Baek, Soo Jeon, Soohee Han

Abstract: Highlights•Overcoming challenges in applying learned policies from simulation to a real robot.•Generative Adversarial Network-based simulation-to-reality translation.•Extracting domain-invariant and task-relevant features from depth images.•Minimal performance drop and eliminates the need for post-deployment tuning.