Squint: Fast Visual Reinforcement Learning for Sim-to-Real Robotics
This addresses the problem of expensive visual reinforcement learning for robotics, offering a fast training solution for sim-to-real applications.
The paper tackled the challenge of slow visual reinforcement learning for robotics by introducing Squint, a method that trains policies in 15 minutes on a single GPU, with most tasks converging in under 6 minutes, and demonstrated sim-to-real transfer to a real robot.
Visual reinforcement learning is appealing for robotics but expensive -- off-policy methods are sample-efficient yet slow; on-policy methods parallelize well but waste samples. Recent work has shown that off-policy methods can train faster than on-policy methods in wall-clock time for state-based control. Extending this to vision remains challenging, where high-dimensional input images complicate training dynamics and introduce substantial storage and encoding overhead. To address these challenges, we introduce Squint, a visual Soft Actor Critic method that achieves faster wall-clock training than prior visual off-policy and on-policy methods. Squint achieves this via parallel simulation, a distributional critic, resolution squinting, layer normalization, a tuned update-to-data ratio, and an optimized implementation. We evaluate on the SO-101 Task Set, a new suite of eight manipulation tasks in ManiSkill3 with heavy domain randomization, and demonstrate sim-to-real transfer to a real SO-101 robot. We train policies for 15 minutes on a single RTX 3090 GPU, with most tasks converging in under 6 minutes.