Measuring Visual Generalization in Continuous Control from Pixels
This work addresses the challenge of visual generalization for reinforcement learning agents in real-world environments, though it is incremental as it builds on existing techniques.
The authors tackled the problem of visual generalization in continuous control from pixels by proposing a challenging benchmark with graphical variety, finding that current methods struggle to generalize across diverse visual changes, with data augmentation outperforming self-supervised learning and more significant transformations improving generalization.
Self-supervised learning and data augmentation have significantly reduced the performance gap between state and image-based reinforcement learning agents in continuous control tasks. However, it is still unclear whether current techniques can face a variety of visual conditions required by real-world environments. We propose a challenging benchmark that tests agents' visual generalization by adding graphical variety to existing continuous control domains. Our empirical analysis shows that current methods struggle to generalize across a diverse set of visual changes, and we examine the specific factors of variation that make these tasks difficult. We find that data augmentation techniques outperform self-supervised learning approaches and that more significant image transformations provide better visual generalization \footnote{The benchmark and our augmented actor-critic implementation are open-sourced @ https://github.com/QData/dmc_remastered)