Spatially Visual Perception for End-to-End Robotic Learning
This work addresses the challenge of environmental variability in end-to-end robotic learning, offering incremental improvements for scalable, low-cost solutions in embodied intelligence.
The paper tackles the problem of robust generalization in imitation learning for robotic control across diverse camera observations, particularly under lighting changes, by introducing a video-based spatial perception framework that integrates a novel image augmentation technique and a monocular depth estimation model, resulting in a significant boost in success rates where previous models fail.
Recent advances in imitation learning have shown significant promise for robotic control and embodied intelligence. However, achieving robust generalization across diverse mounted camera observations remains a critical challenge. In this paper, we introduce a video-based spatial perception framework that leverages 3D spatial representations to address environmental variability, with a focus on handling lighting changes. Our approach integrates a novel image augmentation technique, AugBlender, with a state-of-the-art monocular depth estimation model trained on internet-scale data. Together, these components form a cohesive system designed to enhance robustness and adaptability in dynamic scenarios. Our results demonstrate that our approach significantly boosts the success rate across diverse camera exposures, where previous models experience performance collapse. Our findings highlight the potential of video-based spatial perception models in advancing robustness for end-to-end robotic learning, paving the way for scalable, low-cost solutions in embodied intelligence.