Beyond Domain Randomization: Event-Inspired Perception for Visually Robust Adversarial Imitation from Videos
This addresses the challenge of visual robustness in imitation learning for robotics and AI systems, offering a novel approach that is more efficient than domain randomization.
The paper tackles the problem of imitation learning from videos failing due to domain shifts like lighting or texture differences, by proposing an event-inspired perception method that converts RGB videos into sparse event-based representations to achieve robust imitation without expensive data augmentation, achieving strong performance on benchmarks like DeepMind Control Suite and Adroit.
Imitation from videos often fails when expert demonstrations and learner environments exhibit domain shifts, such as discrepancies in lighting, color, or texture. While visual randomization partially addresses this problem by augmenting training data, it remains computationally intensive and inherently reactive, struggling with unseen scenarios. We propose a different approach: instead of randomizing appearances, we eliminate their influence entirely by rethinking the sensory representation itself. Inspired by biological vision systems that prioritize temporal transients (e.g., retinal ganglion cells) and by recent sensor advancements, we introduce event-inspired perception for visually robust imitation. Our method converts standard RGB videos into a sparse, event-based representation that encodes temporal intensity gradients, discarding static appearance features. This biologically grounded approach disentangles motion dynamics from visual style, enabling robust visual imitation from observations even in the presence of visual mismatches between expert and agent environments. By training policies on event streams, we achieve invariance to appearance-based distractors without requiring computationally expensive and environment-specific data augmentation techniques. Experiments across the DeepMind Control Suite and the Adroit platform for dynamic dexterous manipulation show the efficacy of our method. Our code is publicly available at Eb-LAIfO.