3D manipulation policies derive actions from explicit scene geometry rather than color images alone, allowing them to reason over metric distances, object relations, and robot-scene reachability. But commodity RGB-D sensors often return missing measurements, boundary artifacts, and plausible but wrong surfaces around transparent, reflective, thin, or textureless objects. The paper observes a sharp depth-noise cliff across modern 3D policies, even when they are trained with matched noisy depth.
To close this residual gap, the authors propose Noise-Invariant Spatial Encoding (NISE), a 3D policy architecture that learns trustworthy geometry features despite imperfect depth. NISE uses Geometric Cross-View Self-Supervision to shape shared policy features with calibrated multi-view depth, and Gaussian Marginalized 3D Attention to carry position uncertainty into deployed attention rather than treating every lifted token as exact.
Across simulated benchmarks with sensor-realistic corruption and real-robot tasks involving transparent and reflective materials, NISE improves depth-noise robustness while preserving clean-depth performance. The work supports 3D policies designed for commodity RGB-D deployment rather than clean-depth benchmarks alone.