AISPO: Enhancing Depth Reliability for Robotic Manipulation of Non-Lambertian Objects via Affine-Invariant Shape Prior

2026-06-24

AI Summary

This work addresses the challenge of robotic grasping failures caused by unreliable depth perception on non-Lambertian objects, such as transparent or highly reflective surfaces. To this end, the authors propose a depth completion method that fuses multi-scale RGB-D features with an affine-invariant shape prior. By incorporating geometric consistency constraints and prioritizing physically plausible depth predictions, the approach significantly enhances both the structural completeness and physical realism of reconstructed depth maps. Compared to existing methods, the proposed technique demonstrates superior generalization to unseen objects and novel scenes, achieving state-of-the-art performance across multiple benchmarks and real-world grasping experiments—particularly yielding a marked improvement in manipulation success rates for transparent objects.

Abstract

Reliable depth perception is critical for robotic manipulation, especially for non-Lambertian objects such as transparent or highly specular surfaces, where raw depth measurements are often corrupted or missing. These failures frequently propagate to motion planning, resulting in invalid grasp poses and execution errors. We propose AISPO, a depth completion framework that improves depth reliability for manipulation in challenging sensing conditions. AISPO combines multi-scale RGB-D feature fusion with an affine-invariant shape prior to enforce geometric consistency and mitigate catastrophic depth failures. Unlike methods that focus primarily on average depth accuracy, our approach emphasizes physical plausibility and structural integrity of the predicted depth maps. Extensive benchmark evaluations demonstrate competitive performance and strong generalization to unseen objects and novel scenes. Real-world grasping experiments further show that enhanced depth reliability significantly improves manipulation success rates, particularly for transparent objects where many existing methods fail to produce physically usable depth estimates.

Problem

• depth reliability
• non-Lambertian objects
• robotic manipulation
• transparent objects
• depth perception

Innovation

• affine-invariant shape prior
• depth completion
• non-Lambertian objects
• robotic manipulation
• geometric consistency