Efficient Camera Pose Augmentation for View Generalization in Robotic Policy Learning

2026-03-30

AI Summary

Existing 2D visual motion policies suffer from limited generalization under novel viewpoints due to their reliance on static observations. This work proposes GenSplat, a framework that reconstructs scenes in high-fidelity 3D through a single forward pass of 3D Gaussian Splatting (3DGS) from sparse, uncalibrated images, leveraging a permutation-equivariant network architecture. To prevent geometric collapse, the method introduces 3D prior distillation as a regularization mechanism and trains the policy using synthetic data generated via multi-view rendering. By grounding agent decisions in consistent underlying 3D structure, GenSplat significantly enhances robustness to viewpoint variations, substantially outperforming baseline approaches under severe spatial perturbations and markedly improving task success rates in unseen viewpoints.

Technology Category

• Computer Vision: 3D Computer Vision
• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Planning, Routing, and Scheduling: Replanning and Plan Repair

Application Category

• Economics, Online Markets and Human Computation: Uses of LLMs and GenAI for marketplace design, bidding, and strategic interactions
• Search and Retrieval-Augmented AI: Agentic search
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Prevailing 2D-centric visuomotor policies exhibit a pronounced deficiency in novel view generalization, as their reliance on static observations hinders consistent action mapping across unseen views. In response, we introduce GenSplat, a feed-forward 3D Gaussian Splatting framework that facilitates view-generalized policy learning through novel view rendering. GenSplat employs a permutation-equivariant architecture to reconstruct high-fidelity 3D scenes from sparse, uncalibrated inputs in a single forward pass. To ensure structural integrity, we design a 3D-prior distillation strategy that regularizes the 3DGS optimization, preventing the geometric collapse typical of purely photometric supervision. By rendering diverse synthetic views from these stable 3D representations, we systematically augment the observational manifold during training. This augmentation forces the policy to ground its decisions in underlying 3D structures, thereby ensuring robust execution under severe spatial perturbations where baselines severely degrade.

Problem

• view generalization
• camera pose
• visuomotor policy
• 3D representation
• novel view

Innovation

• 3D Gaussian Splatting
• view generalization
• permutation-equivariant architecture
• 3D-prior distillation
• camera pose augmentation