HERE: Hierarchical Active Exploration of Radiance Field With Epistemic Uncertainty Minimization

2026-01-12

IEEE Robotics and Automation Letters

AI Summary

This work addresses the inefficiency in exploration and inaccurate identification of unobserved regions in active 3D reconstruction with neural radiance fields (NeRFs). To this end, the authors propose a hierarchical active exploration framework grounded in epistemic uncertainty minimization. They introduce evidential deep learning into NeRF for the first time to quantify epistemic uncertainty arising from data scarcity, and leverage this uncertainty to design a synergistic exploration strategy: local viewpoint planning based on voxel-level uncertainty and visibility, coupled with global path planning optimized for scene coverage. Experiments demonstrate that the proposed method significantly improves reconstruction completeness and fidelity across multi-scale simulated environments, and its practicality is further validated through real-world deployment on physical hardware.

Technology Category

• Reasoning under Uncertainty: Uncertainty Representations
• Machine Learning: Active Learning
• Computer Vision: 3D Computer Vision

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization

Abstract

We present HERE, an active 3D scene reconstruction framework based on neural radiance fields, enabling high-fidelity implicit mapping. Our approach centers around an active learning strategy for camera trajectory generation, driven by accurate identification of unseen regions, which supports efficient data acquisition and precise scene reconstruction. The key to our approach is epistemic uncertainty quantification based on evidential deep learning, which directly captures data insufficiency and exhibits a strong correlation with reconstruction errors. This allows our framework to more reliably identify unexplored or poorly reconstructed regions compared to existing methods, leading to more informed and targeted exploration. Additionally, we design a hierarchical exploration strategy that leverages learned epistemic uncertainty, where local planning extracts target viewpoints from high-uncertainty voxels based on visibility for trajectory generation, and global planning uses uncertainty to guide large-scale coverage for efficient and comprehensive reconstruction. The effectiveness of the proposed method in active 3D reconstruction is demonstrated by achieving higher reconstruction completeness compared to previous approaches on photorealistic simulated scenes across varying scales, while a hardware demonstration further validates its real-world applicability.

Problem

• active 3D reconstruction
• neural radiance fields
• epistemic uncertainty
• camera trajectory planning
• scene exploration

Innovation

• epistemic uncertainty
• neural radiance fields
• active exploration
• evidential deep learning
• hierarchical planning