Preference-Driven Active 3D Scene Representation for Robotic Inspection in Nuclear Decommissioning

2025-04-02

AI Summary

In constrained environments such as nuclear facility decommissioning, conventional 3D scene representations neglect operational objectives—including safety coverage and task-oriented viewpoints. To address this, we propose an operator-preference-driven active 3D scene representation framework. Our method is the first to integrate Reinforcement Learning from Human Feedback (RLHF) into active 3D perception, unifying interactive preference elicitation, explicit geometric modeling, and implicit policy optimization to enable human-robot collaborative inspection path planning and adaptive focus control. Evaluated on a reactor floor tile inspection task, our approach achieves a 37% improvement in critical detail detection rate while simultaneously enhancing trajectory safety and task utility—outperforming all baseline methods significantly.

Technology Category

• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Knowledge Representation and Reasoning: Preferences
• Humans and AI: Human-Aware Planning and Behavior Prediction

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Active 3D scene representation is pivotal in modern robotics applications, including remote inspection, manipulation, and telepresence. Traditional methods primarily optimize geometric fidelity or rendering accuracy, but often overlook operator-specific objectives, such as safety-critical coverage or task-driven viewpoints. This limitation leads to suboptimal viewpoint selection, particularly in constrained environments such as nuclear decommissioning. To bridge this gap, we introduce a novel framework that integrates expert operator preferences into the active 3D scene representation pipeline. Specifically, we employ Reinforcement Learning from Human Feedback (RLHF) to guide robotic path planning, reshaping the reward function based on expert input. To capture operator-specific priorities, we conduct interactive choice experiments that evaluate user preferences in 3D scene representation. We validate our framework using a UR3e robotic arm for reactor tile inspection in a nuclear decommissioning scenario. Compared to baseline methods, our approach enhances scene representation while optimizing trajectory efficiency. The RLHF-based policy consistently outperforms random selection, prioritizing task-critical details. By unifying explicit 3D geometric modeling with implicit human-in-the-loop optimization, this work establishes a foundation for adaptive, safety-critical robotic perception systems, paving the way for enhanced automation in nuclear decommissioning, remote maintenance, and other high-risk environments.

Problem

• Integrates operator preferences into 3D scene representation for robotics
• Uses RLHF to optimize robotic path planning with expert feedback
• Improves safety-critical coverage in nuclear decommissioning inspections

Innovation

• Integrates expert preferences into 3D scene representation
• Uses RLHF for robotic path planning optimization
• Combines geometric modeling with human-in-the-loop optimization