Active Next-Best-View Optimization for Risk-Averse Path Planning

2025-10-07

AI Summary

This paper addresses safe navigation under uncertainty by proposing a unified planning framework that jointly optimizes risk-averse path planning and active perception. Methodologically, it constructs a tail-sensitive risk map from an online-updated 3D Gaussian splatting radiance field to refine coarse reference trajectories, while simultaneously optimizing the next-best-view (NBV) on the SE(3) manifold to maximize information gain. Its key contribution lies in the first integrated formulation of risk-aware trajectory optimization and NBV selection, enabled by a scalable Riemannian gradient decomposition algorithm that supports efficient online updates in complex environments. Experimental results demonstrate that the framework significantly reduces navigation risk, enhances trajectory safety and robustness, and enables efficient, locally feasible perception–motion co-planning in dynamic settings.

Technology Category

• Planning, Routing, and Scheduling: Replanning and Plan Repair
• Reasoning under Uncertainty: Sequential Decision Making
• Humans and AI: Human-Aware Planning and Behavior Prediction

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: Efficient manipulation of static and dynamic Web-related graphs
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web

Abstract

Safe navigation in uncertain environments requires planning methods that integrate risk aversion with active perception. In this work, we present a unified framework that refines a coarse reference path by constructing tail-sensitive risk maps from Average Value-at-Risk statistics on an online-updated 3D Gaussian-splat Radiance Field. These maps enable the generation of locally safe and feasible trajectories. In parallel, we formulate Next-Best-View (NBV) selection as an optimization problem on the SE(3) pose manifold, where Riemannian gradient descent maximizes an expected information gain objective to reduce uncertainty most critical for imminent motion. Our approach advances the state-of-the-art by coupling risk-averse path refinement with NBV planning, while introducing scalable gradient decompositions that support efficient online updates in complex environments. We demonstrate the effectiveness of the proposed framework through extensive computational studies.

Problem

• Integrating risk aversion with active perception for safe navigation
• Refining coarse paths using tail-sensitive risk maps from 3D reconstruction
• Optimizing Next-Best-View selection to reduce critical motion uncertainty

Innovation

• Refines paths using tail-sensitive risk maps
• Optimizes NBV selection on SE(3) manifold
• Couples risk-averse planning with active perception