Lightweight Path-Aware Frontier Re-ranking for LiDAR-Based UAV Exploration

Duc Manh Do; Thanh Nguyen Canh; Ngoc Minh Do; Mai Thanh Thai; Xiem HoangVan

cs.RONov 6, 2026

Lightweight Path-Aware Frontier Re-ranking for LiDAR-Based UAV Exploration

Authors: Duc Manh Do, Thanh Nguyen Canh, Ngoc Minh Do, Mai Thanh Thai, Xiem HoangVan

Organizations: Vietnam National University, University of Engineering and Technology, Hanoi, Viet Nam. · College of Engineering and Computer Science, VinUniversity, Hanoi 100000, Viet Nam. · School of Information Science, Japan Advanced Institute of Science and Technology, Nomi 923-1292, Japan. · Center for AI Research, VinUniversity, Hanoi 100000, Viet Nam.

Abstract

Autonomous 3D exploration with UAVs requires efficient target selection under limited flight time and onboard computation. In conventional Euclidean-distance-based frontier pipelines, geometric distance may poorly represent the actual collision-free travel cost in cluttered environments, leading to inefficient frontier choices. This paper presents a lightweight path-aware frontier re-ranking method for LiDAR-based UAV exploration. The method first retains a small top-N set of promising frontier candidates using a lightweight utility function and then applies time-bounded RRT* planning to estimate their feasible path costs for re-ranking. The path generated for the selected frontier is further reused for navigation to reduce redundant global planning. Experiments in three warehouse-like environments show that the proposed method consistently produces shorter exploration trajectories than conventional frontier baselines, with traveled-distance reductions of up to 14.1%, while reducing exploration time by up to 9.5% in two environments and remaining competitive in the third. The method maintains map detail comparable to the conventional OctoMap-based baseline, demonstrating an effective trade-off between exploration efficiency and mapping quality.

Figures & tables

Fig. 1 : In Euclidean-distance-based frontier selection, a UAV may prefer a geometrically close frontier even when the corresponding collision-free path requires a long detour. The proposed method instead evaluates candidate frontiers using estimated collision-free path costs, allowing the UAV to favor candidates that can be reached more efficiently in cluttered environments.

Fig. 2 : Overall architecture of the proposed UAV-based autonomous exploration system. The UAV uses 3D LiDAR measurements together with its pose estimate to construct and update an OctoMap, denoted by ( $V$ ), where the current robot position is represented by ( $p_{rb}$ ). The exploration module detects frontier voxels, clusters them into frontier candidates, and selects the best frontier ( $v_{bf}$ ) by evaluating candidates according to their estimated collision-free path costs. The reusable-path mechanism stores and reuses the feasible path generated during frontier evaluation ( $P_{bf}$ ), thereby reducing redundant planning and accelerating subsequent navigation.

Fig. 3 : Illustration of frontier detection and candidate generation in 2D and 3D occupancy maps. Detected frontier cells are clustered to reduce storage and computational cost, and each cluster is represented by a frontier candidate point for subsequent evaluation.

Fig. 4 : Warehouse simulation environments used for evaluation. The left column shows three environments with different obstacle configurations, while the right column presents their corresponding 3D ground-truth maps. For each environment, the UAV is initialized at one of three starting positions, labeled 1, 2, and 3.

Fig. 5 : Evolution of the 3D map and UAV trajectory in Env. 3 from start position 1. The red point denotes the UAV starting position, and the dashed boxes indicate the initial explored region and its corresponding location in the final map.

Environments	Methods	Metrics
Environments	Methods	$T$ (min) $\downarrow$	$S$ (m) $\downarrow$	$M$ (%) $\uparrow$
Env. 1	MRF [ 11 ]	$8.32\pm 1.33$	$383.61\pm 32.22$	$97.60\pm 0.43$
	FCS [ 2 ]	$12.35\pm 1.96$	$507.47\pm 50.21$	$\mathbf{99.65}\pm 0.11$
	EPIC [ 4 ]	$\mathbf{6.98}\pm 1.05$	$350.32\pm 20.21$	–
	Ours	$7.53\pm 0.23$	$\mathbf{332.79}\pm 22.42$	$97.45\pm 0.45$
	Ours w/o reusable path	$7.68\pm 0.29$	$332.93\pm 20.54$	$97.79\pm 0.43$

TABLE I : Quantitative comparison of exploration performance in three warehouse-like environments.

T

= exploration time;

S

= traveled distance;

M

= map detail; the best results are highlighted in bold.

Parameter	Environment	Value	Metrics
Parameter	Environment	Value	$T$ (min) $\downarrow$	$S$ (m) $\downarrow$	$M$ (%) $\uparrow$
$N$	Env. 1	2	7.46	332.41	97.79
		3	7.53	332.79	97.45
		4	7.92	335.45	97.78
	Env. 2	2	8.75	387.93	97.44
		3	8.42	363.67	97.32

TABLE II : Parameter sensitivity analysis of the proposed method.

T

= exploration time;

S

= traveled distance;

M

= map detail; the best results are highlighted in bold.

Autonomous Robot Exploration

See all 38 papers

Explore From Sketch: Accelerating UAV Exploration in Large-scale Environments with Prior Maps
Jun 10, 2026Tiancheng Lai, Yuman Gao, Xiangyu Li +8Next-Best-View Planning Point Cloud Registration
Institute of Cyber-Systems and Control, College of Control Science and Engineering, Zhejiang University, Hangzhou 310027, China · Huzhou Institute, Zhejiang University, and Huzhou Key Laboratory of Autonomous System, Huzhou 313000, China. · Differential Robotics Technology Company, Hangzhou 311100, China. +1
Palm-sized Omnidirectional Vision-Based UAV Exploration with Sparse Topological Map Guidance
May 8, 2026Zirui Wang, Xinjia Luo, Haotian Sun +3Robot Navigation UAV Navigation
Department of Mechanical and Energy Engineering, Southern University of Science and Technology, Shenzhen, 518055, China
WOLF: World Model Guided LiDAR Exploration with Predictive Frontiers
Sep 20, 2026Yuyang Tian, Penghui Yang, Pengyuan Wu +7World Model Learning UAV Navigation
University of Science and Technology of China, Hefei, China. · Shanghai Artificial Intelligence Laboratory, Shanghai, China. · East China University of Science and Technology, Shanghai, China. +3

Jun 10, 2026J/K move · Enter open · S save

Explore From Sketch: Accelerating UAV Exploration in Large-scale Environments with Prior Maps

Tiancheng Lai, Yuman Gao, Xiangyu Li, Ruitian Pang+7

Institute of Cyber-Systems and Control, College of Control Science and Engineering, Zhejiang University, Hangzhou 310027, China · Huzhou Institute, Zhejiang University, and Huzhou Key Laboratory of Autonomous System, Huzhou 313000, China. · Differential Robotics Technology Company, Hangzhou 311100, China.+1

Next-Best-View Planning Point Cloud Registration

Open paper AskPDF arXiv

Autonomous exploration with UAVs in large-scale, topologically complex environments often suffers from low efficiency due to suboptimal scheduling and detours. Prior maps (e.g., construction drawings), although usually imprecise and flawed, are readily available in many scenarios and have the potential to provide global structural guidance. This paper presents a novel exploration framework that leverages sparse, unaligned, and even discrepant 2D prior maps for LiDAR-based UAV exploration. First, a robust 2D-3D point cloud registration pipeline is proposed to align LiDAR observations with prior maps. The registration pipeline combines a GeoContext descriptor for single-frame candidate retrieval, a multi-frame verification mechanism for coarse transformation estimation with outlier rejection, and a Scale-ICP algorithm for refinement. The registration module can handle map discrepancies and provide multiple hypotheses when geometric ambiguities arise. To effectively utilize the registration results for exploration planning, we further develop a hierarchical viewpoint planning strategy under localization uncertainties. The hierarchical strategy first spatially attaches local viewpoints to prior guidepoints and adopts a Monte Carlo Tree Search solver to determine their traversal sequence under each registration hypothesis. To mitigate registration uncertainty, a risk-aware selector evaluates prior sequences using confidence-weighted travel risk, and a fixed-endpoint traveling salesman problem is formulated to generate an efficient local coverage path under the selected prior guidance. Benchmark evaluations reveal up to 34.2% improvement in exploration efficiency and 37.9% reduction in flight distance compared to state-of-the-art methods, while extensive simulations and field experiments further demonstrate robustness to prior map incompleteness and deformations.

Abstract

Autonomous Robot Exploration

Explore similar work

Explore From Sketch: Accelerating UAV Exploration in Large-scale Environments with Prior Maps

Palm-sized Omnidirectional Vision-Based UAV Exploration with Sparse Topological Map Guidance

WOLF: World Model Guided LiDAR Exploration with Predictive Frontiers