Core Concepts
Star-Searcherは、複雑な未知の環境での自律目標検索に特化した効率的な航空システムです。
Abstract
I. Abstract
Star-Searcher introduces a specialized aerial system for autonomous target search in complex unknown environments.
Path planning challenges are addressed through hierarchical planning with visibility-based viewpoint clustering.
The system demonstrates superior performance in simulation and real-world experiments.
II. Introduction
UAVs are essential for exploring hazardous environments and conducting target searches.
Autonomous target search requires exploration and inspection simultaneously, posing unique challenges.
III. Problem Formulation
Autonomous target search involves thorough observation of occupied spaces to detect targets.
The task is considered complete when all areas have been searched.
IV. System Design
The system utilizes specialized sensors like LIDAR and cameras for efficient mapping and inspection.
Mapping modules continuously update occupancy probability and observation distances for thorough inspections.
V. Inspection-Aware Lidar-Camera Mapping and Viewpoint Generation
Traditional occupancy mapping is enhanced by integrating measurements from LIDAR and cameras.
Viewpoints are generated based on information gain and viewing angles to ensure accurate target detection.
VI. History-Aware Hierarchical Planner
Visibility-based viewpoint clustering divides viewpoints into clusters for efficient path planning.
History-aware global path planning maintains consistency between replanned paths, enhancing efficiency.
VII. Results
A. Benchmark Comparisons (Simulation Experiments)
Our method outperforms existing approaches in terms of flight time, path length, and completeness in various scenes.
B. Ablation Study
Viewpoint clustering reduces detours, while history-aware global planning improves planning consistency.
C. Real-world Experiments
Real-world experiments validate the effectiveness of the proposed method in complex scenarios.
VIII. Conclusion
The Star-Searcher system offers an efficient solution for autonomous target search in complex unknown environments, demonstrating superior performance in both simulation and real-world experiments.
Stats
Star-searcherは、飛行時間が153秒であり、パス長が191メートルである。
提案手法は、最高の完全性を示しました。