Star-Searcher: A Complete and Efficient Aerial System for Autonomous Target Search in Complex Unknown Environments

TL;DR

Star-Searcher is introduced, an aerial system featuring specialized sensor suites, mapping, and planning modules to optimize searching in complex unknown environments and employs a history-aware mechanism to reduce motion inconsistency from frequent map changes, significantly enhancing search efficiency.

Abstract

This paper tackles the challenge of autonomous target search using unmanned aerial vehicles (UAVs) in complex unknown environments. To fill the gap in systematic approaches for this task, we introduce Star-Searcher, an aerial system featuring specialized sensor suites, mapping, and planning modules to optimize searching. Path planning challenges due to increased inspection requirements are addressed through a hierarchical planner with a visibility-based viewpoint clustering method. This simplifies planning by breaking it into global and local sub-problems, ensuring efficient global and local path coverage in real-time. Furthermore, our global path planning employs a history-aware mechanism to reduce motion inconsistency from frequent map changes, significantly enhancing search efficiency. We conduct comparisons with state-of-the-art methods in both simulation and the real world, demonstrating shorter flight paths, reduced time, and higher target search completeness. Our approach will be open-sourced for community benefit at https://github.com/SYSU-STAR/STAR-Searcher.

Figures (8)

• Figure 1: (a) A test of autonomous target search conducted in a complex scene with six apriltags. (b) The apriltag search results and executed trajectory. Video of the experiments is available at: https://youtu.be/08ll_oo_DtU.
• Figure 2: An overview of our star-searcher system. (a) The hardware platform of our aerial system. (b) The UAV utilizes LIDAR point clouds and camera projection to update map information, cluster frontiers and uninspected areas. (c) For each cluster of frontiers and uninspected areas, a set of viewpoints is generated and scored based on information gain and viewing angle. The higher the score, the larger the size of the viewpoint. The one with the highest score is selected. (d) Visibility-based viewpoint clustering is performed, and history-aware global path planning is conducted based on this. (e) Local path planning.
• Figure 3: Illustration of the visibility-based viewpoint clustering. If a viewpoint has collision-free rays to all viewpoints in the current cluster, it is assigned to that cluster.
• Figure 4: A comparison between the planned trajectories with (fig. a) and without (fig. b) visibility-based viewpoint clustering (VBVC). Without viewpoint clustering, the UAV calculates the shortest path to visit all viewpoints and selects the viewpoint behind the obstacle as the next target. Once it reaches this viewpoint, it discovers a new area and generates additional viewpoints. As a result, it replans a trajectory as indicated by the red dashed line, leading to revisits later, as shown in fig. c.
• Figure 5: Illustration of the history-aware global path planning. After finding the anchor centers, multiple collision-free shortest paths are calculated using the A* algorithm between pairs of anchor centers and concatenated to update the history-aware global path. Without the history-aware path, an indecisive trajectory occurs, as indicated by the red curve in (d).