SOAR: Simultaneous Exploration and Photographing with Heterogeneous UAVs for Fast Autonomous Reconstruction

TL;DR

The paper addresses rapid autonomous 3D reconstruction of complex scenes by coupling a LiDAR-equipped explorer with camera-equipped photographers in a heterogeneous multi-UAV setup. It proposes a hybrid pipeline that uses surface frontier-based exploration to quickly gather geometric information, incremental viewpoint generation to cover surfaces efficiently, and Consistent-MDMTSP to assign viewpoints to photographers with consideration for task consistency. The approach is validated in realistic simulation against state-of-the-art baselines, showing improved efficiency and reconstruction quality, and includes ablations to demonstrate the benefits of incremental viewpoint generation and the consistency-driven assignment. The work advances practical large-scale reconstruction by reducing planning time and promoting robust, parallel data capture, with future work targeting real-world communication and occlusion challenges.

Abstract

Unmanned Aerial Vehicles (UAVs) have gained significant popularity in scene reconstruction. This paper presents SOAR, a LiDAR-Visual heterogeneous multi-UAV system specifically designed for fast autonomous reconstruction of complex environments. Our system comprises a LiDAR-equipped explorer with a large field-of-view (FoV), alongside photographers equipped with cameras. To ensure rapid acquisition of the scene's surface geometry, we employ a surface frontier-based exploration strategy for the explorer. As the surface is progressively explored, we identify the uncovered areas and generate viewpoints incrementally. These viewpoints are then assigned to photographers through solving a Consistent Multiple Depot Multiple Traveling Salesman Problem (Consistent-MDMTSP), which optimizes scanning efficiency while ensuring task consistency. Finally, photographers utilize the assigned viewpoints to determine optimal coverage paths for acquiring images. We present extensive benchmarks in the realistic simulator, which validates the performance of SOAR compared with classical and state-of-the-art methods. For more details, please see our project page at https://sysu-star.github.io/SOAR}{sysu-star.github.io/SOAR.

Figures (6)

• Figure 1: (a) Illustration of the proposed framework's execution process. (b) 3D reconstruction result of the above scene produced by the proposed framework.
• Figure 2: The system overview of the proposed LiDAR-Visual heterogeneous multi-UAV system for fast aerial reconstruction.
• Figure 3: The process of incremental surface extraction. (a) and (b) represent surface clustering and explored surface detection at a specific moment, while (c) and (d) depict the same process after the map update at the next moment.
• Figure 4: Example of the multi-chromosome representation for 9 VCTs with 3 photographers. It shows a single individual of the population, which represents a single solution of the problem. The first chromosome represents the sequence of VCTs that 1st photographer needs to visit, which are 1st, 5th, and 7th VCT. Likewise, the remaining chromosomes represent the sequences of tasks for the other photographers.
• Figure 5: Trajectories generated and reconstruction results by our method, SSearchers, and Multi-EE in two scenes. Except for the explorer (the black trajectory) in our method, which does not participate in image capture, all other UAVs are involved in image acquisition tasks.