Adaptive Sampling-Based Next Best View for UAVs in Large-Scale Mapping

Robotics
ACQUAL
Staff Involved
Topic description

Recent developments in 3D scene reconstruction and autonomous inspection have significantly benefited from Next-Best-View (NBV) planning. NBV planning is one of the main ways for viewpoint optimization to obtain high-quality 3D data efficiently. NBV planning aims at minimizing data redundancy and operating costs while maximizing model quality. Adaptive Sampling-Based NBV for UAVs in Large-Scale Surface Mapping focuses on optimizing UAV trajectories for efficiently mapping surfaces like terrains, solar farms, or roofs. By integrating voxel-based adaptive sampling, the approach dynamically prioritizes high-information areas while minimizing redundancy. This method is expected to ensure accurate surface reconstruction, reduce computational load, and improve flight efficiency. 

Topic objectives and methodology

The objective is to develop an adaptive sampling-based NBV framework for UAVs to optimize large-scale mapping. The methodology involves:

  • Integrating VSLAM for real-time UAV localization and mapping.
  • Designing a voxel-based representation to dynamically evaluate surface coverage.
  • Implementing adaptive sampling algorithms to identify high-information viewpoints.
  • Planning efficient UAV trajectories while minimizing redundancy.
  • Validating performance through simulation platforms like Blender, AirSim, Isaac SIM, or another suitable tool.
References for further reading

Take Your Best Shot: Sampling-Based Next-Best-View Planning for Autonomous Photography & Inspection
Shijie Gao, Lauren Bramblett, Nicola Bezzo