A Vision-Aided Navigation System by Ground-Aerial Vehicle Cooperation for UAV in GNSS-Denied Environments

摘要

Modern UAVs are expected to be able to fly safely through the area where GNSS signal is weak or even could not be received. In order to realize it, the intuitive idea is by adding some airborne sensors and processor chips, which is, however, not practically feasible for the UAVs that have limited payload capability and online-computing resources. To address this issue, this paper presents a technical solution for building autonomous navigation system based on the concept of Ground-Aerial Vehicle Cooperation. The key features of the solution lie in four aspects: 1) A local reliable self-organizing network is constructed for transmitting control commands and estimated pose information between the UAV and a reusable ground robot; 2) A small and lightweight monocular camera is equipped on the UAV to achieve visual simultaneous localization and mapping (SLAM) considering the limitation of UAV's carrying capacity; 3) The modified ORB-SLAM2 system equipped on the ground robot rebuilds a global dense map and reconstructs the 3D environment based on the Octomap; and 4) The low-cost, commercial ground robots and rotor craft were chosen to facilitate other researchers to build a similar platform quickly. Four in-door flight experiments are conducted to verify the effectiveness of the technical solution. Other potential and extendable applications of the cooperation system include: real-time reconstruction of environmental maps, collaborative obstacle avoidance and path planning, and motion synchronization control.