3D Obstacle Avoidance for Unmanned Autonomous System (UAS)

Abstract

The goal of this thesis is to design a real-time, three-dimensional algorithm, named as the vector mesh (VM) algorithm, for unmanned aerial vehicles (UAV) to generate collision-free motion in indoor or outdoor environments with unknown obstacles. This promising technology can be utilized in both military and commercial applications. The VM approach employs three data reduction phases to compute optimal navigation directions while on-board scanning range sensor continuously updates depth data. In order to develop the VM, vector filed histogram (VFH) which applied in 2D space was first simulated in Matlab. Then a 2D autonomous navigation was implemented on a developed Vision-based Ground Vehicle (VGV) and the entire system was controlled by a modified VFH method which was computing in the Robot Operating System (ROS). Also, the VM algorithm was simulated in ROS and integrated into Gazebo simulator which is an effective graphic based robot simulator in complex indoor and outdoor environment. In this study, it has been shown that the proposed VM can be an effective 3D obstacle avoidance algorithm for typical small-UAVs if 3D information is continuously provided.