Autonomous Quadcopter Navigation Using Depth Camera and Real-Time Kinematic GPS

Abstract

Advancements in materials, control techniques, and computing have led to improvements in modern quadcopters. However, autonomous quadcopters that are available in the market today are extremely cost-prohibitive and closed source, making them rather inaccessible to the average drone enthusiast. As such, the goal of this research was to enhance the capabilities of the Intel Aero Ready to Fly (RTF) Drone and contribute to the open source community of UAVs. More specifically, the goal of this project was to enable the Intel Aero to autonomously conduct precise landings while avoiding any obstacles that it may encounter during its flight.

Building upon previous work, this was done by developing an obstacle detection and avoidance algorithm using the on-board depth camera and integrating the Intel Aero with a Real-Time Kinematic (RTK) GPS. This GPS would allow the drone to achieve centimeter-level accuracy while landing. Targeted landing and obstacle detection and avoidance were developed separately through Python simulations and indoor testing. The field testing of the targeted landing and obstacle detection and avoidance was not ultimately successful because of some hardware and software incompatibility issues that came up later. Although field testing was not successful due to hardware failures, the algorithms and framework developed for this project can potentially be used on any quadcopter platform with a depth camera and an RTK GPS. As such, they are an important contribution to furthering the autonomous capabilities of quadcopters and to the open source community.