Intelligent autonomous drone

Abstract

Drones are used in multiple industries such as military, construction, maritime and more. It has allowed many companies to work more efficiently while reducing the manpower labor and enhancing the worker’s safety at the same time. This technology has been constantly evolving and improving. Now, with the help of Artificial Intelligence (AI) algorithm, drones will be able to perform more automated tasks and be always adaptive to its environment which includes collision avoidance.

In this project, I aimed to develop an autonomous drone application to perform inspection tasks. To achieve this objective, I will develop a Proportional-Integral-Derivative (PID)-based flight control algorithm for efficient navigation and tracking purposes. Then, I will evaluate and integrate some of the state-of-the-art AI object detection algorithms such as You-Only-Look-Once (YOLO). The AI will be able to detect structural defects that are commonly found such as cracks and corrosions.

The PID controller’s parameters were optimized through trial and errors by real flight tests. It was designed to track defects detected by the object detection algorithm and maintain its position so that picture can be clearly captured by the drone. All image processing and computation will be leveraging the NVIDIA Jetson NX edge onboard computer installed onto the DJI drone.