A Lightweight Controller for Autonomous Following of a Target Platform for Drones

Abstract

Drones or Unmanned Aerial Vehicles (UAVs) are providing interminable opportunities to capture high-quality video feeds that were previously impossible and have transformed the digital era. Many applications today require computer vision (CV) and machine learning (ML) techniques to extract the useful information captured from the onboard camera, and send it to an embedded controller that can make independent decisions. For instance, maneuvering the drone to follow a target platform by using only the onboard camera feed is critical in target tracking. Therefore, in this paper, we exploit the applicability of a low-computational embedded tracking controller to follow a target platform e.g. a car or pedestrian, and thus, react in real-time, adjusting the drone steering angles and velocity. We developed a system that enables drones to follow a target platform autonomously without requiring continuous human intervention on an embedded state-of-the-art STM32 Nucleo board. The system includes a lightweight controller that controls the drone's movement and enables it to track and follow a target platform accurately. To validate the performance of our embedded controller, we performed a number of experiments in an open-source AirSim simulator. The tracking controller footprint and memory consumption was less than 3%, and was able to reliably track and trail the target platform in different environmental conditions.