Enhancing drone autonomy through cloud integration: a comprehensive software architecture for navigation, visual servoing, and control

This work proposes a comprehensive software framework for cloud-enabled autonomous drone navigation, featuring precise target tracking via image-based visual servoing (IBVS) coupled with a control scheme. In this study, a low-cost quadcopter running the ArduPilot firmware is evaluated within a simulation-in-the-loop (SITL) environment using a Gazebo-based simulation of a real-world mission. The tested software architecture can be seamlessly integrated with an onboard companion computer for real-time execution. The mission involves waypoint tracking, precise identification and descent onto visual markers using IBVS, along with real-time data visualization on a remote client connected via a cloud interface. Because the software architecture is versatile, it can accommodate any conventional or knowledge-based controller. To demonstrate the efficacy and robustness of the proposed architecture, the quadcopter was tested under challenging weather conditions, where it successfully completed the mission despite disturbances and sensor noise. Finally, the complete software architecture has been tested and implemented in the robot operating system (ROS).