UAV Autonomy Research - Challenges and Advantages of a Fully Distributed System Architecture

Abstract

A fully distributed system concept solely based on networked sensor and processing subsystem modules, both on-board of an UAV and on the side of the user interface, allows scalable systems, adaptability of the control system hierarchy without hardware changes, and a quick exchange of individual system modules. Reliable results in UAV autonomy research are requiring documentation and detailed reasoning of events and decisions during the flight tests, especially in experiments with real UAV demonstrators. This must also be ensured in the case of not available communication links. The optimal interaction of various modules on the ground and in the air is essential for an efficient overall system. To prove this, a synchronous recording of data from various sensors and automation modules as well as the recording of all user inputs is necessary. This article discusses the challenges and advantages of a fully distributed system approach. The discussion is based on experiences from an open source based implementation and deployment on two research UAV demonstrators of different payload classes (electrical glider and turbine-driven rotorcraft).