Fully autonomous aerial systems (FAAS) are an increasingly relevant and intriguing research topic which require considerable systems support. FAAS use unmanned aerial vehicles and edge and cloud compute resources to dynamically sense and respond to their environments without human piloting. FAAS are useful in a number of domains, but are very difficult to develop. FAAS applications require different UAV specifications, autonomy policies, machine learning algorithms, and compute architectures to execute domain specific tasks, all operating on a tight power budget.

To help support FAAS research and development, we have developed SoftwarePilot, an open source Fully Autonomous Aerial Systems middleware. SoftwarePilot provides easy to use flight control and data sensing utilities, and allows users to configure custom machine learning models for pathfinding, object detection, and data analysis to build FAAS for any application.

In this half-day virtual tutorial, we will discuss state of the art FAAS research, dive deep into SoftwarePilot’s design, and instruct users to build an example FAAS application using SoftwarePilot’s simulation environment.

Organizers:
* Jayson Boubin, The Ohio State University
* Christopher Stewart, The Ohio State University

Target audience:
Computer systems researchers with building and improving robotic, edge, and cyber-physical systems.

Prerequisites:
None