Authors: Jacob Moore, Ian Reid, Brandon Sutherland
Mentors: Tim McLain
Insitution: Brigham Young University
Autonomous unmanned aerial vehicle (UAV) research depends on autopilots capable of integrating new, low-level estimation and control algorithms. Current autopilot software stacks (eg. PX4 and Ardupilot) are bulky and complex, making it difficult for researchers to easily integrate their own algorithms. The codebases for these autopilots are fully-featured, meaning complete understandability is impossible. Furthermore, these autopilots are targeted towards industry and plug-and-play use, not researchers. A research-centered autopilot, with a lean codebase, capable of easy algorithm integration is a needed tool for productive research. ROSflight is an autopilot designed from the ground up with researchers in mind. Currently it is based on a ROS1 framework, which is old and outdated and is reaching end of life. Our research is to update ROSflight to ROS2, the most current version of ROS. Our work includes revamping the existing fixed-wing autopilot, ROSplane, and multirotor autopilot, ROScopter, and extending ROSflight to support vertical takeoff and landing (VTOL) aircraft. Since we started the project, ROSplane and ROSflight now work under a ROS2 framework. We have verified autonomous flight on a fixed wing aircraft, capable of rejecting disturbances. Hardware demonstration of a GPS waypoint-following fixed wing aircraft is in progress. Future work includes hardware demonstration of a multirotor aircraft autopilot and support for a VTOL aircraft.