We’re a spinoff from the University of Minnesota UAS Research Labs and we’re developing a low-latency, deterministic, scalable flight control system. The UAS Research Labs primarily uses drones as a low cost means for taking navigation and control law research through flight test, but in more recent years, we’ve worked with other departments and organizations to research bears, invasive species (plant and insect), and precision agriculture.

You may wonder why we’ve created our own flight control system, when there is the PixHawk and many similar systems available. First, the UAS Research Lab started in 2006, and at the time the only relatively low-cost option was the Cloud Cap Piccolo, which was not open source enough to use for research. So the lab started with an MPC-5200B Tiny processor running eCos and a custom daughterboard. The University continuously made improvements to that system and has been using it up until only a few years ago. In fact, we made similar systems for San Jose State University, NASA Armstrong Flight Research Center, and DLR (the German Space Agency) to conduct navigation and flight controls research.

A few years ago, we recognized that our system was growing outdated. We developed an interim solution combining the Teensy 3.2 with the MPC-5200B in a little processor / big processor arrangement to conduct morphing wing research for NASA. The Teensy handled sensor and actuator I/O and the MPC-5200B handled the navigation and flight control algorithms. And we developed a system with a local UAS company around a PIC little processor with a BeagleBone Black big processor, before internally developing a flight control system around a Teensy 3.6 little processor with a BeagleBone Black big processor.

We developed our own systems, rather than use PixHawk or similar systems, primarily because of latency and determinism. To effectively conduct navigation and flight control research, we need to know that our algorithms are running at a constant rate with a constant latency between taking sensor samples and outputting actuator commands; otherwise, we can’t assess the performance and robustness of the system. Latency and determinism also play a role in the other research we conduct. Using PixHawk or similar systems in precision agriculture, for example, huge amounts of time are spent post-processing images because the UAS position, orientation, and the time of the shutter aren’t well known. And that’s primarily due to latency and determinism issues with the current flight control systems that are available. The same can be said about any use of drones to collect data, accurate knowledge of the drone position, orientation, and when data was collected is key to getting accurate data output without needing to spend nearly as much time and money in post-processing. With better flight systems, you can reduce the overlap needed in collecting data, move towards direct georeferencing instead of costly manual post-processing, and get better, more usable data output.

We also needed a flight control system that could scale well from relatively simple drones to extremely complex aircraft with huge amounts of sensor and actuator I/O. We created a system of Nodes that can connect to our little processor / big processor central flight computer using overclocked I2C to transfer data and two digital lines to keep data collection and actuator commands in sync. This allows us to scale our system to a virtually unlimited number of sensors and actuators while maintaining determinism and a constant, well defined latency. You can read more about the components and architecture here.

The University is using this system on upcoming morphing aircraft flight tests as part of a grant with NASA. This grant is to demonstrate flexible aircraft that can morph their shape to optimize performance across the flight envelope. This technology will enable future commercial aircraft that are much more fuel efficient and quiet.

These tests will be conducted on a vehicle called mAEWing2. The mAEWing2 aircraft is a 14 ft wingspan, 40 lb vehicle with 10 trailing edge control surfaces and 2 leading edge control surfaces. To measure and control the wing motion in flight, it has IMU packages located at the fore and aft outboard wing, midspan wing, and in the center-body as well as two IMU's for the rigid body control laws for 12 IMU's in total! Additionally, it has a five hole probe to measure wind angles, control surface position sensors, servo voltage sensors, and motor voltage / current sensors.

Earlier this year, we collaborated with Tao Systems to test its hot-film sensors in the Texas A&M University Low Speed Wind Tunnel on a prototype of the mAEWing2 wing using a prototype of our flight control system node. The wing from this vehicle is flexible enough for Tao Systems to collect data and research the incorporation of hot-film sensors for real-time force/moment sensing as part of aeroservoelastic control laws. The Raven Sensor and Actuator Node and hot-film sensors were integrated with the wing at the University of Minnesota in February and the wind tunnel testing was conducted in March. The sensor and actuator node collected data from the 4 wing IMU’s and 6 control surface position sensors as well as issued commands to the 6 wing control surfaces. Data collection and actuation was conducted at a 500 Hz frame rate with no dropped frames. A higher frame rate could have been used if it was just the Raven Sensor and Actuator Node running, but the wing data was being collected and logged by a National Instruments DAQ, which was also collecting wind tunnel data and hot-film data. It was several very full days of wind tunnel testing, but the data collection was very successful and the data looked excellent.

Click image for larger version. 

Name:	IMG_0054.jpg 
Views:	32 
Size:	91.3 KB 
ID:	12208

The mAEWing2 will likely start flight tests in the spring. A great walk around video of the aircraft is here:

Last year, a prototype of the flight control system was used by the University in partnership with Barron Associates to conduct spin testing on a modified Ultra Stick 120. They’re writing flight software to detect spins and automatically recover. In that case, we were measuring the aircraft motion, control surface positions, and wind angles with a couple of angle of attack and sideslip vanes.

Currently, an Ultra Stick 120 is being prepared to conduct flying qualities research in partnership with STI. Flying qualities is an important characteristic to making aircraft that are safe for pilots to fly. It’s a well researched field for manned aircraft, but not for drones. So they’ll begin collecting aircraft parameter and performance data as well as pilot feedback over a wide ranging series of flight tests.

The University is also building an X-UAV Talon aircraft for invasive insect data collection in partnership with the MN Department of Agriculture. They’ll be flying with deployable petri dishes under the aircraft wings and collecting samples from a variety of altitudes to map the presence and concentration of these insects.

Finally, the University is also using the flight control system for high altitude balloon and small satellite experiments. In the past, these have researched using Pulsars for low cost satellite navigation and they're currently measuring X-ray and gamma ray emissions.

We're current working on flight software for the FMU, Node, and BeagleBone Black. In the meantime, we have low level drivers available for all of our sensors available on our GitHub and we've started selling the hardware outside of the University, primarily targeted at businesses, research institutions, and makers who enjoying playing with code and are willing to get their hands dirty in software. You can find much more information on our website, http://bolderflight.com, including a technical overview, detailed information on our flight computer, nodes, air data sensors and breakout boards. Plus I'd be more than happy to answer any questions here!