From Saturday, October 13 until Wednesday October 17, we had the opportunity to travel with Prof. Ankur Desai and our ATM OCN 404 (Meteorological Measurements) students to Champaign, Illinois, to participate in the SAVANT field campaign led by Prof. April Hiscox (University of South Carolina).
In addition to the unique opportunity for AOS undergraduates to see a meteorological field experiment in action and to make contributions with their own complementary measurements, it was also our first chance to fly our newly acquired Elanus Duo fixed wing unmanned aerial vehicle (UAV, or drone) carrying a FLIR Dual Pro R camera that captures both visible and thermal infrared images. The purpose was to map surface temperature at sunrise over the 100-acre farm field, where SAVANT scientists are studying night-time drainage winds.
We had originally planned to fly our manned ultralight airplane for this purpose, which would have given us considerably greater flexibility with respect to altitude, among other things, but the pilot (me) is temporarily grounded for medical reasons.
Having never flown a fixed wing drone under autopilot control before, let alone one carrying an expensive instrument like the FLIR camera, I was admittedly anxious about all the things that could go wrong. I spent most of the three weeks prior to our departure to Illinois carefully testing and debugging all the critical systems, including motors, control servos, telemetry radios, and autopilot programming. There is unfortunately no ideal place near Madison (but away from the suburbs) to test-fly a fixed wing drone of this size and speed, so it would have to wait until we reached the SAVANT field site.
Our maiden flight was on the afternoon of October 15. The goal for this daytime flight was not to obtain scientifically useful thermal IR images but simply to do a dry run, including getting the hang of hand launches of the drone while everyone was comfortably warm and there was good light.
I asked student Leo Mikula to be the one to toss this relatively large bird into the air while I manned the remote control box. The procedure essentially involves putting the drone into autopilot mode, waiting a couple seconds for the motors to reach full power, throwing it like a football into the air at a 30-degree upward angle, and then hoping that the power of the motors would accelerate the drone to a safe flying speed before it hit the ground! Having never been through this before, it was hard to guess how forgiving the drone would be of mistakes.
I could tell Leo was feeling a bit of pressure (as I would have too), but I tried to make clear that if anything went wrong, I wouldn’t hold it against him. None of us had any experience with this, and any one of us launching the drone would have had an equal chance of flubbing it (or watching something else go wrong). In the end, all of our anxiety proved unnecessary, as the launch went off perfectly (click for video):
We all watched in awe as the drone reliably and precisely flew my pre-programmed “lawnmower” survey pattern back and forth at the FAA-mandated maximum altitude of 400′ over the 100-acre field, taking about a half hour to do so. Average ground speed was 18 m/sec, or about 40 mph, so it effectively flew about 20 miles during the time it was aloft. Throughout this time, the FLIR camera was snapping away through a hole cut in the belly of the aircraft at a rate of one image per second.
The next step, and the part I worried about most, was the landing. As they say in aviation, “Takeoffs are optional; landings are mandatory.” The autopilot has an auto-land function that I hoped I could trust, but I also knew that it would depend on my having set up the landing sequence in a way that would allow the aircraft to descend on a reasonable glide slope toward its planned touchdown point. It also depended on the autopilot responding appropriately to guidance from the laser altimeter despite the fact that the ground we planned to land on was far from level.
On the first approach to landing, it was clear that the aircraft was too hot and high, so I initiated a go-around, put the airplane into a temporary loiter mode in which it steadily circled a fixed point, and reprogrammed the landing sequence to try to bring it to a lower altitude before turning final. The adjustment wasn’t perfect, but the airplane did make a successful, if admittedly inelegant belly landing (click for video):
Back in my hotel, I retrieved the 1800+ FLIR images from the tiny SD card in the camera and trimmed the total selection down to the 340 that gave the essential coverage of the field. Later, these were uploaded to the Pix4D cloud-based software system for stitching the geotagged visible images into a photomosaic (orthomap) and for photogrammetrically deriving a map of terrain height.
Below is just a small sample of the individual images used, followed by the straight orthomap as well as the orthomap colored-coded for terrain elevation (blue = low; red = high):
The next step would be to fly the UAV at sunrise so as to capture the nighttime patterns of ground temperature over the SAVANT site, but I’ll cover that (successful) flight when I have the yet-to-be-derived thermal orthomap in hand.
PS: This is my first blog post in almost a year. Suffice it to say that 2018 got off to a very bad start, healthwise, and I am only just now finally getting back to doing blog-worthy things.