Welcome, Commercial Drone Pilots!
Join our growing community today!
Sign up

Fully autonomus landing demo (video) with augmented reality trajectory gates.

clolsonus

Well-Known Member
Joined
Feb 22, 2019
Messages
229
Reaction score
148
Age
55
Just something fun I was fiddling around with this weekend ... drawing the trajectory gates for the AuraUAS autonomous [fixed wing] landing system. We are using it on an in-house built ag drone trying to sample/capture insects at different altitudes. The main beast of interest is the spotted wing drosophila (an invasive fruit fly) that decimates fruit and berry crops:

 
  • Like
Reactions: JDSimpson
That is pretty cool, other than seemed like a pretty hard landing, might try programming a little flare.
 
Thanks for the checking out the video. :) The camera is angled down which contributes to making it look like it didn't flare much (and on roll out it's probably only 8" above the ground so that also probably looks weird to someone who is used to the full size picture.) The tail skid actually touched first, and also I don't have a radar altimeter installed and the gps/baro just isn't accurate to the cm in altitude. Anyway, for this particular airplane size/mass it's actually a nice safe plunk down.
 
I'm not lazy, I just apply my energies in less visible areas ... and at different times. :)
 
I'm not lazy, I just apply my energies in less visible areas ... and at different times. :)

I'm sorry, I wasn't implying you were lazy. I'm just an old school guy that learned to fly without all the digital help. I flown a few with glass panels and they are really cool, and makes instrument approaches almost idiot proof.
 
No worries, after I posted my reply I realized I might have implied I thought you were implying I was lazy which wasn't my intent at all. :) I'm an old school computer geek and pilot wanna-be that always found full scale aviation about 2-5x more expensive than I could afford, so I have fun with RC airplanes and managed to wiggle my way into a UAV related day job.
 
  • Like
Reactions: BigAl07
Uncle Sam taught me to fly for free, wasn't that nice of him. RC aircraft seem like they would be fun, but if I get anymore expensive hobbies wife will shoot me.
 
  • Like
Reactions: clolsonus
Excellent work. Thanks for sharing your video.

How are you capturing insects etc in flight?
 
We have 4 petri dishes converted to traps (with sticky stuff on the flat surface). These are attached to brackets connected to retract mechanisms. At launch and landing the traps fold up flush with the wing, and once at altitude we extend the traps into the windstream. They caused a lot of drag so we only do 4 traps.
 
  • Like
Reactions: BigAl07
I created another video of the UMN AEM auto-land system. This is an augmented reality hud overlaid on action cam footage. I draw trajectory gates in 3d. All in all the rendering is tool I use to evaluate how well the autopilot system is doing. If there are any imperfections (like the horizon doesn't line up exactly) this is good because it shows I'm not cheating and I have more room for improvement. Did you catch the nearly full moon that shows up on the downwind and base leg? Did you watch the end of the video where I pick the airplane up and look back up the approach where I just came from? You can see the trajectory gates along with my actual path threading through them. The actual landing approach starts about t=1:35 into the video. We can fly up to sunset +30 minutes if the aircraft is properly lit, so we have it decked out with bright led light strips. The purpose of the mission is to hang insect traps out in the air and catch insects at various altitudes ... which is why we are flying at dusk when the insects are most active.


The full 30 minute flight (along with many others) is available on my raw/boring flight video channel: AuraUAS
 
I find what your doing very interesting. Autonomous landing have been around for a while. As the technology progresses I can see a day when we only have one pilot on many airline flights. Reality is they don't have much to do now in most cases.
 
I find what your doing very interesting. Autonomous landing have been around for a while. As the technology progresses I can see a day when we only have one pilot on many airline flights. Reality is they don't have much to do now in most cases.

Very true ... there isn't a lot of magic in doing an autonomous landing, but it is fun to build up a system yourself from scratch and then have it work pretty well. Here's a quick video from last night taken from the ground perspective looking at the airplane:

 
I take it then you are programming the approach, glide slope, approach speed yourself. I have a cousin the works for Garmin aviation division in Virginia and it truly amazes me what what they can do today. If I were a young man and had the aptitude for it I would get into software engineering. Keep at it your doing a great job. From what he has told me the computer flight control systems can handle just about any situation they are programmed for, he said the problem comes when a problem is encountered that the flight systems are't programmed to handle.
 
I take it then you are programming the approach, glide slope, approach speed yourself. I have a cousin the works for Garmin aviation division in Virginia and it truly amazes me what what they can do today. If I were a young man and had the aptitude for it I would get into software engineering. Keep at it your doing a great job. From what he has told me the computer flight control systems can handle just about any situation they are programmed for, he said the problem comes when a problem is encountered that the flight systems are't programmed to handle.

Yes, our lab at the U of MN has developed our own inhouse flight control hardware and software to support our various research projects. So for the autonomous landing stuff, I've programmed and set that up all myself (along with testing, tuning, and occasional mishaps.) I probably started messing with autoland 6-7 years ago and the code/strategies have been evolving and improving ever since. Also the augmented reality HUD and visualization I posted above is all developed in house as a tool for seeing what our autopilot is thinking and doing ... but when we get out at dusk, it also makes for cool videos.

Our systems intentionally try to keep things simple. Other people can do everything for everyone on every kind of computer board, sensor, and vehicle. We try to narrow our focus and do a good job at just a few things. Our system pretty much assumes all the sensors will be working correctly. We have a ground control station with a bunch of gauges and data so hopefully in many cases, the operator can see a problem with a bit of advance warning before it goes flight critical. If something does completely fail, then we refer back to our definition of line of sight (we can see the speck and see it well enough to judge orientation and manually fly like an RC airplane.) At that point it's up to the pilot on the ground to take the best action possible and hopefully save the airplane (with the highest priority of course being the safety of the people.)

I've had an elevator servo fail hard over (full up elevator). That led to the expected stall, plunge, stall, plunge sequence ... I manually took over control of the airplane and did my best to roll the airplane 90 degrees knife edge at the peak of it's stall and then roll back level as it started to plunge ... this got me a little bit under control. I timed the last stall just right and plopped the airplane down right at my feet. Now every time I crash a plane doing something stupid, I think back to the one time I got lucky (or good?) and actually saved the airplane. So I remember I even have video of that flight ... jump to about 8 min 20 seconds into this video to where the issue starts:

A few months later I had a similar servo failure on one of the aileron servos. In that case I figured out something was really wrong, tried to fly it manually but completely failed, so I put it back on autopilot and that was successful enough to get the airplane back closer at which point I took back manually control again and sort of flopped it into the bean field next to our club field. Fairly minor damage.

So the moral of the story is this: the fancy $2.87 digital servos from hobby king (versus the $1.98 analog servos) still are terrible. It's worth spending money on quality components and tossing anything that is even slightly suspect. In both cases I was flying a $1000 camera on board ... so why risk that to save a couple bucks on el cheapo servos? I dunno ... I knew better and did it anyway and got sort of lucky, but ended up needing to replace the servos with better ones anyway.
 
That's impressive, you timed that stall perfectly. You are correct, cheap components will enevitably end it bad results. Keep up the great work. Are you a software engineering student?
 
That's impressive, you timed that stall perfectly. You are correct, cheap components will enevitably end it bad results. Keep up the great work. Are you a software engineering student?

My personal background is I'm a life-long model airplane enthusiast with a computer science degree. When I started school I wanted to be an aero engineer, but for a couple reasons bailed on that idea and finished with a csci degree. The air force wasn't too impressed with my eye sight. In my next life maybe I'll get better eyes and could go be a real pilot. Later in grad school I got hooked up with a student job in mech. eng. (via a former roommate connection.) One thing led to another and I found myself working on driving simulators full time and doing flight simulator stuff on the side. Along the way I found a few connections in the aero eng. department at my school. Fast forward a few more years and I was winding down a few projects where we were trying to develop a marinized drone for ocean survey work. One of my aero engineering connections called me up and said "I know you wouldn't be interested, but would you be interested ..." so long story short, life came full circle on me and now I'm working (or at least trying to fake it) as an aerospace engineer. Without saying my exact age, my oldest just started college this fall and lots of things hurt when I walk, and I take my glasses on and off a lot ... :)

I know there are many great paths through life, but one of my favorite things about working at a university is being around people that know so much more than me and I can constantly be learning. I'm also finding a lot of joy in working with students that are just starting out. It's fun to see their enthusiasm, fun to see how quickly they learn and how much they grow in just a couple years of learning. Most of these kids are going to pass me up in no time, so if I can help a few of them scoot ahead just a bit quicker, I'm happy to do that. The rest we send over to the parachute testing lab. :)

I'm super grateful for the opportunities that have come my way over the years and really enjoy being able to work on a variety of interesting projects, continue to build things once in a while, do a lot of software work which I guess is what I'm good at, and still get to head outside and fly now and then along with all my other duties.
 
In my next life maybe I'll get better eyes and could go be a real pilot.

Personally I think you are too intelligent to be a pilot. Pilots today aren't what they once were, it is many hours of nothing to do. Most of the work is entering the flight, taxi out, take off, and sit on you a.... while auto takes you for a ride. Military aviation isn't much different other than you get to blow things up once in a while. I was in anti submarine warfare, hours on end flying around the ocean tracking a sub when you can find it or just flying a grid and wait for one to pop a periscope up or do something stupid.

This is the thing that gets me about aviation today, airline pilots spend very little time actually flying, but they log all the time auto is flying as if they were flying. Think of trying to keep situational awareness for several hours when you have nothing to do. Stick with engineering.
 

Members online

Forum statistics

Threads
4,277
Messages
37,605
Members
5,969
Latest member
KC5JIM