Replies to This Discussion

Permalink Reply by Julian on August 18, 2011 at 5:16pm

Interested in the same answer, im in the midst of designing a Tailsitting VTOL for the DARPA UAVforge.net competition and i need to pick an autopilot soon. Ive been leaning towards the Ardupilot for its opensource and such, so i hope that it does have to capabilities to handle both my vertical and horizontal flight...

• ▶ Reply

Permalink Reply by Edward Strickland on August 19, 2011 at 12:36am

I only ever found one autopilot which used Quarternions, the microbot autopilot. The problem with the "1 degree" fudge in Euler frame of reference is as much as a 1,000 feet inaccuracy at very high speeds. 1 solution maybe to have no hover and hold.....or never fly horizontally. Ive a Matlab autopilot simulation connected to microsoft flight simulation and it may be possible to introduce quarternion frame of reference for simulation purposes. This all takes a lot of time and money though. Its one of the things which is holding back the VTOL transitioning into forward flight.  I have an alternative airframe design which would be more like a harrier, although this would need building, the autopilot problems arent as great. The Olecopter design is much simpler to build and has an extremely high performance BUT then the stability and autonomous issues (as above) are complex to solve.

• ▶ Reply

Permalink Reply by Julian on August 20, 2011 at 12:00pm

Sadly i know very little about the control behind UAVs, im outsourcing that job to some other students at my school lol. I settled on a tail sitter VTOL for simplicity and im hoping to do the same with the autopilot system. if the 90 degree limit does exist, could one circumvent it by physically rotating the autopilot? Barbaric, i know, but if i want something flying by the time December rolls around desperate measures have to be taken. What about the paparazzi platform. it would require alot of work, but it looks like its alot more customizable than the ardupilot in both the software and hardware sense. For the testing phases i may just grab a 3 axis gyro for manual vertical flight and let the ardupilot handle the forward flight. Really hoping i can get the airplane to be fully autonomous though :/

• ▶ Reply

Permalink Reply by Edward Strickland on August 21, 2011 at 5:58am

The olecopter, was simple to build and very powerful. I have been getting an eflux of 230 mph at 50% power from the duct. This next few weeks I will conclude the optimisation of the duct thrust, which should top out in excess of 35kgs. Machine weight so far was 17kgs BUT no range to speak of. I can solve the range and power issue easily enough. I originally built the machine and ran it but realised that at full power it would dissapear from radio range (1km) within 15 seconds. That was when I realised that I needed a smaller design although the speed will still be high. funny you mention students, I lecture at a university in engineering and maths. I discussed integrating quarternions into the matlab code for the Auto pilot, Julian have you got a simulator of an autopilot working?

• ▶ Reply

Permalink Reply by Edward Strickland on August 21, 2011 at 6:04am

I have the papparazzi autopilot and that may be worth a further look esp. as the Ground control station is now Apple Lion compatible. The 9DOF unit available is worthy of a look.

 

1 solution to the VTOL transition problem, VTOL take off can be at an angle, sloped, i.e. 45 degrees. Flight can be as per normal AP. Not sure about the landing......maybe at slow speeds switch to a helicopter AP. This would have the advantages of an AP suitable for slow and hover speeds. This could be achieved with a voting computer, Helicopter AP takes precedence at or below 30 mph......Plane AP takes over 30mph +...........needed is a simple logic computer......taking sensor data from airspeed, maybe attitude also? need to think on this one. Quarternions are the way to go for autonomy BUT having a voting computer also enables redundancy in the AP systems. ed

• ▶ Reply

Permalink Reply by Julian on September 4, 2011 at 7:24am

Nice job on the duct work!

 

The voting computer is what i was figuring, also have to interface another computer system that would be doing the optical processing as well. Im still focusing on the airframe design, but school starts tomorrow so im going to start looking around for a team of EEs who can help me with the simulation/programming.

• ▶ Reply