Well, I’ve taken a little break on the microcontroller stuff for the RV-6 autopilot because the plane needed a little love. It’s been flown 50+ hours since Cedric and I bought it in February, 2010 and there was some maintenance that needed to be done before a trip to Tennessee.

We took the cowling off, changed the oil, re-wired some of the electronic ignition, re-sealed the right fuel tank, cleaned up some of the wiring, replaced the fuel filters, re-plumbed some of the vacuum lines, and a couple other things. We worked every evening from about 17:30 until about 23:00.

In the end we were able to complete all the necessary work in three evenings of work and the plane was ready to go by the 3rd of July.

The goal of this project is to add an autopilot to my RV-6. There are many smaller steps to be completed along the way so I’ll lay out the end goals, then I’ll break things down a bit and then we’ll start from the top.

The RV-6 is a great airplane. Mine will typically cruise at about 180-190 mph on 7.5 gallons of fuel per hour. Do the math, that’s about 25 mpg! I don’t believe that there are any cars that can get you somewhere at 180 mph while only consuming 25 mpg. It’s the Honda Civic of airplanes except that it runs on 100LL.

Anyhow, the RV-6 is a two person side-by-side experimental airplane that comes as a kit sold by Van’s Aircraft. It is one of the most popular kit airplanes ever produced -a tribute to it’s incredible performance. It is rated for all kinds of aerobatic maneuvers and it’s high speed makes it a great cross country airplane as well.

Unfortunately, some of the aspects of the airplane that make it a great aerobatic airplane become drawbacks when flying cross country. Typically maneuverability is achieved with low stability, however stability would be really nice when flying cross country. It’d be great to have a plane that just tracked a course with little no pilot input. Well, that’s where the autopilot comes into play. We will use a feedback look to take what is an unstable system (or at least not very stable system) and make it stable.

Here are the steps necessary to get to where I have an autopilot in the RV.

1. Control servo with microcontroller
2. Control multiple servos with microcontroller
3. Incorporate buttons and control servos with microcontroller and package it all
4. Incorporate servo into elevator to control the trim tab
5. Put entire mechanism into RV-6
6. Verify that the trim tab has sufficient control authority (test flights)
7. Add servo control to ailerons
8. Verify that aileron servo has sufficient control authority (test flights)
9. Add sensor package to airplane that interface through microcontroller
10. Interface microcontroller with Matlab/Simulink
11. Develop control algorithms
12. Verify stability of control algorithms
13. Gain tuning
14. Port control algorithm from Matlab/Simulink to C

I guess it falls under the category of “if you want something done right you’ve got to do it yourself.” Well, I’ve been searching for a great LED flashlight, but I don’t want to be buying batteries all the time, and I don’t want to remove the batteries to charge them. What’s the solution? A lithium ion battery powered, USB chargeable, constant-current, LED flashlight. The above picture shows my fully functional prototype. I epoxied the LED to the front end of an aluminum rod for heat-sinking purposes, and I dead bug’d all the components. The collimator came with an adhesive back, and is designed to fit around the specific LED I used. I can now use this flashlight, and charge it with my iphone, or ipod charger, which I never go anywhere without. Any USB port will also work for charging. The only problem is the packaging, but if I get around to it I’m going to gut a cheapo harbor freight flashlight and put the guts from a second prototype inside.

I’ve even made my first board. PCB with photoresist already spun onto it is available somewhere on the internet, but my roommate just happened to have some and the chemicals for etching it. Anyhow, I took a stab at drawing it all up in Eagle, then I printed the traces and pads onto transparency film and taped that over the PCB and exposed it under UV light. From there, it’s just like old fashion black and white photography. Dip it in this, then that, then rinse and voila, you’ve got a PCB. I’m obviously not making the best use of the space on the PCB, but it was a first attempt and it actually worked. This picture of it is after I moved, and the inductor for the buck/boost is missing. 


BOM:

• Luxeon LED
• Maxim Li-ion IC: MAX1555EZK+T
• National Semiconductor LED drive IC: LM3410XMF/NOPB
• Various resistors, capacitors, and inductors
• Mini-USB plug
• Li-ion batteryCollimator lens

I recommend checking out www.maxim-ic.com (not maxim.com as I learned the hard way) before you start your project because they have free samples of various IC’s available.

I've completed my thesis. You can find it below and download it.

We conducted another flight test over the weekend and it did not go well, but we did learn a lot from it.

The airplane was stable when under power, as it had been in previous flights. It even flew relatively well in a very stalled scenario. (It is not necessarily visible from the video, but it was very stalled throughout even the beginning of the flight.) Our problems resulted from pilot error/misinterpretation as well as from a poor configuration of the airplane prior to the flight.

Prior to take off we noticed that the rear wing of the airplane was slightly pitched down with respect to the fuselage and that the canard was slightly pitched up when the joystick was neutral. We decided that it was not extremely important for both canard and main wing to be parallel with the fuselage for flight, which we still believe is true, (I will discuss this separately) but we did not comprehend the full impact and consequences of this configuration until too late.

We had not realized that this configuration would mean that "normal" fuselage attitude during climb would actually put the canard and main wing at high angles of attack. As a result of this we took off in a stall and maintained a stall throughout without comprehending what had happened and why the airplane wasn't performing as it had in our previous flights. With the fuselage in an appropriate climb attitude it was not immediately apparent what the problem was.

Flight Analysis
The airplane took off well from a hand launch in zero wind conditions. It began to climb, but as it climbed it became apparent that it was mushing through the air rather than establishing a proper climb. The pilot then controlled down slightly, but it was not enough and the airplane continued to mush through the air. As the airplane came to a complete stall it slid right and entered a sort of flat spin. It recovered from this and began to gather speed ending the stall, but at this point we decided to autorotate in an effort to gather what we could from a failing test.


Unfortunately, prior to the flight we had decided on an autorotation sequence that was improper and could never have worked given our greater understanding of the problem at hand. This sequence was: 

1. Kill engine
2. Change remote control to high rate setting (this allows the pilot to rotate the wings over large angles for autorotation rather than the low angle limits used during conventional flight).
3. Begin wing rotation (~80 degrees)
4. Approach ground and increase angle to ~100 degrees so as to generate sufficient lift to arrest descent.

The problem with this scenario is that as soon as we killed the power we lost yaw stability and the airplane went into a bit of a flat spin. As you may recall the airplane has no vertical stabilizer. It is therefore very unstable in yaw, and cannot be maintained in coordinated, controlled flight without thrust vectoring. The ability to use thrust vectoring is lost if the motor is turned off. Thus our flight plan was doomed to fail because we had planned to cut power before entering the autorotation. The airplane was already out of control before the autorotation sequence was initiated control was never regained.

The damage to the aircraft was not as serious as it may look on video. The part of the fuselage made of aluminum bent and I will need to machine a new one. Other than that, the airplane is in good condition for future flight tests and we will be conducting these as soon as possible.

We conducted one flight test today and the results are pleasing and will certainly lead to more testing of the current airframe. The most incredible part of the flight test was that the airplane actually flew in coordinate, controlled, straight, and climbing flight. Due to various design constraints, the airplane has no vertical stabilizer and it is therefore unstable about the yaw axis. A gyroscope, and gyroscope-controlled thrust vectoring were incorporated into the design making coordinated flight possible. That said, it would be impossible for a human to control the airplane without the gyroscope. Additionally it is interesting to note, although not entirely surprising that the airplane was controllable in roll. As previously explained, there are no ailerons on the airplane. Instead, the whole main wings rotate in opposite directions to achieve roll control. I know of no other airplane that controls roll in this manner and we were all pleased to see the airplane fly as well as it did.

The rotoplane was hand launched into the wind from a moving vehicle. It climbed in coordinated, controlled flight to an altitude high enough for an autorotation. At this point the motor was cut and the airplane was set into a roll. This roll merged into a situation where the airplane was spinning about its longitudinal axis, and parallel to the gravity vector as designed.

The pilot then changed the pitch of the wings to a “positive” angle of attack. It is unclear why, but the airplane then reversed its autorotation such that the what was supposed to be the top part was now the bottom. It continued in this nose upward attitude for a brief period without rotating significantly until the controls were adjusted to accomodate an autorotation in this orientation. At this time the airplane began to rotate about an axis through the center of its fuselage although pointed upward.

Without enough altitude left to rectify the situation, the test pilot let the airplane spin up and prepared for the flare at the end of the autorotations. At an altitude of about ten feet he flared. By this point, the airplane was below the tree-line and isn’t as clearly visible as I would have liked. However, it did arrest its descent as is noted by the brief pause only a couple feet off the ground.

This test was a great success for the rotoplane even though it wasn’t a perfect demonstration of the technology being developed. It did demonstrated that an airplane of the rotoplane’s configuration could fly, and that not only could it fly; it can autorotate. Future tests will be conducted with the same airframe in a couple weeks once I have time again. 

Future work includes gathering data from the autorotation such as rotational rates, wing angle and descent rate. In order to accomplish this I would like to install electronic sensors. I am in the process of learning to program microcontrollers. I eventually hope to install some custom made data loggers in the airplane to gather data during autorotations.

I would like to thank William Grossman (UC Berkeley) who flew the airplane for all test flights from the very beginning of the project. I would also like to thank Cedric Jeanty (Caltech) for all the useful ideas and advice during the design and construction of the airplane. I would also like to thank professor Culick for his help and guidance along the way and Caltech for having a senior thesis program. Finally I would like to thank my cousin Beth for filming.

On February 2nd, 2008 we did high speed taxi and preliminary flight tests for the rotoplane. The results are positive, but the airplane design leaves some to be desired. The most important result of the test is that the airplane has enough yaw stability to maintain straight and level flight due to the gyroscope-controlled thrust vectoring. Many high speed taxis without take offs were conducted and then we graduated to low altitude flights (1-3 feet above ground). 

All flights unfolded in approximately the same manner. The takeoff runs were smooth and transition into flight was easily accomplished with pitch control from the canard. We were able to maintain level pitch attitude and the wings parallel to the runway as well as maintain heading. The problems were always with the landings. The airplane can maintain heading when it’s motor is powered because it can vector its thrust. Upon reducing the thrust to land, the ability to maintain heading is lost and one wing slips ahead of the other. During this slip one wing generates more lift  than the other sending the plane into an awkward sort of cartwheel/roll landing as you’ll see in the videos. Fortunately, the airplane isn’t designed to land in a conventional manner.

Because the landing gear was an afterthought, designed solely for high speed taxis and initial testing, it interferes with large rotations of the main wings and thus we cannot attempt an autorotation landing from a ground roll take off. We would have liked to fly again and take off without landing gear so that we can attempt an autorotation, but it began to rain and we will have to wait for better weather.

I believe, based on my experience with the mockup as can be see in the droptest video, that we could accomplish an autorotation. Although the current airplane is heavier than that used in the droptest, the wings are much stronger and hopefully will not break as one did in the droptest. In fact, autorotation simulations in Matlab and wing loading tests confirm that the wings are strong enough and we look forward to a chance to autorotate.

See video below.

The Rotoplane is assembled and ready for a test flight. Below you'll find an in-shop demonstration of the aerial platform.

The very first step in the development of the rotoplane was to build a mockup of the plane with limited functionality. I decided to determine whether an airplane with with wings that could rotate about their quarter chord could be made to auto-rotate. This was the basic concept of the rotoplane and if this didn’t work I thought I might as well find out sooner rather than later.

I built a set of wings out of light foam and balsa wood. Then I modified some servos so that I could attach the wings to a mockup airplane fuselage. I would have preferred to strap the wings directly to the servos, but the bearings in the servos cannot take the radial loads that would be applied in an autorotation. I needed to connect the wings to something structural yet be able to control the pitch of the wings. I opted for a solution that took advantage of how servos are designed.

Servos are essentially a geared-down brushed DC motor with a gear train. The output of the gear train is the output shaft of the servo and is connected to a potentiometer. The potentiometer is setup as a voltage divider, which feeds back into the servo control circuitry. Based on the voltage of the potentiometer, the servo position is known to the control circuitry. The control circuitry works to keep the servo at a given position for a given input.

I disconnected the internal potentiometer and re-wired the connections to an external potentiometer to which I connected the wings. Then I coupled the potentiometer and wings to the servo with an additional set of gears. In this manner I was able to de-coupled the loads on the wings from the servos allowing me to use relatively lightweight servos.

Using modified hobby electronics I was able to control the “pitch” of the wings much like the pitch of a propeller is controlled on an aircraft, or the blades of a helicopter are controlled in what is called collective. I connected a battery to the setup and an remote control receiver.

It was time to find something to drop the mockup rotoplane from to see if it could autorotate.

I live in Pasadena, known for the elegant Colorado Street Bridge. I climbed up underneath it and dropped the mockup while my brother controlled the pitch of the wings via remote control.

You will notice that in the beginning of the video the wings are facing more downward than horizontal. As the mockup dropped the rotational velocity increased and the pitch of the wings was gradually increased until they passed from being driven by the air they were falling through to pushing enough air down in order to rise up and arrest the descent of the craft.

The very first step in the development of the rotoplane was to build a mockup of the plane with limited functionality. I decided to determine whether an airplane with with wings that could rotate about their quarter chord could be made to auto-rotate. This was the basic concept of the rotoplane and if this didn’t work I thought I might as well find out sooner rather than later.

I built a set of wings out of light foam and balsa wood. Then I modified some servos so that I could attach the wings to a mockup airplane fuselage. I would have preferred to strap the wings directly to the servos, but the bearings in the servos cannot take the radial loads that would be applied in an autorotation. I needed to connect the wings to something structural yet be able to control the pitch of the wings. I opted for a solution that took advantage of how servos are designed.

Servos are essentially a geared-down brushed DC motor with a gear train. The output of the gear train is the output shaft of the servo and is connected to a potentiometer. The potentiometer is setup as a voltage divider, which feeds back into the servo control circuitry. Based on the voltage of the potentiometer, the servo position is known to the control circuitry. The control circuitry works to keep the servo at a given position for a given input.

I disconnected the internal potentiometer and re-wired the connections to an external potentiometer to which I connected the wings. Then I coupled the potentiometer and wings to the servo with an additional set of gears. In this manner I was able to de-coupled the loads on the wings from the servos allowing me to use relatively lightweight servos.

Using modified hobby electronics I was able to control the “pitch” of the wings much like the pitch of a propeller is controlled on an aircraft, or the blades of a helicopter are controlled in what is called collective. I connected a battery to the setup and an remote control receiver.

It was time to find something to drop the mockup rotoplane from to see if it could autorotate.

I live in Pasadena, known for the elegant Colorado Street Bridge. I climbed up underneath it and dropped the mockup while my brother controlled the pitch of the wings via remote control.

You will notice that in the beginning of the video the wings are facing more downward than horizontal. As the mockup dropped the rotational velocity increased and the pitch of the wings was gradually increased until they passed from being driven by the air they were falling through to pushing enough air down in order to rise up and arrest the descent of the craft.