BXFlyer Four Rotor Helicopter
by
Bruce J. Weimer, M.D.
THE ROBOT:
This is the finished BXFlyer with the control board attached to the
airframe. The aluminum arms measure 23 inches in length from motor to
motor. And the rotors measure 11.5 inches from tip to tip. The 9.6V
battery pack (not shown) is then velcroed underneath the control board for
flight.
BACKGROUND:
I decided to build an autonomous flying robot about three years ago. I've
converted RC cars to autonomous operation, so I thought that an RC
helicopter might be a good place to start. I quickly discovered that the
standard helicopter design would probably be too unstable in flight for my
simple navigational and programming skills - but in searching the web I
stumbled across the four rotor electric helicopter platform...and I was
hooked!
The first four rotor RC heli seems to have been the Japanese "Engager":
As far as I can tell, the Engager was never marketed here in the US -
although you can sometimes see it on Ebay. However an upgraded design is
available here in the US as the "AFOT Roswell" - and through a Canadian
source as the "DraganFlyer":
I, of course, bought one. As it happens, I came across the DraganFlyer
first, so that's the one I bought.
There are others - Stanford, NASA and JPL together are developing a four
rotor "Mesicopter" for Mars exploration, but I couldn't find a price quote ;
- )
THEORY OF FLIGHT:
The helicopter has four electric motors driving four rotors - all pointing
"up." Looking down on the helicopter from above, the motors / rotors can be
thought of as being located on the points of a compass - North, East, South
and West. The North and South motors spin clockwise and the East and West
motors spin counterclockwise - I'll come back to this in a minute. Now if
you rev up all four motors / rotors equally, then the helicopter will lift
off going straight up. And if you then slow down the North motor while
simultaneously speeding up the South motor, then the helicopter will tilt
towards the North and it'll start traveling in that direction - that allows
you to control "pitch". You control "roll" using the East and West motors.
And remember the clockwise / counterclockwise thing? Well, as long as all
four motors / rotors are all going at the same speed, then the North / South
pair cancels the East / West pair's angular momentum. But by speeding up
the North / South pair and slowing down the East / West pair (or
vise-versa), you can make the heli rotate to control "yaw" motion! The
really cool part is that by always adding to one motor the same amount that
you're taking away from another motor, you control the heli without changing
thrust!! (Okay, in an ideal world, ignoring the effects of friction,
etc.....but close enough.)
There are at least two other considerations.
The four rotor helicopter has four rapidly spinning rotors. And to some
extent they should act as four independent "gyroscopes". Therefore, at
least in theory, there should be an intrinsic "gyroscopic stability" to the
helicopter. In addition, since the battery pack's attached below the
helicopter, the center of gravity / mass should be below the helicopter
creating a "pendulum effect" that should also tend to stabilize the
helicopter in flight.
But that's all just theory...I later had to keep reminding myself - crash
after crash - that I HAD seen the DraganFlyer flying in videos on the
internet!
AIRFRAME:
The DraganFlyer's actually a remarkably well-designed and really well-built
RC helicopter. Their initial design used a light-weight "sandwiched"
plastic airframe - they're now using carbon fiber rods. However, I thought
that I could make it lighter and stronger - so I rebuilt the airframe using
two hollow aluminum archery arrow shafts joined in the centers at 90 degrees
to each other to make a "+" shape. And I then reused the DraganFlyer motors
/ motor mounts and rotors gluing them to the ends of the shafts with JB Weld
epoxy. This ended up being lighter than the original DraganFlyer - and
lighter even than carbon fiber rods. I also dispensed with the RC receiver
- this thing's supposed to be autonomous - so I saved some additional weight
there. And less weight means more battery / flight time! (As you'll see, I
needn't have worried.)
This is a close-up of one of the motor mounts. I used the original
DraganFlyer motors and motor mounts and attached them to the aluminum shafts
using JB Weld epoxy. I found the little wooden "spools" at a crafts shop.
ELECTRONICS:
I used a BX24 microcontroller because I'm familiar with it, it's
programmable in Basic, it's (relatively) fast, it features floating point
math and it has multi-tasking. I also used a serial connection from the
BX24 to a PAK pwm co-processor. I then used the PAK to drive 4 MOSFETs.
And, as I've already said, I reused the DraganFlyer motors and rotors. All
this powered off the DraganFlyer's 9.6V rechargeable battery pack. Then the
problems started. The motors generate a LOT of EMI - enough that they kept
resetting the BX24 and the PAK. I could run one motor at a time - no
problem. Two sometimes. But all four at once - forget about it! It took
me over a year of playing around with various combinations of PCB boards,
flyback diodes and noise suppression capacitors to finally get the thing to
work!
My solution eventually involved putting flyback diodes and capacitors
across the motor terminals and 0.1uF capacitors across the MOSFETs while
running the PAK in "equal" pwm mode.
This is the circuit board that I designed for the BXFlyer - I've removed
the BX24 microcontroller from it's socket so that you can see more of the
traces underneath. The PAK pwm coprocessor is labeled. The Spectron 2-axis
tilt sensor sits in the center of the board. And the Dinsmore compass sits
just "north" of the tilt sensor. The four MOSFETs are on the edges of the
board. And a header connects each MOSFET to it's corresponding motor.
SENSORS:
Tilt sensors:
I wanted this to fly autonomously. And that meant sensors. The RC
DraganFlyer uses ceramic rate gyros to dampen it's responses so that a human
operator can control it. I didn't want that. I wanted it to fly by itself.
And, to me, that meant it had to know to keep itself level. And I reasoned
then that "level" meant a Spectron 2-axis tilt sensor. So that's what I
decided to use. If you haven't used them, these are great little sensors -
electrolytic fluid filled capsules with four outside pins and a fifth pin in
the center. You then use them in a "voltage divide" configuration.
Specifically, I programmed the BX24 to put the "North" pin high, the "South"
pin low and the "East" and "West" pins tristate while taking an ADC reading
off the center pin. That gave me the 'pitch' angle - reproducibly accurate
to a degree or two. And measuring changes over successive readings gave me
the 'angular rate of change' - no need for a rate gyro after all! Then you
just do the same thing using the "East" and "West" pins and you've got the
'roll' info.
Almost.
If you keep pumping current in the same direction through these sensors
you'll eventually electroplate the sensor's electrodes making it unusable.
The solution's to take a reading in the "North" / "South" direction followed
immediately by another reading in the "South" / "North" direction - a
pseudoAC current - that cancels out the electroplating. Of course you have
to do the same thing with the "East" / "West" pins. This means that you're
in effect taking two readings in each direction each time, but you can
average them together to reduce noise if you want to - I didn't.
Now someone's sure to object that if the helicopter goes into a steep bank,
then centrifugal forces are going to create an acceleration that cancels the
pull of gravity and the tilt sensor's going to be giving useless data.
You're right - technically. I'd counter by telling you that if the heli's
in a steep bank, then you don't have to worry about taking measurements -
it's out of control and it'll crash before you can say "I told you so!"
Compass:
So far I haven't mentioned 'yaw' control. My idea was to try to keep the
helicopter pointing North. And to me that meant using a compass. I choose
the Dinsmore 1525 analog compass which was easy enough to use - you just
give it 5V and take readings from it using an ADC pin on the
microcontroller. I was able to reproduce a benchtop accuracy of about +/- 1
degree - so I was hopeful that this would work. And the compass readings
didn't seem to be affected appreciably by the four motors even running at
full speed. In fact, I programmed in PID control for the helicopter's
maiden flight, but because the four spinning rotors create their own
intrinsic "gyroscopic stability", I'm not convinced that the compass idea
actually worked...but I'm not sure that it didn't.
SOFTWARE:
As I mentioned, one of the reasons that I choose the BX24 was because it
gave me floating point math in Basic.
The right way to do this then would have been to code some appropriate
linear algebra flight equations into a control program - since
mathematically there actually is some interaction between pitch, roll, yaw
and thrust:
But as a reasonable and simple approximation, I treated them all as
independent variables and programmed pitch, roll and yaw using simple PID
equations. For as far as I got with the project, this seemed to work well
enough.
TEST FLIGHTS:
I had lots of test flights. At first I "flew" the heli tethered and mounted
to two wooden posts so that it could "roll", but it couldn't actually go
anywhere - that let me test my PID software. Then later I tested it in
short unconstrained flights with the battery pack - flights of only 3 - 5
seconds. My hands were never far from the heli during those "flights." By
the way, I can attest to the fact that the plastic rotors sting when they
hit you - but they don't actually hurt. Nevertheless, it's a good idea to
wear eye protection and keep innocent bystanders away. The major
frustration that I had with the project was that the heli can literally tear
itself apart in the blink of an eye if one of the rotors touches anything
solid - the floor, a wall or a piece of furniture. And then rebuilding it's
time consuming, expensive and ultimately very very frustrating.
MAIDEN FLIGHT:
Anyway, after the better part of two years - and after having solved most of
the major control problems - I was ready for the BXFlyer's Maiden Flight.
Remember, I didn't have an altitude sensor. So I did the best that I could
under the circumstances - I had it fly straight up for 10 seconds,
decelerate for 2 seconds and then turn-off to land / crash (hopefully
gently) on the ground.
And it worked! In fact it survived!! I had matched the Wright brother's
first flight time - I was off to celebrate!!!
SUGGESTIONS:
If I had it all to do over again, I'd start by looking for a sponsor and
funding. No kidding. I easily put in a couple of thousand dollars over the
two or three years that I worked on this project. And I didn't get as far
as I would have liked. I'd estimate that to do the project right would take
a budget of $5000 - $10,000. I'd also make the helicopter BIGGER. The new
DraganFlyer X-Pro is their scaled up "industrial" version with a one pound
payload carrying capacity - enough for all the electronics and sensors you'd
ever need...but it costs $4997.95. Still, in this case, bigger means more
stable - and therefore easier to fly. I'd also consider adding inertial
measurement units (IMUs). And an altitude sensor. And, of course, a
live-feed video camera.
ACKNOWLEDGEMENTS:
This was easily the most difficult robotics project I've ever undertaken.
The shortcomings are my own, but the successes I owe to a number of others
including Steve Vorres, Ed LeBouthillier, Alex Brown, Zenon Dragan, Mike
Dammer, Richard Friedrich, Mark Johnson, Al Williams, R.C. Dinsmore and Jim
Ubersetzig. And, of course, to my wife Ana - who never once asked "So what
useful purpose is this thing going to serve?"
Bruce J. Weimer, M.D.
June 2003
