Initial Quadrotor Design
I have been researching a variety of multi-rotor helicopter setups for some time. I’ve been trying to identify the strengths and weaknesses of each design type. I have come up with an initial design for my quad-rotor helicopter. In attempts to create a very aggressive aircraft, I have designed the propellers to be a close as possible, as large as possible, and the motors to have a high power to weight ratio. All of my assumptions about quadcopter design are very spectulative since I haven’t ever built one before.
My first matter of design is a powerful CPU. My fly-by-wire T-Rex 600 project used 4 Arduinos and they left a sour taste in my mouth. Arduinos are great for quick prototyping but anything with substance needs a better processor. Besides, I’m a Computer Engineer so I can’t justify using someone else’s poorly designed microcontroller libraries. My microcontroller of choice for this project is the LPC1768 ARM Cortex-M3 by NXP Semiconductors. It is a powerhouse! I’ve written most of the low-level hardware drivers and a few of the higher level routines, such as PID controllers and Collective Cyclic Throttle Mixing (CCTM). The combination of the ARM’s Cortex-M3 core with NXP’s hardware peripherials makes this as amazingly powerful design. ARM+NXP=Happiness!
For aircraft attitude measurement, I’m planning on using the 9DOF Sensor Stick from Sparkfun. Version 2 is still under design so I’ll have to wait on that. I’m going to implement 3 different types of sensor fusion algorithms and see which type works the best. The 3 algorithms are:
- Complimentary Filter
- Direction Cosine Matrix
- Extended Kalman Filter
In the animation above, the two circles representing the propellers show the two sizes I will test. The inner circles are a 12×3.8″ APCprop and the outer circles are a 14×4.7″ APCprop. I haven’t seen another helicopter use the 14″ props before so I’ll get the 12″ props working first.
This is an explanation of the animation per level:
Level One (bottom):
- LiPo battery
Level Two (between the two metal plates):
- 4x Electronic Speed Controllers
Level Three (top):
- 3-Axis of Gyroscopes, Accelerometers, and Magnemeters
- 4x Motors
- 4x Props
I know that having the props close together makes the attitude harder to stabilize. On the other hand, having the props close together will (I hope) induce much more torque on the frame from the motors. This will help me overcome the lame yaw response of most quadrotor helicopters. I’m banking on the fact that my CPU will be running at 100 MHz and I’ll hopefully have the sensor fusion filters and PID controller running at 400+Hz. This should allow me to precisely adjust each axis of stabilization. For better discrete calculus computations (integration and differentation) I’ll try running the sensor fusion algorithms above 1kHz and only commanding the ESCs at their maximum speed (50Hz-400Hz). This will make the computations more accurate because each time step will produce less error.