Short story

My goal was never to create a quadcopter, I always knew it is very difficult task, especially, because of the fact that I like reinventing the wheel and do everything from scratch. At the beginning, I just wanted to make BLDC motor driver, it worked, then, I attached a propeller and an accelerometer and tested if I was able to write a speed controller for a motor attached at the end of a long arm. I was. The arm was reaching desired angles very fast. Then I made a double driver and a simple quad frame, attached them to it, it worked, the controller was able to keep the frame at "fixed" position. Then I made a quad driver, a remote controller, tons of PCBs, thousands lines of code, and it is, my own quadcopter.

Flight controller

Flight controller is the heart of the whole system. Its main task is to take care of processing accelerometer, gyroscope and magnetometer sensors reads, computer current rotation quaternion, extract Roll-Pitch-Yaw information from it, compute motor speeds (PWM duty) using Roll-Pitch PI controller and Yaw P controller and send this data to the motors driver board. Other module tasks:

• Supervise battery cells voltages
• Process GPS data
• Transmit and receive data to/from remote controller and PC interface
• Process barometer reads
• Manage critical situations like connection lost with remote controller

Motors driver

Instead of buying several regulators like this, I decided to make my own, which is described on separate page -> BLDC driver Main task of the driver is to drive four [alt="Brushless DC"]BLDC[/alt] motors with [alt="Pulse Width Modulation"]PWM[/alt] duty factors obtained from flight controller. Moreover, the module manages motor states. That is, when one motor stops due to mechanical reasons (e.g. hit something), driver can restart it or stop other motors depending on the settings provided from flight controller. The module is also equipped with current Hall-based sensor ACS712 which can measure current up to 25A in both directions.

Remote controller

Whole flight control is done by proper rotating of one part of the remote controller. It is equipped with the same set of sensors and algorithms as main flight controller board. Thanks to this, quadcopter can follow remote controller's movements. Both systems know absolute and relative orientation of each other. It allows to easy controlling without taking care of quadcopter's frame orientation. For example, if I turn remote controller right, the quadcopter will always fly right, even, if I turn myself around.
Second part of the remote controller controls throttle. PS2 joysticks have very poor resolution, they start changing voltage very late (after about 1-2mm) and get saturated very early. My solution is to take inclination angle of the knob instead of its absolute position and restrict mechanical movements to prevent results of being saturated. It is simple arctan(y/x) in fact, but allowed me to still use those joysticks.

PC interface

Interface between remote controller, quadcopter and software on PC. KDcopter software on PC provides a lot information about current state of the quadcopter like cells voltages, sensor reads, quadro and remote controller positions and lot of other things.

Radio communication

Radio communication is done by RFM69HW modules. My own protocol includes simple collision detection algorithm and some kind of time division algorithm. Every devices sends its data every 20ms. After transmission, the module is switched to Receive Mode and starts to listening for a packet. If there is some packet being transmitted in the air, module receives it and postpones transfer until packet is received. There is also some random time added after each reception to allow recovery in case of collision and to allow other transmiter to switch to Receive Mode.

Control panel

-

GPS

Quadcopter, as well as remote controller are equipped with GPS receivers. They are not currently used to control position, but they work and provide real location.

Resources

• [1] Open source IMU and AHRS algorithms by x-io Technologies - Excellent IMU algorithms.

Gallery

See full gallery on Flickr