Drone flight controller and autopilot for Raspberry Pi. It does not require any additional board, the Raspberry Pi pinout is connected directly to ESCs and sensors. It has been developed from the scratch, the code for stabilisation and reaching waypoints is original.
Why to do it?
This project began as a professional assignment. The aim was to develop a flight controller and autopilot where it is easy to connect various hardware components such as gyroscopes, GPS and distance sensors and perform series of tests. Another goal was potentially unlimited precision (sub centimeter) of flight.
Unfortunately, the company that commissioned the software cancelled the project before it had reached its first milestone. By that time the most of the autopilot code had been written. I published the code and continue working on it on my own. Here is a video of one of the first succesfull flight
Why raspberry pi?
Raspberry Pi fits the purpose of the project. It is powerful enough to support sensors that require a Linux operating system such as Intel Intellisense devices. The computer is light enough to be mounted on small drones. Raspbery Pi natively supports double precision floating-point arithmetic, which we use for all our calculations.
Of course, there are obvious drawbacks to using Raspberry Pi. After connecting the battery, you have to wait until Linux boots. You have to shut down Linux before you disconnect the battery. And, of course, Linux is not a real-time operating system, and precise timing can be a problem.
However, none of these points outweigh the advantages we have gained. Raspberry Pi offers plenty of processing power and a standard development and debugging environment. Connected via wifi we can edit, compile, run and debug the code directly on the drone. No need to flash the firmware every time we change the code. We can fly with new versions of our software again and again without physically touching the drone.
What does raspilot provide?
Raspilot is a small software at the moment. It implements the necessary to allow a drone to fly and carry out its mission. The autopilot provides basic stabilisation, it reads sensors and sends commands to the motors. The overall architecture is similar to that of the PX4. Specialised routines that manipulate the sensors are running in separate processes and are connected to the autopilot via Linux pipes. Software runs asynchronously and the input is read as it comes. The autopilot itself runs an infinite loop at an adjustable frequency, usually 100Hz. Software that physically controls motors (or ESCs) runs in a separate process that is connected through Linux pipes as well. It makes it easy to integrate specific ESC protocols or specific motor hardware.
From the user point of view, the whole autopilot behaves like a library. The main program initialises the autopilot and then executes a mission for the drone to follow. The mission is a C function coded by the user. It basically calls autopilot functions like "goto_waypoint(X,Y,Z,Yaw)", which means that the drone will go to the point X,Y,Z and approach it with the yaw orientation. When the waypoint is reached, the function returns and the mission can continue. Such an architecture gives the user (who must be a C programmer) full control over all flight variables and configurations to perform any sophisticated computations he may wish during the flight.
The physics that Raspilot uses to stabilise drones is very rudimentary. Only inertia rules are taken into account. Changes in rotation and speed of movement are controlled by PID controllers. This model is able to stabilise the drone very well at 100Hz (main loop frequency). We were able to fly drones with frequencies between 20Hz and 400Hz. At lower frequencies the drone is unstable, higher frequencies did not make sense for us as we do not have sensors with such update rate.
Hardware
For the moment Raspilot supports T265 Intel intellisense positioning and orientation sensor; MPU-6050 family of gyroscopes; HC-SR04 distance sensor; any NMEA GPS sensor and HMC5883 compass.