The request was made to write controlling software for a quadcopter drone. Since this code was not written to control any specific hardware, it is not fully a functional system, but more of a demonstration of how to approach this task. Further down, this README describes how a real controller could be implemented.
A Quadcopter employs 4 motors with propellers, 2 spinning clockwise (CW) and 2 spinning counter-clockwise (CCW). Rotor pairs spinning in the same direction are at diagonals from each other. Each motor has an Electronic Speed Control (ESC) pin. Applying voltage to the pin changes the speed.
Front
O CCW
|
CW O--+--O CW
|
O CCW
Back
Imagine three lines running through an airplane and intersecting at right angles at the airplane’s center of gravity.
- Rotation around the front-to-back axis is called roll.
- Rotation around the side-to-side axis is called pitch.
- Rotation around the vertical axis is called yaw.
- 4 motors
- 1 Gyroscope (3 vectors x,y,z that provide velocity in each direction)
- 1 Orientation sensor (provides pitch and roll)
- Each motor has a power indicator from 0 to 100
- Motor status (on, off)
- drone status (off, hovering, moving)
- Pitch is aligned to the x axis
- roll is aligned to the y axis
| function | Performed By |
|---|---|
| take_off | increase speed of all rotors while maintaining while staying level |
| rotate_left | increase speed of the two CW turning engines |
| rotate_right | increase speed of the two CCW turning engines |
| move_left | increase speed of two CW turning motors |
| move_right | increase speed of two CCW turning motors |
| move_back | increase speed of front motor, decrease speed of back motor |
| move_forward | increase speed of back motor and lower speed of front motor |
| stabilize | set all motors to same speed |
| status | return current status off, hovering, moving, takeoff |
| land | stabilize and slowly lower to ground |
| tap | simulates a hand tap. The drone attempts to restabilize |
| break_engine | simulates a failed engine. Error prints on server and drone langs |
Dependencies: ruby >= 2.5
Open first terminal window and enter the following commands to start the server
1. git clone [email protected]:dsadaka/drone.git # Clone this repo to a new directory
2. cd drone
3. gem install drone-0.1.0.gem
4. drone server # Start server (simulates drone)
Open second terminal window to run the console
1. cd /path/to/directorycreatedabove
2. drone console
You should see "Ready to receive drone commands. Type "exit" to quit."
Enter connect to connect to server
You should see Ready to fly!
Now enter any of the functions listed above, starting with "take_off"
Successful command will return a true status (in blue)
Errors return a red message. Ex: Trying to take_off when you're already flying
The current state of the drone is also displayed.
You can also keep an eye on the server terminal window for log output and distress signals.
The controller cannot blindly send commands to the drone. In the real world, external forces act upon the craft. As opposed to performing complicated calculations connecting power, weight, blade spin rates, gravity, etc, a feedback loop is implemented that compares the requested command (target) with the input from the sensors. This is called the PID process (Proportional, Integral, Differential). It rapidly re-estimates the current best guess output.
The PID class is fed “target” value and an “input” value. The difference between these is the “error”. The PID processes this “error” and produces an “output” which aims to shrink the difference between the “target” and “input” to zero. It does this repeatedly, constantly updating the “output”.
The “P” of PID stands for proportional – each time the PID is called its “output” is just some factor times the “error” – in a quadcopter context, this corrects immediate problems and is the direct approach to keeping the absolute “error” to zero.
The “I ” of PID stands for integral – each time the PID is called the “error” is added to a grand total of errors to produce an output with the intent that over time, the total “error” remains at zero – in a quadcopter context, this aims to produce long term stability by dealing with problems like imbalance in the physical frame, motor and blade power plus wind.
The “D” of PID stands for differential – each time the PID is called the difference in error since last time is used to generate the output – if the “error” is worse than last time, the PID “D” output is higher. This aims to produce a predictive approach to error correction. The results of all three are added together to give an overall output and then, depending on the purpose of the PID, applied to each of the motors appropriately.
The input referred to above comes from two sensors: gyroscope and accelerometer.
A gyroscope rotor spins and maintains it's position within a "cage", as the vehicle moves (rolls, pitches, yaws) around the cage, the gyro "senses" this movement. It can then output the rate of rotation, degree of tilt, and angular velocity of the drone. Our simplifed gyroscope simply outputs 3 velocity vectors (x, y and z). These vectors go to zero when the drone stops rotating. Gyroscopes, however, cannot detect linear movement.
Most modern day drones use an inertial measurement unit (IMU) to provide orientation. It ensures that the drone maintains its orientation towards the ground by itself and does not drift. This device combines a gyro, accelerometer and magnetometer (and/or GPS) to determine orientation and velocity vector relative to the earth. Each of these components have the their strengths and weaknesses but the IMU is smart enough to use their respective strengths to overcome these weaknesses.
https://howthingsfly.si.edu/flight-dynamics/roll-pitch-and-yaw
https://diydrones.com/profiles/blogs/faq-whats-the-difference
https://github.com/blacktm/tello
https://www.raspberrypi.org/magpi-issues/MagPi19.pdf
https://github.com/PiStuffing/Quadcopter
Also purchased a Ryze Tello to get a hands-on feel. I totally recommend this model if you're a beginner like I was. It's under $200 including the separate controller but it includes DJI electronics for a smooth, easy ride.