Drone

A drone is an aircraft that can fly without an onboard human pilot. In robotics, “drone” usually means a multirotor UAV, especially a quadcopter.

A quadcopter flies by changing the speed of its four propellers. The propellers create thrust, and differences in thrust create roll, pitch, yaw, and vertical motion.

Main components §

• Frame
• Propellers
• Brushless DC motors
• Electronic Speed Controllers (ESCs)
• Battery
• Power distribution board
• Flight Controller
• Sensors: IMU, barometer, magnetometer, GPS, camera, lidar, or optical flow
• Radio receiver or onboard computer

The frame holds everything together, the battery provides power, the ESCs drive the motors, and the flight controller runs the control loop that keeps the drone stable.

Motors §

A quadcopter has two clockwise motors and two counterclockwise motors.

Diagonal pairs spin in the same direction. This makes the torque generated by the spinning propellers cancel out during hover, enabling stable flight.

Typical X-frame motor layout:

Front

      M1 CCW     M2 CW
          \     /
           \   /
           /   \
          /     \
      M4 CW      M3 CCW

Each motor produces upward thrust:

where:

• is thrust from motor
• is a thrust coefficient
• is motor angular velocity

Motor torque also scales roughly with motor speed squared:

This is why small motor speed changes can create large control effects.

Control axes §

A drone controls motion by changing relative motor speeds.

Motion	How it happens
Throttle	Increase or decrease all motors together
Roll	Increase thrust on one side and decrease thrust on the other side
Pitch	Increase thrust in the back/front and decrease thrust in the opposite side
Yaw	Speed up one diagonal pair and slow down the other diagonal pair

The drone does not directly “move forward.” It pitches forward, tilts its thrust vector, and part of the thrust pushes it forward.

Force intuition §

In hover, total thrust equals weight:

If total thrust is greater than weight, the drone accelerates upward.

If the drone tilts, the thrust vector is no longer perfectly vertical:

• vertical component keeps it in the air
• horizontal component accelerates it sideways or forward

This is why aggressive forward flight usually requires more throttle than stationary hover.

Flight controller §

The Flight Controller is the real-time computer that stabilizes the drone.

Basic loop:

1. Read sensors
2. Estimate current state
3. Compare desired attitude/position to current attitude/position
4. Use a controller such as PID to compute corrections
5. Convert corrections into motor commands
6. Send commands to ESCs
7. Repeat hundreds or thousands of times per second

The flight controller usually runs nested loops:

• rate control: controls angular velocity
• attitude control: controls roll, pitch, yaw orientation
• velocity control: controls translational velocity
• position control: controls position in space

Lower-level loops run faster because stability depends on quick attitude correction.

State estimation §

The drone needs to estimate its state:

• position
• velocity
• orientation
• angular velocity
• sensor biases

Common sensors:

Sensor	Measures	Problem
Gyroscope	Angular velocity	Drifts over time
Accelerometer	Acceleration and gravity direction	Noisy during motion
Magnetometer	Compass heading	Disturbed by metal and current
Barometer	Altitude from air pressure	Noisy and weather dependent
GPS	Global position	Low update rate and poor indoors
Camera / optical flow	Visual motion	Needs texture and light

Sensor fusion combines these imperfect measurements into a better estimate. A common method is the Extended Kalman Filter.

Attitude representation §

Drone orientation can be represented with:

• Euler angles: roll, pitch, yaw
• rotation matrices using Transformation Matrices
• quaternions

Euler angles are intuitive, but quaternions are often used internally because they avoid gimbal lock and are numerically stable.

Motor mixing §

Controllers usually output desired total thrust and torques:

The mixer converts this into four motor commands:

This depends on frame geometry, motor direction, propeller direction, and arm length.

Software stack §

Typical drone software layers:

Layer	Example
Low-level firmware	PX4, ArduPilot, Betaflight
Middleware	ROS 2
State estimation	Extended Kalman Filter
Control	PID, MPC, geometric control
Planning	trajectory generation, obstacle avoidance
Perception	camera, lidar, optical flow, VIO

For learning/research, the stack often looks like:

Sensors -> State Estimation -> Controller -> Motor Mixer -> ESCs -> Motors

Important problems §

• Stability: the drone must constantly correct attitude
• Latency: slow sensors or slow control loops cause oscillation
• Vibration: motor vibration corrupts IMU measurements
• Battery sag: available thrust changes as voltage drops
• Weight: every gram reduces flight time
• Prop wash: airflow from propellers affects nearby surfaces and sensors
• Localization: GPS is weak indoors, so drones often need visual-inertial odometry

Related §

• Flight Controller
• Extended Kalman Filter
• PID
• Quaternion
• Robot Operating System
• PufferLib Drone
• BetaFPV Pavo20 Pocket Quadcopter