Suppose the physical sensor flies on a quadrilateral. If you stand at an angle, Pythagoras said that the ground looks farther than it really is. It was tilted enough that we couldn't see it at all, even though we were only a few centimeters away from the crash.

In order to correct this, we need to know our tendency. That means ...

Gyroscope and accelerometer are called "IMU" together, but the reason why they often appear together is that they naturally cover up each other's main weaknesses. They are a classic fusion pair.

However, individually, the gyroscope is a good and clean source of sensor data, and you will get continuous position measurement.

Without too much detail, the gyroscope is made of micromechanical tuning fork etched on the surface of the chip. When the chip rotates, some tuning forks will be powerful, and they will change the tone of each other.

MEMS gyroscopes are hardly affected by any movement except rotation. Even violent vibration will not affect them (within the specification range), because this is lateral acceleration.

If you look at the internal control loop of a multi-rotor, it is the gyroscope used by the flight controller to keep level in the air. Some rate mode pilots turn off the accelerometer completely, giving the gyro more bandwidth because the gyro has many important clues.

If you know where you were half a millisecond ago and you want to know what has changed since then, then the gyroscope is fast, accurate and reliable. And they don't need flexible surfaces or satellites.

The weakness of gyro is drift. No matter how hard you try, they seem to rotate slowly around a random axis. Each rotation takes several minutes, but even a motionless brick will rotate gently. Accumulating relative samples to obtain "excellent" estimates will also add up the error bars. The error of the ascending gyroscope needs to be "zeroed" periodically by an external excellent reference mark.