Kalman filter

A Kalman filter is a method for estimating something you can’t measure perfectly, like a drone’s position, by combining what you expect with what your sensors say. Each step it predicts where things should be from the last estimate and a model of how they move, then takes a noisy measurement and blends the two, trusting whichever is less uncertain more. It also keeps track of how uncertain its own estimate is, so that trust adjusts automatically over time. That’s why it’s everywhere in navigation: it fuses a GPS that is accurate but slow and jumpy with an accelerometer that is smooth and fast but drifts, getting a better answer than either alone.