# Orientation estimation using gyro and accelerometer when sensor platform has high acceleration

I am trying to estimate the orientation of a sensor platform using gyroscope and accelerometer. I am using a Kalman filter based approach. I integrate the readings from gyro to obtain the orientation about x, y and z axis. This is will act as the process model. I am planning to use accelerometer readings as observation. I know how to estimate the angles of orientation when the accelerometer is reading only gravity, by using acceleration components in x, y and z.

However, when the sensor platform is moving or accelerating (i.e. gravity is not the only acceleration the sensor is reading), can accelerometer provide orientation/ tilt estimation without the help of any other sensor? Can this approach work without using a third sensor when the sensor platform has high acceleration?

## 1 Answer

General movement is not a problem even at high velocity but, as you already described, acceleration is. There are a couple mitigating factors that allow many platforms to still (mostly) assume the accelerometer is a gravity measurement

• Short bursts of acceleration can be filtered out (e.g. if $||a|| >> ||a_g||$ ignore the reading)
• High frequency acceleration oscillations average out to zero over time. This just means you have to increase the measurement noise parameter of the filter
• If you happen to know the maneuver acceleration you can simply adjust the accelerometer measurement accordingly. This case is rather unlikely though (there's a reason we need sensors in the first place)

When this is not the case, such as for a missile which has long periods of sustained non-gravity acceleration, you simply need a different sensor.