50.3.2 Choice of model for the KF using IMU readings

Search IconIcon to open search

Parent: IMU index ,  50.3-modelling-imu-in-kf

According to  MKok 2017 , here are some models that assume either a constant acceleration or a constant angular velocity:

  • Constant acceleration model unknown_filename.png

  • Constant angular velocity model unknown_filename.1.png (Notation: angular velocity of the body with respect to world (n), expressed in body CS)

If motion is unknown, there is also the option of modelling the states using random walk equations.

A common model in the literature for EKF sensor fusion using IMUs use the IMU readings as inputs

  • This results in a partial state vector for the pose, s.  50.3.1-states-for-imu-motion-model
  • Variables such as biases can be appended to the states to be estimated and are modelled as a random walk
  • If gravity is to be estimated, it also appears as a state in the estimator

A such model is defined below in  Solà 2017

Image.png

Note: [The initial gravity vector/orientation for the IMU ESKF](the initial gravity-vector_orientation-for-the-imu-eskf.md)