Simplified 1-D Kalman Filter

Variable Naming Convention

$x_t$​: state

$z_t$: measurement

$u_t$​: control action

Mean Calculation


$\mu$: Mean of the prior belief

$\sigma^{2}$: Variance of the prior belief

$\nu$: Mean of the measurement

$r^{2}$: Variance of the measurement

Measurement update Formulas


The new mean is calculated as a weighted sum of the prior belief and measurement means.

$$\mu'=\frac{r^2\mu+\sigma^2\nu}{r^2+\sigma^2}$$

The new variance can be calculated as:

$$\sigma'^2=\frac{1}{\frac{1}{r^2}+\frac{1}{\sigma^2}}$$

This is implemented as measurement_update function in the code.

State Prediction Formulae


The state prediction is simply the sum of the means and variances to produce a new state prediction and variance respectively.

Posterior Mean: $$\mu'=\mu +\nu$$

Posterior Variance: $$\sigma'^2=\sigma^2_1 + \sigma^2_2$$

This is implemented as state_prediction function in the code.

This write was a part of Udacity’s Robotics Nano-Degree. The C++ code for this can be found here.