The repository includes an example of propagating poses through SE(2) process model and demonstrates how to compute a first-order uncertainty bound.
The script heavily follows this MATLAB script by UMich-CURLY-teaching.
Note that for longer trajectories, the confidence bounds are no longer consistent. This may be due to simplified Jacobian computations (e.g., the left/right Jacobians are not used).
As can be seen here, the effect is minimal for the presented example. Note that the effects may be more substantion for different examples (I'm trying the dark-themed plots here).