Time independet ema for path planning smoothing

[Flova]

Summary

Currently, the path planning exponential smoothing is dependent on the update frequency of the path planning. But this can be formulated in a time independent discrete way by using the time constant tau (see https://en.wikipedia.org/wiki/Exponential_smoothing#Time_constant). It described the amount of time a unit impulse takes to decay to 1 - 1/e which is roughly 63%. Using this, we can calculate a smoothing factor alpha for arbitrary time deltas instead of using a fixed one. This accounts for changed rates as well as fluctuations in the timers rate.

Here you can see two impulse responses with different discretization, but same tau (= 0.39):

Note: k and alpha are used as synonyms here for legacy reasons.

Proposed changes

• Fix typo
• Add flexible time delta
• Use tau instead of k/alpha
• Use ROS 1 parameter value for smoothing. ROS 1 used a slower update frequency for the controller, but with this we can translate the smoothing parameter to the new 20Hz update rate (we need to check that on the real robot!)

Checklist

• Run colcon build
• Write documentation
• Test on your machine
• Test on the robot
• Create issues for future work
• Triage this PR and label it

Idea: We can add this also to the IMU filter!

[SammyRamone]

lgtm

[Flova]

We did some parameter tuning, but further testing is required