Taekyung Kim

Hi, @SivertHavso Do you have a plan to upload the ROS2 ported version? It will be greatly helpful for me even if the implementations are dirty.

@SteveMacenski This is basically correct. But, 'gamma/std' should be 'gamma/(std^2)'. $\Sigma$ in the paper means variance, so the denominator should be square of std $(\sigma^2)$.

That is a valuable observation. First of all, the std in the control cost should be squared in theory. The control cost is theoretically derived from the free energy and...

> ... which is really annoying. I'm not sure which, either, or neither is right facepalm This is one of the most confusing part of MPPI. Since they have 3...

That's an insightful opinion. I agree there is no generally acceptable __golden__ parameters for `lambda` and `gamma`. They are 100 % adjustable. Plus, I agree that the difference may come...

I temporarily closed #98, so that we could resolve the issues from gamma and control cost first. @artofnothingness Please see the latest comments from @SteveMacenski in #98 about the ideas...

I have looked over your thoughts and results today. The bottom line is that this is valid but it would be better to add some components here. First, the “clamping”...

> Clamping now is just handling the hard constraint enforcement, so the control_sequence_ optimal that comes out shouldn't be shifted very much due to enforcement - this critic handles the...

> I think this is happening because we're evaluating trajectories that aren't clamped yet and are weighting towards shortcutting. I'm not sure but I think taking shortcutting may not originate...