[Torcs-users] Developing a Kalman Filter-based controller

[Rendy A. Y <set...@gm...>]

Hi, everyone! I recently discovered TORCS after trying to look for
something for my final paper project in my college. I am interested in
making my own controller and am still in the learning phase.

For my final paper project, I decided to use Kalman Filtering for my
controller, based on this algorithm as shown by Rhudy et. al. in the paper
"A Kalman Filtering Tutorial For Undergraduate Students" (2017). The
algorithm I plan to use would be as shown as in section 3.2. which details
how the filtering is done each state. The eventual product will be used
with the SCR Server application and will be in C++ programming language.

In my project's paper, the purpose for the development of this controller
would be to see the performance of the controller in situations consisting
practice, qualifying, and racing against other AIs, driving the same set of
cars (in this case ) on five different tracks. I would envision that
algorithm would find a good racing line in each state, in addition to the
behavior where the controller would try to avoid an AI car when trying to
overtake.

Anyone has suggestions on the filtering part or a better algorithm to use
to achieve this? It's rare to see a final paper centered around TORCS and I
can't recall any of a final paper in my college that would be of a
reference, which is the reason why I consult to this place.

Thank you very much in advance, and hope you guys have a nice day!

~[R]

P.S.: Any forum recommendations on where I can talk about AI/controller
developments, especially regarding TORCS controllers? I would appreciate it!