Model Predictive Control vs Pure Pursuit

I was building a robotic car that can navigate through points on a field that are predefined. However to do this, I have two possible approaches in order to steer accurately and quickly. I can use a pure pursuit algorithm which sometimes has errors steering but is generally simpler to implement than an MPC. An MPC is much more accurate and can be much faster. I wanted to get everyone’s take on this problem and what they would use.


Hi Aditya,

very interesting question. I never had a project with a “car” involved, but i guess i would start with a PurePursuit because, as you said it is easy to implement. So you have an early running version you can do measurements on.
And then i would try to implement MPC, because i think - as you also said - it will give better results. Pretty sure.
Doing it this way you first of all have fast visual/usable progress on your project and can then compare and really measure HOW much better the results are and maybe find some configurations to optimize it.

Looking forward to see what worked best (and how much) :slight_smile:

Hello @aditya_mangalampalli,

I didn’t work with these methods, but generally I would also start with simpler version.

While doing some short research, I’ve found this short but cool video :slight_smile: