Make a robust human controller

[moorepants]

The current method of finding full state feedback with LQR on the Carvallo-Whipple model is not robust in controlling the non-linear model with linear or non-linear tires. One option is to use a very simple controller that we can manually tune, like $T_7 = k u_4$. This can work but will likely fail with the large lateral kick plate forces. I think the next thing to try is to linearize the new model and find an LQR controller for it. This should be robust for a reasonable slip angle range but may still break down for high slip angles when we get into the non-linear areas of the Pacejka tire model.

[moorepants]

Another option would be to use opty to solve for the open loop controls to manage the kick.