the model is inspired from Udacity's MPC course . here i have used car's x,y,velocity, psi,cte, error of psi as teh state vectors and two controls speed , steering angle . the model predicts the best set of control input which has low cost by trying out a combination of input controls .
this is the model which i used in this MPc to control the car . it has x,y,orientation , velocity , cross track error and psi error
after trying many combination of values and going through the walkdown video provided by udacity , i choose N = 10 and dt = 0.1 .
as suggested in the walkthrough video , i just converted the wy points to car centric points and fit a polynomial onto it .
after going through various notes , previous student works , i found out that the actuation is based on the previous state . the latency of 100 milliseconds makes this actuation of no use because the current state is different from the measured state . i have add modifications to the state
if (t > 1) {
a = vars[a_start + t - 2];
delta = vars[delta_start + t - 2];
}
a cost function is also added (combo of velocity and delta) to make the car go slow and take precious turn at corners