normalized state and input

Hi alexliniger,

Line 73 in 3ec7805

    
           stages_[time_step].cost_mat = normalizeCost(cost_.getCost(track_,xk_nz,time_step));

In this place, you did the normalization for the problem. And I guess it was done by replacig "x'=T_x \times x", "u'=T_u \times u" and "s'=T_s*s".
Then,

MPCC/C++/MPC/mpc.cpp

Line 235 in 3ec7805

    
           State x0_normalized = vectorToState(normalization_param_.T_x_inv*stateToVector(x0));

but here you just dealt with state x while doing nothing with input x and slack variable s? I don't know why?

The first part is normalizing the matrices of the optimization problem and thereby indirectly the optimization variables, the second line is normalizing the "measured" state x0
This function

MPCC/C++/MPC/mpc.cpp

Line 119 in 3ec7805

    
           std::array<OptVariables,N+1> MPC::deNormalizeSolution(const std::array<OptVariables,N+1> &solution)

is then de-normalizing all the variables back to their metric values.

Normalizing the optimization variables is only depending on the matrices and the initial state x0, so I don't have to normalize anything else. But once I get the solution I have to get it back to my normal space.

I hope this helps, otherwise I suggest you just write down all the math for normalizing a QP.

Thanks a lot. I got it.