Understanding Quadrotor Tracking Example

[nfc35]

Hello,

I am looking for some help understanding the outputs from the quadrotor tracking example here: https://github.com/TinyMPC/TinyMPC/tree/main/examples. I am trying to implement another computational technique for doing the LQR optimization, and hoping to use TinyMPC as a reference to evaluate the correctness of my implementation.

My implementation is suggesting that the optimal solution for the tracking example included with TinyMPC is control outputs of 0, and that this will produce no error between the desired and actual states given the first state as an initial condition. Looking at the A dynamics matrix in the example, this actually seems logical to me.

However, this is not the solution that TinyMPC produces. TinyMPC produces a solution that has non-zero control signals, and has error between the achieved states and the desired states. Is this an artifact of the optimization technique TinyMPC uses, or the precomputation done to make it more efficient? Or am I misunderstanding something about the model / dynamics?

Thanks!