Dynamic portfolio optimization problem

Hi all,

To compute the portfolio weights of 2 risky and 1 risk-free asset, I took the following steps:

For the risky assets, I simulated per step in time 10000 asset paths. My objective is to adjust every month the weights of the risky and risk-free asset so there are twelve timesteps per year. The asset paths are predicted by dividend yield, (or state variable Z.)

The risk=free rate is fixed at 6 percent.

So now i have two matrices, for each risky asset one, and a matrix voor the state variable.

I think solving backwards through recursive least squares is the best solution for this optimization problem but I have no clue how to start with my variables. So any suggestions would be very helpful.

Thanks!