How do solve the initialization problem?

Dear everyone, I run the comand osr to compute the optimal simple monetary policy. But the optimal reaction coefficients rely on the initial values. Professor jpfeifer told me it the initialization problem. I checked it and found, actually, it is. I change my algorithm to opt_algo=9. But the results sitll rely on the initial values heavily.
So, could any one help me how to choose the initial values?

There is no unique starting value you could chose. You should try several starting values and then take the result that represents the global optimum.