Theorem
For the LQ problem, let JNā=QNā and Ptā=Q+ATPt+1āAāATPt+1ā+B(R+BTPt+1āB)ā1BTPt+1āAwith final condition PNā=QNā.
The optimal cost is γāĪAāinfāJ(x,γ)=J(x0ā)=x0TāP0āx0ā+E[wtTāPt+1āwtā]with our optimal control being γtāā=ā(BTPt+1āB+R)ā1BTPt+1āAxtā