Total: 1
In this work, we first prove that the separation principle holds for switched LQR problems under i.i.d. zero-mean disturbances with a symmetric distribution. We then solve the dynamic programming problem and show that the optimal switching policy is a symmetric threshold rule on the accumulated disturbance since the most recent update, while the optimal controller is a discounted linear feedback law independent of the switching policy.