A dual control method for MIMO stochastic systems

A new dual control method is proposed for the MIMO stochastic systems with unknown constant parameters, which converts the unsolvable dynamic programming problem into multiple single-step minimum variance control problems. In every instant, innovation is introduced to improve learning. Unknown parameters are estimated using Kalman filter and the present minimum variance control can be solved by use of the estimation. Then, the innovation in the next instant is obtained and the Kalman filter is reused. Thus, the dual control can be implemented. Finally, simulation results show that the control has dual property that can achieve preferable estimation results and satisfactory control performance.