Decentralized guaranteed cost control of interconnected systems with uncertainties: A learning-based optimal control strategy

A novel learning-based optimal control approach is constructed to attain the decentralized guaranteed cost controller design for a class of continuous-time complex nonlinear systems with dynamical uncertainties and interconnections. This is performed by combining robust decentralized control formulation with adaptive critic learning technique. By expressing the interconnected subsystems as a whole system and introducing a new cost function for the overall plant, the decentralized guaranteed cost control problem is formulated as an optimal control problem for the nominal overall system. Then, a policy iteration based learning control algorithm is employed to solve the modified Hamilton-Jacobi-Bellman equation with respect to the nominal plant iteratively. A critic neural network is constructed to approximate the optimal state feedback control law and then the uniform ultimate boundedness stability issue is analyzed. Meanwhile, a simulation experiment is conducted to verify the good performance of the control approach.