DNN‐Based AI‐Driven H2/H∞ Control and Two‐Coupled H2/H∞ HJIE‐Supervised Adam Learning Algorithm to Rephase HJIE‐Embedded Algorithm

ABSTRACT This study results in the exposure of a deep neural network (DNN)‐based control scheme to go with the mixed reference tracking control design for a sort of nonlinear dynamical systems in contingent to the external disturbance. To be exact, the crafted mixed reference tracking control design has got to solve two complex partial differential hamilton jacobi equations (HJIEs) at the same time, one for the optimal reference tracking control design and one more for the robust reference tracking control. In contrary with the traditional fuzzy interpolation, global linearization and gain scheduling schemes for interpolating the local linearized systems to approximate nonlinear systems, the Hamilton Jacobi equations has been rephrased to a frame of set of two dissimilar linear matrix inequalities to attain the robust reference tracking performance against external disturbance and optimal reference tracking performance, simultaneously. The main difficulty in the traditional nonlinear mixed control design lies in how to solve the nonlinear partial differential hamilton‐jacobi‐isaacs equation (HJIE), especially for nonlinear dynamical systems. Hence, an Adam learning algorithm is pledged to train the HJIEs‐embeded DNN to realize the mixed reference tracking control strategy of the nonlinear dynamic systems by solving the two developed HJIEs concurrently. Further, the theoretical mixed reference tracking control results of the proposed DNN‐based control design can be guaranteed if the training error drives to zero. Also, the projected mixed DNN‐based reference tracking control can retrieve considerable training data and computational time for more practical applications. Thereby, the Adam algorithm converges, we can exhbit that the put forwarded HJIE‐reinforcement DNN‐based decentralized tracking control strategy can bring off the theoretical result for the considered nonlinear dynamical systems. At long last, the efficaciousness and supremacy of the proposed DNN‐based mixed reference tracking control is signified with an application based example of unmanned ariel vehicle (UAV) reference tracking system with external disturbance.