In order to realize the decentralized robust adaptive control for a class of the distributed multi-input system(DMIS), based on the state extending and back-stepping diturbance attenuation method, a new autonomous robust adaptive decentralized control method is proposed. Based on the direct feed-back linearization and the optimal control method, the optimal autonomous robust adaptive decentralized control method is realized. Simulation results show that the proposed method can stabilize the DMIS, be adaptive to the uncertain parameters and can be applied to minimize the impact of disturbance by solving the LMI to obtain the optimal control law against the worst disturbance.