Adaptive neural network passivity-based control with state observer for partially unknown nonlinear systems

To overcome the limitations of interconnection and damping assignment passivity-based control (IDA-PBC) in partially unknown nonlinear systems, particularly the reliance on large damping coefficients that may cause control inputs to exceed physical constraints, this paper proposes a novel adaptive neural network passivity-based control strategy incorporating a passivity-based observer. In the proposed framework, neural networks are introduced to relax the strict requirement of the desired closed-loop port-controlled Hamiltonian (PCH) model and to construct dynamic compensation mechanisms. These mechanisms are embedded into the interconnection and damping assignment passivity-based observer (IDA-PBO), resulting in an adaptive neural network IDA-PBO (ANNIDA-PBO) with enhanced dynamic compensation capability. On the control side, neural-network-based adaptive laws and compensation terms are developed to accurately approximate and compensate for unmodeled dynamics. This design alleviates the inherent limitation of conventional IDA-PBC methods that depend on large damping coefficients to achieve robustness. As a result, the proposed framework enables a well-conditioned selection of control gains within physical constraints and effectively decouples the strong dependence between robustness performance and damping parameters in standard IDA-PBC designs. Furthermore, the closed-loop system is shown to be semi-globally uniformly ultimately bounded through port-controlled Hamiltonian modeling and Lyapunov stability analysis. Finally, simulation studies on two representative nonlinear systems are conducted to validate the proposed method, demonstrating improved control accuracy and enhanced robustness against external disturbances.