The physical model of an electrohydraulic actuation system with pressure-reducing cylinder end cushioning has been obtained. In order to arrive at the closure of the model, experimental models for actuator friction and the characteristics of the relief, non-return and proportional valves have been constructed. The variation in discharge through the proportional valve with both pressure and command signal has been modeled by training a neural network with experimental data. For the characterization of the discharge through the proportional valve, besides square root of the pressure drop in each metered orifice, polynomial forms of command signal for the discharge coefficients have been used. Such a direct characterization of the discharge with the command signal eliminates the orifice opening as an intermediate variable. A simple friction model that retains all the features of the existing complex models has been developed. Parameters such as maximum dynamic friction and the corresponding velocity have been introduced for this purpose. All these nonlinear subsystem models have been integrated together in MATLAB/Simulink frame to predict the actuation dynamics. The variations in the predicted and experimental displacements of the piston against different command signals have been found to be quite close to each other.