The pendulum-swing problem is a factor that limits the development of parafoil-unmanned aerial vehicle (UAV) systems. Autonomous attitude control based on parafoil and UAV control mechanisms is considered an effective solution to this problem. However, due to the coupling effect of the two control mechanisms, conventional control methods are not suitable. The design of attitude control for parafoil-UAV systems has become a challenge. For this problem, a model-independent optimal control method called multivariate extremum seeking with the Newton method (ES-NM) is introduced in this paper. To assess the performance of multivariate ES-NM control for parafoil-UAV systems, a multibody dynamic model based on the flexible line assumption is built. The aerodynamic coefficients of this model are estimated via computational fluid dynamics (CFD) and corrected using flight data. Using this model, the coupling effect of the two control mechanisms is investigated, and the control range is determined. Finally, the effectiveness of multivariate ES-NM controller for a parafoil-UAV system is verified. Simulation experiments performed under various conditions demonstrate that the multivariate ES-NM control can manipulate the UAV control mechanism and the parafoil control mechanism simultaneously and produce the desired UAV attitude track. Additionally, comparisons to proportional-integral-derivative (PID) control reveal the better performance of the proposed control method. INDEX TERMS Parafoil-UAV system, nonlinear multibody dynamic model, multivariate ES-NM control, attitude optimal control, attitude tracking.