Pneumatic artificial muscles (PAMs) belong to the group of nonconventional actuators with re-markable force/weight ratio that can be used for the construction of soft mechanisms safe in
contact with humans. In order to be able to design an effective control of 2-link soft robot arm actuated with PAMs, a dynamic model of this system needs to be derived. We use a PAM dynamic model derived usingfirst principles modeling (for contraction, pressure, and airflow dynamics) and ANFIS-based approximation based on the experimental data for the muscle force function. To derive the dynamics of the robot arm, we use Lagrangian mechanics approach for planar arm with the inertial and mass data based on the 3D CAD model. To val-idate the complete dynamic model of the soft robot arm, we used a gravity test (without PAM actuation) and pulse excitation for PAM control. The results confirm good validity of the dy-namic model for all relevant variables (joint angles, muscle contractions, and pressures) as well as the dynamic coupling between the joints.