In pneumatic soft robotics systems, pneumatic soft actuators (PSAs) are the most important components, so it is necessary to study the precise modeling and control over PSAs for application. This article focuses on a self-designed planar PSA, named pneumatic bellow actuator (PBA), and proposes a modified three-element model (MTEM) and a nonlinear disturbance observer-based sliding mode control (NDO-based SMC) method to achieve the robust tracking control objective of the PBA with multiple uncertainties. First, the dynamic characteristics of the PBA are analyzed based on experimental data obtained from open-loop experiments of the PBA. Then, based on these dynamic characteristics, a semiphenomenological MTEM is established considering coupling between the dynamics of the PBA and load as well as friction, which can describe the dynamics of the PBA with changing loads in the rough horizontal plane precisely. Model parameters are then identified by the Levenberg–Marquardt algorithm. Next, identification errors, device disturbances, and external disturbances are regarded as the multiple uncertainties of the PBA, and the NDO-based SMC method is proposed based on the MTEM to realize the robust tracking control objective of the PBA. The stability of the control system is proved. The effectiveness of the proposed method is verified by experiments.