A vast majority of the available biomechanical models of the human shoulder has been developed based on inverse dynamics, e.g. [1,2]. This imposes a number of limitations on their application. For instance, the glenohumeral joint is approximated as an ideal joint in an inverse-dynamics simulation. Therefore, the models fall short to predict the joint translations . The different approaches developed to overcome the recurrent limitations of the models can be broadly divided in two categories. The first category tries to tailor an available inverse-dynamics model to a specific application, e.g. [3,4]. The second category aims to develop a framework allowing forward-dynamics simulation, e.g. [5,6]. Indeed, few studies have developed forward-dynamics simulations of the human body. In , dynamic optimization was used to develop a forward-dynamics model of the lower extremity. Dynamic optimization typically demands many times integration of the equations of motion. Given the computational expense incurred by the integrations, the method is impractical for common applications. In this study, a framework for forward-dynamics simulation of the human shoulder is developed. In contrast with the dynamic optimization, the developed framework requires a single integration of the system equations. It is based on a joint application of a biomechanical model of the shoulder and a controller. The controller defines the muscle forces allowing the model to be simulated in forward dynamics. Different control scenarios are considered to investigate the model convergence in terms of accuracy and computational effort.