A parallel manipulator is a closed-loop mechanism composed of an end-effector having n degrees of freedom (DoF) and a fixed base linked together by at least two independent kinematic chains. Compared to serial manipulators, parallel manipulators have a higher payload-to-self-weight ratio. The rigid-link parallel manipulators provide high positioning accuracy, better stiffness and better dynamic characteristics. The usable workspace of the parallel manipulators is very conservative (by choice) due to the presence of kinematic or gain-type singularities, link interference and the actuator and joint limits. In addition, the parallel manipulators are over-designed so as to simplify the control. Such choices in design and restriction of the usable workspace increase the cost and also limit the performance capabilities of the parallel manipulators. The focus of this research is to analyse the dynamic behaviour of the parallel manipulator over its workspace and relate it to the kinematic singularities. Further, a real-time nonlinear model-based control framework is proposed, which would ensure safe operation with explicit consideration of the kinematic singularities. Thus allowing better exploration of the workspace of the parallel manipulators without the need for offline programming. The proposed control framework will be experimentally evaluated on two 6-DoF parallel manipulators: the semi-regular Stewart platform manipulator and 6-RSS parallel manipulator.