Parallel robots are dominating industry, aerospace and medicine due to the low inertia, high acceleration and accuracy with respect to serial robots. Robot calibration is a major concern in practical application of robotic systems for identifying the performance characteristics of robotic platforms, decreasing the cost with an increase in robot performance. Robot calibration consist of three main steps, modelling, measurement and identification and it’s output model have a broad range of appalications in model based control theories and fault detection schemes. The closed kinematic chains of parallel robots leads to a challenging modeling procedures using conventional methods and this research aims at providing a promising solution to the dynamic modelling with a novel approach. At first, we introduce a systematic approach, based on the Gibbs-Appell formulation, for the modelling of robots, including serial robotic arms, parallel robots and cable driven robots. Providing the closed form and linear regression models are of the main results of this research applicable in robot control and identifications, which are verified through simulations and experiments. Due to motion constraints, all the dynamic parameters are not involved in motion behavior of the robot, using and representing the optimal regression model, we have obtained the motion behavior of the robots with the minimum number of parameters. Due to motion constraints, all the dynamic parameters are not involved in motion behavior of the parallel robots. One consequence of our study is utilizing the Gram-Shmidth theorem and singular values and obtaining of optimal regression model which represents the motion behavior of the parallel robots with the minimum number of parameters. In the following the identification is performed using a suitable excitation signal and using KF, BLS and LS methods. Based on the simulations and experimental results this method has a better performance comparing the popular identification schemes such as OLS. Improving the identification methods we aim at using the identified models in control schemes .In this regard we have developed the existing control schemes for the direct (explicit) use of optimal regression model and the performance is investigated in tracking problem. Covering the existing uncertainty like friction and flexibility a nonlinear model NARX, NARMAX,NOE are used as a hybrid model accompanying regression model. The other consequence of our study is introducing a linear-local method, based on optimal regression model, which has shown a suitable performance to overcome unmodeled dynamics, preserving the useful properties of linear model. Finally for the verification of the proposed method, experimental implementation is done on the Nasir Cable Driven Parallel Robot.