The important component required for an e ective human-robot collaboration in a shared autonomy is the prediction of human intention. In this thesis work, a Bayesian probabilistic estimation of user intention for grasping an object is presented. A Virtual Reality (VR) device is used as a control interface for the user to have an e ective communication in robotic teleoperation tasks. Robot Operating System (ROS) framework is used for the communication between user and robot. The proposed method is implemented in a supervised learning approach to learn the likelihood distribution of the data. Experiments are conducted with the collected trajectories and results show that the proposed model performs better for the unseen data too.