Empowering a robot with ability to grasp has been a long dream in robotics. Despite serious research effort over many decades, even the most advanced robots still struggle in the area of grasping. The main challenge is how to deal with errors and uncertainty. In grasp planning, traditional methods such as ICR and caging deal with the uncertainty problem at the contact level and suffer the poor performances. This thesis explores a more global outlook by aiming at finding a promising grasping plane (instead of individual points) to place the three contact points. The thesis presents a heuristic method to find a plane through the object to place the three finger contact. The input of the method is the input mesh model of the object. Physical experiment is also presented to show the effectiveness of the method.
