In many real lie situations, we grasp objects that change their internal stiffness rom time to time. Holding a live hamster is an example. In such cases how do we learn the best grip force without over-squeezing the object? Recently we published (PDF) our findings in a probabilistic approach to grasp such objects when the critical grip force was known. In this project, the student will extend the grip controller to with autonomous learning of the critical grip force.