Unified model for Grasp stability estimation

Grasp stability is one of the main concerns in the field of robotics, as it directly impacts the effectiveness and safety of robotic manipulation tasks. Grasping an object and evaluating the grasp is an important task when we use robotic hands. Establishing robotic hands that can securely hold and manipulate objects is crucial for a wide range of applications like manufacturing, logistics and healthcare. A reliable grasp stability model is essential to prevent accidents, reduce damage to objects, and improve the efficiency of tasks performed by robots. The goal of the thesis is to create an unified model for grasp stability prediction using machine learning techniques and to generalize this model over various known objects with different geometric features and properties of the object. The main challenge of this thesis is to create an unified model that overcomes the hardware constraints of the Mia hand, particularly its inability to precisely detect contact points and measure forces. A significant part of this thesis involves extending the model’s applicability to soft objects, which vary in their geometric characteristics and degrees of softness. The successful development of such a unified model for grasp stability is crucial for enhancing the functionality of robotic hands, making them more effective and applicable in a range of real-world situations.