Iit Annual Report Year 2007 Grasping Strategies: Humans Vs Humanoids

In traditional robotics, the problem of grasping and manipulation is typically addressed by considering simplified situations, i.e. a strongly structured environment or a simulated one. In many cases, in industry, all the required actions are a limited subset of the possibilities of humans, which allows using standard (often kinematically deficient) manipulators and simple end-effectors (grippers). This further simplifies the planning and execution of tasks. In these cases the problem becomes tractable from a mathematical point of view, exploiting a-priori knowledge of the object properties (hardness, weight, affordance, position and orientation). Even when the end-effector is an antropomorphic hand, constrained kinematic and dynamic equations can be used to model, plan and execute form closure or force closure grasps [11]. Exact knowledge of the geometrical and friction properties are required. In this area a considerable amount of work has been carried out in the last few decades [12, 3]. In a real-life, the scenario is different and this approach is definitely unfeasible. This is not an approach for a humanoid robot to manipulate, and it is not definitely the approach “implemented” by the human brain. Although we don’t know yet exactly how our brain plans and executes the wide repertoire of movements that we use daily, it is probably true that we employ efficient procedures for solving the kinematic and dynamic problems of everyday’s tasks. It seems that there is something like a direct mapping between perception and action, that does not go through a precise and fully comprehensive representation of the world (or, in other terms, that our world representation has to be intended in terms of action and perception over the world itself, and not in terms of a list of objects features [1]). Following recent trends in neuroscience, the development of knowledge in human beings seems to be the result of a continuous interaction with the environment and can be identified with the co-development of both perceptive and motor systems, which occurs following a process of mutual improvement: indeed, a better perception permits the execution of more and more sophisticated tasks and a more mature motor system improves dramatically active perception [4, 2, 18].