Investigation and modelling of human upper limb kinematics for robotic applications

Investigation of human behaviour is important in robotics for many rea- sons, as the implementation of human-like movements in robots and to shape the design of simple yet effective robotic devices as witnessed by the employ- ment of hand synergies, broadly defined as common covariation of hand joints to develop and control robotic manipulators with a reduced number of control inputs and actuators. In this thesis, I propose a multi-sensor experimental procedure to study human upper limb movements in Activity of Daily Liv- ing (ADL); the procedure consists of a protocol to select the most significant human actions based on a state-of-the-art taxonomies on grasping and human kinematic workspace; an experimental Setup for MoCap acquisitions which also considers the integration of other sensing modalities such as force measurements and EEG, and kinematic model and filtering Techniques to retrieve upper limb joint angles. I performed experiments using this prootocol with external partic- ipants. Among the points, I focused on extracting the most useful features to describe upper limb movements, in order to find representative trajectories that can be implemented to a robotic manipulator to achieve human-like motions. Main motivation for this are the outcomes of several studies that identified an- thropomorphism as one of the key modelling factor for successful, acceptable, predictable and safe HRI in many fields such as human robot co-working and rehabilitative assistive robotics. At the same time the identification of principal motion features can be successfully exploited to simplify the planning phase for anthropomorphic robotic manipulators. To achieve this goal I leveraged upon functional PCA, a statistical method for investigating dominant modes of vari- ation of functional data in time. The use of FPCA is particularly appropriate for this study because it allows to explore some important features of the signal, such as shape and time dependence