• Cartesian planner
• fPCA
• human movement
• motion planning
• robotics

Human-likeness is an important step to ensure safety and acceptability of human-robot interaction. The investigation of human upper limb kinematics is fundamental in different fields, which include generation of human like motions with anthropomorphic robots, but also neuroscience, clinical practice, with important implications also in rehabilitation and assistive robotics. However, ensuring Human-likeliness of robots' behavior is a critical task, far from being solved in literature. This thesis starts from the analysis of a dataset of real movements using functional Principal Component Analysis (fPCA) in order to extract the Cartesian domain features of human motion. These features are then used to develop a closed form Cartesian planner capable to handle different scenarios (free space planning, offline obstacle avoidance, real time obstacle avoidance and dual arm manipulation). Finally, the capacities of this planner were tested both in simulation and with real manipulators.