Investigating and Modeling Human Perceptual and Motor Mechanisms with Application in Rehabilitation through Human-Robot Interaction

This thesis investigates the role of prior experience in human sensory and motor behavior in individual and interactive contexts. The final aim of this research project is to describe human behavior through quantitative models and subsequently develop novel robotic technologies that could support humans in their everyday life. To address this objective, the first study of the thesis has the goal of studying and modeling the role of prior experience in perceptual and motor mechanisms during an interaction. To cope with sensory uncertainty, human perception takes into account previous sensory experience, and potentially leads individuals to divergent percepts. Therefore, how can humans efficiently coordinate in joint actions? This thesis presents the implementation of a novel controllable and repeatable setup, developed for the exploration of perceptual and motor processes intervening during the temporal coordination with the humanoid robot iCub. Thanks to the development of this interactive environment, a perceptual task of interval timing estimation was realized to monitor participants' and the artificial agent's perceptual and motor behavior at each instant of their interaction. Results revealed that participants managed to coordinate in time with their artificial companion, by shifting their temporal estimations towards their partner’s perceptual model. The presence or absence of their interactive agent's motor feedback produced different effects on individuals' perceptual mechanisms. The presence of the robot's motor behavior induced 30 participants to imitate their companion's kinematics, leading towards the generation of separate representations of their own and the robot’s perceptual model. On the contrary, the lack of the robot's motor cues induced 32 participants to integrate the estimations from the previous interaction into their own perceptual model. Even more interestingly, by applying Bayesian techniques to perceptual responses, human perceptual estimations were successfully predicted after the interactive sessions in the presence or the lack of their partner’s kinematics. However, prior experience influences not only the magnitude estimation of sensory stimuli, but also their association with a reward. Therefore, the second study of this thesis aims to investigate the role of prior experience in the perception of reward-related stimuli in an individual setting with healthy and pathological populations, with the final goal of deepening the understanding of human perceptual behavior. In particular, a visual search experiment was conducted with 19 healthy individuals and 24 pathological gamblers. Since reward-related distractors have been demonstrated to delay participants’ reaction time, it was investigated if prior experience could differently affect pathological gamblers’ perceptual processes in presence of reward-related stimuli. Compared to the healthy population, pathological gamblers delayed their response in the presence of reward-related stimuli, independently of their associated value. Due to their uncontrolled impulse to gamble, pathological individuals did not vary their perceptual strategy, after an improvement in performance during the first part of the experiment. After investigating perceptual and motor mechanisms in different contexts, the third study of this thesis aims to induce motor imitation processes underlying human-robot interaction, with the final goal of influencing patients’ prior motor skills in a rehabilitation setting. Therefore, a novel clinical system was developed to exploit the interaction with the humanoid robot NAO during rehabilitative sessions. The system aimed to track participants’ movement and assess their performance through clinical labels, with the final goal of supporting patients’ motor skills, affected by cerebral palsy. Thanks to an effective communication with the system, the robot was endowed with the capability of providing patients with clinical feedback, according to their performance. Once the development of the system was completed, its validity and functioning were tested on 11 physical therapists and rehabilitation specialists. To conclude, the significant results of this thesis will contribute to develop bio-inspired adaptive models, aimed at having a complete understanding of human perceptual and motor behavior and monitoring related anomalies to prevent pathologies. Moreover, adaptive skills will endow robots with the ability of tailoring their behavior, to support humans in their daily tasks and improve their quality of life.