Mechanism of interpersonal sensorimotor interaction

This thesis uncovers the mechanism of interaction between human partners in continuous contact. During such sensorimotor interaction, I propose that people interpret their partner's desired movement through the forces they feel. The evidence of this interpersonal goal integration mechanism comes from three behavioural experiments and computational modelling of their results. Novel dual-robotic interfaces were used in the experiments for two reasons: first, to eliminate the partner's gaze as it was found to influence the behaviour of interacting pairs of subjects. Second, to systematically and quantitatively measure the effects of sensorimotor interaction on task performance and adaptations in muscular activity, i.e. motor force and joint impedance. The first experiment revealed that sensorimotor interaction during a tracking task, where partners track a randomly moving target with their cursor, enabled both the worse and better partners in the pair to improve at tracking. This mutually beneficial effect of interaction, regardless of the partner's skill, cannot be explained by existing theories in psychology and neuroscience, but the discovered interpersonal goal integration mechanism does capture this behaviour; when this interaction mechanism was implemented in a robot partner, people interacting with this robot behaved similarly to those during interpersonal interaction. Adaptations in the muscular activity were observed in the second experiment, where pairs were rigidly or weakly coupled during a tracking task. The better partners increased their motor force and joint impedance, whereas the worse partners decreased both. The interaction mechanism also predicted these changes in the motor force. Finally, an experiment where four partners interacting together during a tracking task showed that the best partner's performance was not hindered by many worse partners, which was also predicted by the interpersonal goal integration mechanism. The interpersonal goal integration mechanism is the first robotic algorithm to exhibit human-like behaviour during sensorimotor interaction. It has applications where continuous contact with a human partner is critical, such as the transport of large objects or in the co-operative control of lower-limb exoskeletons. It offers a new approach for robotic rehabilitation through continuous, physical interaction which exploits the benefits of simultaneous action observation and physical training. Finally, the interaction mechanism may be used as an investigative and therapeutic tool for autism spectrum disorder (ASD). Individuals affected by ASD commonly lack the ability to extrapolate someone's goal by observing their actions alone. Similarly, they may lack the ability to estimate a partner's desired movement from observing haptics forces alone, thus the interpersonal goal integration mechanism may shed light on the specific differences between people with ASD and healthy individuals.