Investigation of Physical Interactions in Humans Mediated By Robots

Humans perform interaction tasks in a wide range of everyday actions whether for interacting with objects or with other individuals. Although research has already been carried out to investigate human motor control in interaction scenarios, many aspects of the strategies employed by the Central Nervous System (CNS) for managing interactions still need to be investigated. This can provide insights into motor neuroscience and also lead to advantages on robotics and human-robot interaction, which could take inspiration from human motor strategies for designing better devices. This thesis investigates human interaction control focusing on different aspects: i) adaptation during physical interactions with novel dynamics, ii) control of a remote system interacting with the environment, iii) management of human-human interactions through robots. Each of these points will be explored in dedicated parts of the thesis. The first part of the work on external disturbance during motor tasks exhibits the individuals’ ability to adapt to complex position-dependent disturbances by understanding the position-intensity relationships. The second part, focusing on controlling a remote system, reveals the reliance of individuals on both limb position and impedance to manage remote interactions during human-based teleoperation tasks, independently from the control algorithm employed for the remote system. Finally, the third part on human-human interactions, shows individuals’ performance progressing from a theoretical study in a laboratory environment to real-life application to handwriting learning in children. It demonstrates that: i) achieving optimal performance with minimal effort in dyads where individuals possess diverse skills can be attained through asymmetric connections with a stiffer link to the less skilled partner; ii) a child’s performance is enhanced in executing a handwriting task when connected with an adult. In addition to these theoretical considerations, the proposed studies pave the way for novel practical applications in both robotic technology and educational methodologies: i) the novel teleimpedance control based on wearable sensor serves as a foundation for using this controller in teleoperation within unstructured environments; ii) the demonstrated improvement of dyadic performance achieved through an asymmetric connection between partners with different skills suggests the use of collaborative robots to optimally modulate the physical interaction in scenarios where partners have highly different skills, such as physical training or learning; iii) the improvement of handwriting skills in children achieved thanks to the connection with a more skilled subject suggests a novel handwriting exercise format that involves parent-child dyads.