Human-Human, Human-Robot and Robot-Robot Interaction While Walking: Data Analysis, Modelling and Control

In everyday life humans perform many tasks with other partners which involve coordination, involuntary communication and mutual control adaptation, as the case of carrying objects together with another person. Humanoid robots may help with such activities by collaborating with humans or even substitute them. With this in mind, understanding how humans perform such tasks would be useful not only to better understand human motor control strategies but also to extract useful information for the implementation of robotic controllers. This Thesis aims to investigate if and how the haptic interaction among two subjects cooperating with each other while walking (e.g. while carrying objects) (a) alter their walking gaits, (b) is exploited by humans to understand the intentions of the other partners (e.g. to start/stop walking, accelerate/decelerate, etc.), (c) determine a walking synchronization and (d) whether such a synchronization can improve the overall task (e.g. by reducing the interaction forces). Moreover the feasibility of applying human-human strategies to develop robotic controllers is investigated. \Human-human studies are performed to assess the research objectives and replicate them through simple mechanical models. The main features extracted from the human experiments are then used to develop control modules for humanoid robots doing similar tasks with humans or with other robots. Robotic experiments are then used not only to assess the feasibility of using humans strategies in robotic control but also to try to answer some questions related to human behaviour. The obtained results show that most of the subjects analyzed (a) alter their walking gait while mechanically paired with another human subject, (b) communicate their intention through a combination of interaction forces and hand velocity, (c) coordinate their walking gait with their partner, (d) selecting specific quadrupedal animals types of gaits. Moreover human-humanoid robot and humanoid robot-robot experiments are successfully performed and used to show the importance of walking synchronization in reducing the interaction forces and guaranteeing more symmetric motion of the two bipedal agents.