Abstract

My talk will present experiments and computational modelling to i) understand how humans learn to skilfully interact with their environment and with each other, and ii) endow robots with similarly successful control behaviours. Impedance control describes how biological or robotic motor commands can shape the mechanical interaction with the environment. We have previously provided the first clear experimental evidence of impedance control in humans and described the underlying computational mechanism, which led to a novel adaptive behaviour for robots. This talk will first present how this behaviour can be/is used for haptic exploration of rigid and soft objects in robots and humans. I will then concentrate on our recent research to understand how do humans exchange haptic information through touch in order to coordinate their actions. By examining the behaviours of individuals when their right hands are physically connected, we could show how haptic information enables humans to estimate partners’ motor plan and use it to improve one own performance. Embodied as a robot partner, this model was verified as it induced the same improvements in motor performance as a human partner. We further elucidated how the sensory exchange is influenced by the interaction mechanics and how its benefits increase with the number of partners. These results elucidate the haptic communication taking place between physically interacting humans and promise collaborative robot systems with human-like assistance.