Towards robust human-robot mobile co-manipulation for tasks involving the handling of non-rigid materials using sensor-fused force-torque, and skeleton tracking data

Abstract

Over the past decades, robots have been extensively deployed in multiple industries. More recently, industrial robots have been taken out of their cages, being more present in dynamic, uncertain environments interacting in the vicinity of humans. Traditionally, robots have been mainly developed to perform pre-programmed tasks. However, some tasks are too complex or expensive to be performed by a robotic system alone. One of these hard tasks is the handling of large fibre sheets in composite part production. This paper presents a force-based approach towards human-robot co-manipulation of tasks involving the handling of non-rigid materials, such as composite fibres. Our approach fuses the data of a force-torque sensor and skeleton tracking data from a 2D camera to control the mobile manipulator in an intuitive manner, by using the intelligence of the operator as much as possible. By using this approach, tools such as path planning, high-level task planning, or modelling of objects to be manipulated are not essential to obtain the results. The overall approach is illustrated with a co-manipulation proof-of-concept experiment in which a mobile manipulator robot handles a flexible textile sheet together with a human operator.