eranbTAU/Robust-MUO

Human-robot collaboration (HRC) in the man-ufacturing context aims to realise a shared workspacewhere humans can work side by side with robots nearby.In human-robot collaborative manufacturing, robots arerequired to adapt to human behaviours by dynamicallychanging their pre-planned tasks.In this Paper, we present a novel method for recognitionobjects using a non-vision method. Our method is based onlearning Deep Neural Networks, we developed a new archi-tecture “Flip-U-NET” that based on the familiar architecture“U-Net” which is used for segmentation in the medical field.The new architecture tries to solve the problem of robust-ness in the field of recognition. We have developed a noveldesign for a lightweight wearable device that will senseand record the dominating features during tasks. Recordeddata will be applied in data-driven modeling approaches tomodel the human arm in various tasks, this while reasoningabout uncertainties and sensory noise.