Computer Vision News - June 2023

17 Georgia Chalvatzaki I am sure you have tried to convince robots to do many things; some have succeeded, and some have not. What can we convince a robot to do, and what are they unable to do yet? What we have succeeded in is that we are now able to have a mobile manipulator robot, which means a robot that has two arms, a head, and a mobile base, so it is like a humanoid mobile robot. We have managed to make it effectively understand from a scene the object that it needs to grasp and then to be able to do sequential pick and place autonomously without us needing to program something on the robot. But when it comes to more challenging tasks, like needing these kinds of robots to open drawers, fridges, doors, put things in cupboards, take things out, or more fine manipulation skills like preparing a coffee for you, that is beyond just picking and placing stuff, we are very far away still from making this robot understand how to operate with all their body in an environment. This is still very challenging. You are a newprofessor – congratulations! Yes, I am a new professor! [ Georgia laughs ] Thank you! You have chosen a very special field for your research. Do you want to tell us about it? I research in the general field of intelligent robotic assistants. In order to enable robotic assistants to become useful to our society, meaning to become service robots in our homes or nurses at hospitals or in elderly houses, we need to endow them with specific capabilities. For example, we want our robots to be able to close this very difficult perception-action loop. We want this to happen not only for low-level actions, so how the robot would move and manipulate objects, how it can effectively grasp objects, and so on, but also for how it can use its perception in order to do reasoning. We are looking into the interplay between perception and action and reasoning to enable these robots to become of good service to us.