Methodologies to create and maintain semantic world models

Description

Current robotic applications rely on the closed-world assumption, in which the real world exactly matches the designer’s assumptions. In other words, unforeseen circumstances are excluded and do not need to be dealt with by the robot. There is a desire in many application domains to expand the use of robots in less structured and less known environments, in which the robot would need to explicitly account for the uncertainty and incompleteness of its knowledge of the surrounding world to take appropriate autonomous actions. The Eindhoven University of Technology (TU/e) Robotics Group develops methodologies to create and maintain semantic world models, which can be used by robots to assess the state of a semi-open world in the context of its task at hand. Semi-open means that the robot will be able to explicitly account for a limited set of potential events and artifacts in the world that may or may not occur. The word “semantic” is important to be able to give meaning to the objects in the world model, such that the robot can reason about the state of the world and the actions it must take to perform its task. The TU/e Robotics Group assists companies in transversing robot applications from closed-(of)-world to semi-open worlds. In the domain of healthcare, the robotics group applies these methodologies to robot-assisted surgery (eye, skin, skull, abdomen regions), general service purpose robots and mobile (indoor & outdoor) logistics solutions. The group actively validates these technologies in the form of hardware demonstrators. It currently (2019) holds 2 World Championship titles in Health-care service robots (DSPL) and soccer (Mid-Size-League). It collaborates with companies like Toyota, and has 5 spin-off companies in the past 10 years, being PRECEYES, MICROSURE, EINDHOVEN MEDICAL ROBOTICS, SMART ROBOTICS and RUVU-BEHAVIOR FOR ROBOTS. It also actively participates in many National and European Research Projects like RoboEarth, EurEyeCase, R3COP, R5COP, AUTOPILOT, ROPOD, KSERA, CRYSTAL