Visual Failure Detection in Robotics

Promotionsprojekt im Überblick

Robots are usually programmed to perform tasks by following a sequence of actions such as moving, looking, picking, etc. When something unexpected happens, a robot often cannot handle the situation because 1) it has not recognized that something has gone wrong and 2) it has not been programmed for the new situation. Recognizing such failure situations allows robots to decide whether to proceed with the task, inform a human about the problem, or try to solve the problem themselves. The project focuses on developing machine learning models to detect unexpected situations that could lead to a failure of the task by using videos from the robot's camera, task information, and other sensors.
Visual Failure Detection in Robotics

Fachbereiche und Institute

Fachbereich Informatik Institut für KI und Autonome Systeme (A2S) Graduierteninstitut

Zeitraum

22.01.2019 to 21.02.2026

Doktorandin/Doktorand

Betreuende Professorin oder Professor

Professor im Ruhestand / retired professor Dr. Paul G. Plöger

Projektbeschreibung

Robots are usually programmed to perform tasks by following a sequence of actions such as moving, looking, picking, etc. When something unexpected happens, a robot often cannot handle the situation because 1) it has not recognized that something has gone wrong and 2) it has not been programmed for the new situation. Recognizing such failure situations allows robots to decide whether to proceed with the task, inform a human about the problem, or try to solve the problem themselves. The project focuses on developing machine learning models to detect unexpected situations that could lead to a failure of the task by using videos from the robot's camera, task information, and other sensors.