The key difference between an AI and a robot is that a robot can interact with, and change, the physical world. While seeing is an important sense, picking up objects is arguably one of the most important skills for a robot to have. Object manipulation research involves building novel hardware, designing flexible software and creating fully integrated systems. A big aspect of this research is machine learning and how to create robotic systems that are robust to changes in the environment.
QUT robotics projects consistently feature interdisciplinary teams working alongside industry partners to advance technology and design. The outcomes of the robotic object manipulation projects continue to establish new benchmarks for industry and robotics as a field of research.
We welcome industry collaborations in this field. For more information about collaborating with our team, please contact us.
Some of our people have gone on to join or found startups including Dorabot and LYRO Robotics.
Deep Learning for Grasping and Manipulation
Deep learning has taken the research world by storm. At QUT, researchers are using advanced deep learning techniques in combination with established approaches to solve new problems.
ACRV Picking Benchmark
There is an ongoing issue in research fields known as the reproducibility issue. Here, QUT researchers are fostering an important bench-marking system for robot pick and place research.
Amazon Picking Challenge (2017)
Our team came first in the global competition, against 16 teams from around the world competed, and winning the $80,000USD first prize. To do this, we created a novel Cartesian manipulator dubbed ‘Cartman’ with a rotating gripper to allow item pick-up using either suction or a simple two-finger grip.