SINTEF Ocean is leading a research project, called iProcess (www.iprocessproject.com ) aiming to develop novel concepts for flexible robot based automation in the food processing industry. An interesting area of research in WP3 (http://iprocessproject.com/wp-3-flexible-processing-automation/) consists of a vision-guided and machine-learning robot that can grasp/manipulate compliant food objects. Given the 3D images from a RGB-D camera, the aim is to develop a grasping concept that finds a suitable grasping pose for the gripper tool. In principle, there are different learning strategies to establish and enable kinematic grasping such as Learning from Demonstration (LfD) and Reinforcement Learning.
In this project assignment, the student will implement Deep Reinforcement Learning for 6DOF pose estimation based on RGB-D images. A dataset of RGB-D images is already collected but the student may collect additional data for training. In reinforcement learning, the goal is to maximize the overall reward for successful grasping, and therefore, in contrast to LfD there is a metric function that enables improvement of learned behaviour i.e. grasping of the compliant food object.
The assignment is as following:
• Based on the dataset of existing RGB-D images and utilizing Tensorflow, implement Deep Reinforcement Learning using an appropriate Deep Learning architecture and reinforcement learning policy (for example Q-learning or Policy Gradient).
• Use Deep Reinforcement Learning to predict a single or a set of potential gripper vectors (3DOF position and 3DOF orientation) for the object in the given RGB-D images.
A state-of-the art workstation with the necessary software for to work with the images will be made available to the student.
• Excellent programming skills in Python
• Knowledge and interest in deep learning, machine learning and computer vision
Co-supervisors from SINTEF Ocean:
Dr. Ekrem Misimi – Senior Scientist (email@example.com) and Alexander Olofsson.
IMPORTANT: If you sign up for this project, please send a) your CV (including a transcript with all of your college grades, and b) a brief explanation of WHY you want to do this particular project to Prof. Keith Downing (firstname.lastname@example.org)