Robotics is theme of Qualcomm EU student innovation program winners

Technology News | May 19, 2016

By Rich Pell

In the European edition, top three applicants Elias Mueggler from ETH Zurich and University of Zurich, Tim de Bruin from TU Delft and Jason Lee from ETH Zurich will receive $40,000 as part of the fellowship along with mentoring by a Qualcomm researcher to pursue their research.

“This year’s proposals focused on hot topic areas of research including, computer vision, machine learning, and autonomous navigation,” said Peter Rauber, Senior Director of Engineering at Qualcomm International, Inc. “QInF helps us support external innovation by working with top PhD students from elite universities across Europe to mentor them and help propel their ideas forward.”

Elias Mueggler’s proposal, “Event-based Vision for High-Speed Robotics”, focuses on investigating how to use event-based cameras in autonomously moving robots such as drones. Supervised by Professor Davide Scaramuzza, the student will bring microsecond-resolution image processing to fast moving mobile robots a step closer, by creating a generic six-degrees-of-freedom Simultaneous Localization and Mapping (SLAM) system with event based cameras.

Tim de Bruin, supervised by Professor Robert Babuska and Professor Karl Tuyls, has been selected for his proposal “Unsupervised Multimodal State-representation Learning for Robotics” which combines reinforcement- and deep learning to make robots learn a uniform representation of their state and their environment to learn new tasks autonomously.

Tim will investigate learning control policies in high dimensional state spaces which still adhere to physical rules. This project will look to improve autonomous learning for robots.

Selected for his proposal “A Unified Neural Language Model for Morphology, Grammar and Coherence” and under the supervision of Professor Thomas Hofmann, Jason Lee aims to bring together the benefits of character-level and sentence-level language models. Jason will model morphology, grammar and coherence jointly with a single neural network-based model.

The model will incorporate past sentences and words to predict the next word. Having no pre-defined linguistic rules to start from, the project will improve language-agnostic natural language processing.

The QInF Europe finalist event was hosted by QUVA, the joint research lab, announced by Qualcomm Technologies. Inc. and the University of Amsterdam, focused on advancing state-of-the-art machine learning techniques for computer vision.

More information about QInF at www.qualcomm.com/research/university-relations/innovation-fellowship