Matt Schmittle

Student Researcher in Drone Navigation
and Machine Learning

About

I am an honors senior Computer Science major at the University of Delaware (UD). My research is in deep reinforcement learning for drone navigation under Dr. Christopher Rasmussen. For more information on my research see below or contact me by email. While an undergraduate at UD, I have also participated in a number of other research projects: Colon cancer cell modelling under Dr. Keith Decker (2015), quadruped robot electronics design and build under Dr. Ioannis Poulakakis (2015), and backend containment and automation for the OpenUAV project under Dr. Vijay Kumar and Dr. Jnaneshwar Das (2017). I am planning to attend the University of Washington this Fall 2018 pursue a Ph.D. in Computer Science.

Outside of research, I am an Outing Club Officer at UD where I plan, and lead trips for groups of students. We go all over the country doing a variety of activities such as hiking, rock climbing, white water rafting, skiing, and more. I lead both introductory and advanced trips for students with a wide range of experience. Our membership is currently 10% of the student body and growing!

News

  • Recieved the Computer Science Outstanding Senior Student Award!

  • I have decided to pursue my Ph.D at the Paul G. Allen School of Computer Science at UW!

  • OpenUAV paper accepted to ICCPS 2018

  • Received Undergraduate research travel award for IROS

  • Abstract accepted with poster presentation at IROS 2017!

Research

Deep Reinforcement Learning and Drone Navigation for robotics

My research interests are in high-speed drone navigation and machine learning. Classical high-speed navigation approaches are too computationally slow and not robust enough to noise. Deep learning has shown to be fast and robust to noise. Specifically, I am looking at how deep learning can safely be used to learn a policy for obstacle avoidance.

Separately, I have done some work with the University of Pennsylvania on the OpenUAV project. OpenUAV is a open source, cloud-based, UAV simulator. The goal of the work is to reduce the barrier to entry for UAV research and education by making something free, hardware independent, and easy to use without much prior knowledge.

[ABSRACT ONLY] Formation Control and Persistent Monitoring in the OpenUAV Swarm Simulator on the NSF CPS-VO
International Conference on Cyber Physical Systems(ICCPS), 2018
Anna Lukina, Arjun Kumar, Matt Schmittle, Abhijeet Singh, Jnaneshwar Das, Stephen Rees, Christopher P. Buskirk, Janos Sztipanovits, Radu Grosu, and Vijay Kumar
[ABSRACT ONLY] Exploring DeepQ Learning for Micro UAV Tree Avoidance
International Conference in Intelligent Robotics and Systems (IROS), 2017
Schmittle M., Rasmussen C.
OpenUAV: A UAV Testbed for the CPS and Robotics Community
International Conference in Cyber Physical Systems (ICCPS), 2018, Acceptance Rate: 26%
Matt Schmittle, Anna Lukina, Lukas Vacek, Jnaneshwar Das, Christopher P. Buskirk, Stephen Rees, Janos Sztipanovits, Radu Grosu, and Vijay Kumar