William Curran Email & Phone Number

Arlington, Va, Us

Arlington, Va, Us

Mclean, Va, Us

Reston, Va, Us

I was the technical lead for Navatek’s efforts related to learning systems, with expertise in reinforcement learning techniques. I frequently wrote white papers and competitive proposals for the DoD and NASA, as well as act as the Principal Investigator and Project Manager for won proposals totaling $2.5 million from 2019-2021. These proposal topics primarily relates to the design of future robotic systems and sensors, autonomous control logic, simulation-based performance evaluation, and new theoretical developments in machine learning and intelligent systems. My research focused on applying novel artificial intelligence techniques in the certification of autonomous behaviors before deployment and then fault identification and sustainment analytics after deployment.I was the Principal Investigator and Program Manager for the following projects:• Reinforcement Learned Adversarial Agents for Active Fault Detection in Space Habitats (NASA). Reinforcement learning techniques to develop a fault prediction and detection solution that improves NASA's ability to reveal latent, unknown conditions that would lead to failure of life sustainment systems. https://www.sbir.gov/sbirsearch/detail/1881977• AI-based MDP Uncertainty Reduction Tool (Navy). Certification and testing of artificial intelligence and machine learning technologies to increase trust, understanding, and reliability.

Washington, Dc, Us

Worked on the Meso-scale Robotic Locomotion Initiative (MERLIN) to develop a small robot that a marine could carry in a backpack and deploy to conduct intelligence, surveillance, and reconnaissance missions. Due to the small size requirements, MERLIN uses hydraulics that have much more energy density, but are a harder engineering and modeling challenge. I developed Control Theoretic/Deep Reinforcement Learning hybrid techniques with the goal of robust, adaptive, and model-free control.

Corvallis, Or, Us

My research interests lie in developing learning algorithms for real-world robot control systems. In particular, my work focuses on scalable model-free reinforcement learning for adaptive robot control in systems where a mathematical model is not easily understood or computable. Specifically, I worked on:Dimensionality Reduced Reinforcement Learning (2015-2017)The complexity of state-of-the-art personal robots lead to large dimensional state spaces, which are difficult to learn in. To alleviate this issue, we developed a technique called Dimensionality Reduced Reinforcement Learning. This technique leverages concepts from dimensionality reduction, transfer learning and reinforcement learning to learn high-dimensional policies quickly.Project Chiron (2014-2016)Project Chiron is a self-driving wheelchair project at Oregon State University. We worked with individuals with ALS and the ALS foundation to develop a small package that can be mounted on a powered wheelchair to provide self-driving capabilities. In 2016, I was a part of a 4-man team sent to the Robots for Good competition in Dubai. We placed 7th out of over 700 entries.As Autonomous As Possible (2014-2015)In this work we introduce a risk-aware task-level reinforcement learning algorithm that adapts an end-user's risk tolerance. A3P learns a task-level policy where states are tasks and actions are approaches in accomplishing that task.

Washington, Dc, Us

I implemented a Memory for Goals cognitive model for human error prediction and embodied this model in a robot perceptual system for online error prediction in human-robot collaborations.Dr. Gregory Trafton, Navy Research Lab, Navy Center for Applied Research in Artificial Intelligence

Washington, Dc, Us

Continued dimensionality reduced reinforcement learning with PCA research. Applied this approach to Mountain Car 3D, Mario, and worked on applying it to an autonomous underwater vehicle.Dr. David Aha, Navy Research Lab, Navy Center for Applied Research in Artificial Intelligence

Moffett Field, California, Us

Developed a new simulation approach leveraging the concept of a multi-fidelity simulation. Computationally complex simulation is used only when agents are in a potential separation violation, while a low fidelity simulation is used for navigation. The key research focus is when to change from one fidelity to another, as small nuanced issues can easily occur that impedes agent learning.Dr. Adrian Agogino, NASA AMES, Moffett Field

Corvallis, Or, Us

For AADI I worked on Air Traffic Control with a focus on multiagent reinforcement learning. This work involved removing congestion and optimizing delay in the national airspace by controlling ground delay imposed upon aircraft. This approach is typically computationally intractable due to the extremely large size of the environment (40,000 aircraft per day). Developed an approach using domain-based agent clustering, essentially clustering agents into semi-independent groups and treating each cluster of agents as an independent problem. I then extended this work to become domain independent, removing the need for domain experts when performing agent clustering.Dr. Kagan Tumer, Oregon State University, MIME Department

Corvallis, Or, Us

Applied Machine Learning techniques on the image processing of multiple radar images. The goal is to design software to autonomously find rare bird roosts using radar images.Dr. Dan Sheldon, Oregon State University, EECS DepartmentDr. Tom Dietterich, Oregon State University, EECS Department

Corvallis, Or, Us

Lead a study exploring how users give feedback to machine learning vision algorithms. Lack of trust is a large issue in the Vision domain of Machine Learning. This study is designed to help develop techniques to alleviate this problem.Dr. Weng-Keen Wong, Oregon State University, EECS DepartmentDr. Sinisa Todorovic, Oregon State University, EECS Department

Corvallis, Or, Us

Explored the juncture between machine learning, human-computer interaction, and testing.Dr. Margaret Burnett, Oregon State University, EECS DepartmentDr. Alex Groce, Oregon State University, EECS Department

Corvallis, Or, Us

Worked with fellow students and the physics department at Oregon State University to launch OSU’s first CubeSat into space. I was the lead programmer and Vice-President of the OSUPSP. I also co-wrote proposals, recruited members, helped obtain funding and co-wrote programs for this project in Python.Dr. William Hetherington, Oregon State, Physics Department

Moffett Field, California, Us

Extended my previous Air Traffic Control work by simulating the entire National Airspace (40,000 planes) rather than a small subset, and developed an automated partitioning system with real world data.Dr. Adrian Agogino, NASA AMES, Moffett Field

Doctor Of Philosophy (Ph.D.), Robotics

Master'S Degree, Computer Science

Bachelor'S Degree, Computer Science