I am a computer engineering graduate student foucused on solving real-world problems with new control strategies and machine learning for robotic systems, particularly with UAV swarms. I am currently exploring career opprotunities for post-graduation in robotics and automation disceplines.
This website is used to provide an overview of projects I have made significant contributions towards, as well as the outcomes and artifacts.
I am a PhD student at the University of Nebraska-Lincoln studying computer engineering. My research interests encompass real-time systems, controls, and their applications to real aeropace technology. Our work has pushed the boundary of decentralized swarming capabilites for size, weight, and power-constrained vehicles.
Project Overviews
Prior to my working on my PhD, I graduated with a BS in mechanical engineernig at the University of Nebraska-Lincoln, and later an MS in mechanical engineering, with a focus on system modeling, dynamics and controls. I began working with Dr. Justin Bradley in the NIMBUS Lab where, as part of my research, I have learned to build and fly multicopters, and have led flight operations for multiple field campaigns.
When the drone batteries are charging, I am usually golfing, climbing mountains, or building something.
Hierarchical Reinforcement Learning for UAS Swarms
Objective: Use model-free reinforcement learning methodologies to generate path-planning and control policies for UAS swarm control and corrdination
Outcomes:
Multiple deployment of 8-drone swarms using learned policies.
Custom swarm-centric ground control station and simulator
Fleet of 20 custom-built UAVs designed for swarm control testing
Publications:
Grant Phillips, Justin M Bradley, and Chandima Fernando. A deployable, decentralized hierarchical reinforcement learning strategy for trajectory planning and control of uav swarms. In AIAA SCITECH 2024 Forum, page 2761, 2024.
Controller (under preparation for ICUAS)
Reachability-Based Collision Avoidance for UAS Swarms
Objective: Leverage formal methods concepts to prevent collisions within a real swarm of UAVs. Highlight the limitations swarming capabilities on size, weight, and power-constrained UAVs and underscore the importance of real-world swarm experimentation.
Outcomes:
Successfully deployed runtime verification to a swarm of 8 quadrotors.
Verification data collected across 140 real UAV flights.
Publications:
(under review)
