Hello! My name is Austin!
I am a Ph.D. Student at the University of Michigan. :)
My research focuses on the intersection between multi-agent reinforcement learning and game theory. I create algorithms that allow agents to find equilibria in both cooperative and competitive environments. I am grateful to be advised by Professor Michael Wellman.
Inspired by ant behavior, I combined learning with a pheromone-based multi-agent coordination algorithm that allowed fleets of automobiles to learn how to navigate to a goal location. End-to-end, they had to learn how to use large boxes to make new paths, choose which paths to take, which boxes/holes to use, and how to avoid collisions with each other.
Fully Online Decision Transformer for Reinforcement Learning
Transformers are not just for NLP! Framing RL as a sequence modeling problem for online settings.
Bayesian Imitation Learning with Robust Optimization
Enabling RL to account for worst-case scenarios by creating a tunably risk-averse agent.
I was born and raised in San Diego and went to UC Berkeley for my undergrad, where I had the opportunity to be advised by Professor Ronald Fearing. I am now based in Ann Arbor, Michigan where my San Diego roots are being challenged by this unpredictable weather. In my free time, I enjoy cooking, eating, hanging out with friends, playing/listening to music, climbing, gaming, and eating. I hope it is clear I very much like food. :D
Fun fact: I am 99.9% sure I can beat you in Super Smash >:)
Please do not hesitate to reach out to me. I'd love to talk to you about my research, or anything for that matter.
Shoot me an email at ngaustin AT umich.edu