Joe McCalmon

I am an undergraduate researcher at Wake Forest University, advised by Dr. Sarra Alqahtani and Dr. Dongwon Lee. I am also pursuing a bachelor's of science in computer science and mathematics, and I am the founder of The Robotics Club at WFU. I have received the Barry Goldwater Scholarship for excellence in undergraduate STEM research, and I was a finalist for the Computing Research Association's Outstanding Undergraduate Researcher award.

Email  /  CV  /  Github

profile photo
Research

I'm interested in the robustness and generalization capabilities in reinforcement learning agents. I am working on methods to identify areas of weakness of RL policies, in order to create agents which are certifiably robust.

CAPS CAPS: Comprehensible Abstract Policy Summaries for Explaining Reinforcement Learning Agents
Joe McCalmon, Thai Le, Sarra Alqahtani, Dongwon Lee
AAMAS, 2022

We create an algorithm to produce interpretable graphs for explaining the policies of black-box RL agents. Our method allows for end-users with little knowledge of reinforcement learning to understand the behavior of RL agents before deployment.

I will present this work at AAMAS 2022.

drone Deep Reinforcement Learning for Adaptive Exploration of Unknown Environments
Ashley Peake, Joe McCalmon, Yixin Zhang, Danny Myers, Sarra Alqahtani, Paúl Pauca
ICUAS, 2021

We develop a reinforcement learning agent for drone navigation in POMDPs which can zero-shot generalize to unseen environments. We are in the process of applying this work towards the detection of illegal gold mining in the Amazon Rainforest.

I presented this work at the AAAI-21 undergraduate consortium.

SaR Wilderness Search and Rescue Missions using Deep Reinforcement Learning
Joe McCalmon, Ashley Peake, Yixin Zhang, Benjamin Raiford, Sarra Alqahtani
IEEE SSRR, 2021

We develop a reinforcement learning agent for search and rescue in POMDPs which can generalize to different target locations.

symb_uav A Symbolic-AI Approach for UAV Exploration Tasks
Yixin Zhang, Joe McCalmon, Ashley Peake, Sarra Alqahtani, Paúl Pauca
ICARA, 2021

We employ a probabalistic information gain map to encourage the efficient exploration of drones in large, unknown environments.

platoon Multi-Agent Reinforcement Learning for Cooperative Adaptive Cruise Control
Joe McCalmon, Ashley Peake, Benjamin Raiford, Sarra Alqahtani
ICTAI, 2020

We propose a multi-agent reinforcement learning approach for autonomous vehicles which communicate in a platoon formation.


Website source code is adapted from Jon Barron.