In this project we aim to develop tools for heterogeneous robot and human-robot teaming for search and rescue missions. Specifically, the aerial drones can cooperatively follow as well as lead human ground operators during search and rescue mission. Some capabilities developed are showcased as part of project Astralis under TL@SUTD.
Y. Loo, G. Chen, M. Meghjani, "A Hierarchical Approach to Population Training for Human-AI Collaboration", International Joint Conferences on Artificial Intelligence, 2023. [Project Page]
G. Chen∗, D. Nguyen-Nam∗, M. Meghjani∗, P. M. Tri, M. B. Prasetyo, M. A. Daffa, T. Q. S. Quek, "Astralis: A High-Fidelity Simulator for Heterogeneous Robot and Human-Robot Teaming", IEEE International Symposium on Safety, Security, and Rescue Robotics (SSRR), 2022. [PDF] [Video] [Code]
Multi-target pursuit evasion
In this project we develop algorithms for multiple agents to track multiple targets that are constantly learning to evade. The agents are trained to perform heterogeneous roles of pursuing and scouting for targets using multi-agent reinforcement learning (MARL).
M. Kouzeghar, Y. Song, M. Meghjani, R. Bouffanais, "Multi-Target Pursuit by a Decentralized Heterogeneous UAV Swarm using Deep Multi-Agent Reinforcement Learning", IEEE International Conference on Robotics and Automation (ICRA), 2023. [PDF] [Video]
[In preparation] M. Kouzeghar, Y. Song, M. Meghjani, R. Bouffanais, "Multi-Agent Reinforcement Learning Enhanced with Dynamic Communication Protocols".