Human-AI collaboration in real-world complex environment with reinforcement learning

Published on .

A work-in-progress version of that paper was first accepted and presented at AAMAS, the International Conference on Autonomous Agents and Multiagent Systems, in 2023. It was later published in Neural Computing and Applications, in 2025.

This paper is the result of a collaboration between AIR, Thales, the University of Alberta and JACOBB.ai.

The paper discusses the use of a novel multi-agent simulator for defense applications, based on Cogment, integrating human and AI capabilities to protect critical infrastructure like airports. By incorporating human feedback and demonstrations into the training of AI agents, the system leverages both human expertise and AI efficiency to enhance decision-making and response strategies in complex, dynamic scenarios. This approach not only improves the learning process and operational performance of AI agents but also maintains meaningful human control over critical tasks.

Paper

Download Open original

Cite

@article{humanaicollaboration2025,
  title={Human-AI collaboration in real-world complex environment with reinforcement learning},
  author={Islam, Md Saiful and Das, Srijita and Gottipati, Sai Krishna and Duguay, William and Mars, Cloderic and Arabneydi, Jalal and Fagette, Antoine and Guzdial, Matthew and Taylor, Matthew E},
  journal={Neural Computing and Applications},
  pages={1--31},
  year={2025},
  publisher={Springer}
}