Multiagent Collaboration for Emergency Evacuation Using Reinforcement Learning for Transportation Systems

Reinforcement learning (RL) has been widely used in intelligent transportation systems. Especially under emergent situations, RL can explore unsafe environments and make optimal decisions to guide the evacuation process for human beings. Multiagent collaboration allows agents to communicate and interact for a more efficient exploration and evacuation process, especially when the uncertainty level is high. In this study, robot agents or drones were used to explore and evacuate from environments with different levels of complexity and with or without collaboration using the RL
$Q$
-learning algorithm. The multiagent collaboration method was found to perform better than the single-agent exploration concerning evacuation time, death counts, and reward. These results as well as future research directions are discussed in the context of emerging literature.