ExitMatrix, a dynamic evacuation guidance system that uses reinforcement learning algorithms combined with mobile technology to improve performances in emergency responses. This paper will provide an overall solution whereby the pathways for evacuation will be continuously optimized through reinforcement learning, while the guidance will be provided in real-time through a mobile application interface. This includes a web-based tool for creating the building layout and Unity-based simulation environments to perform exhaustive testing. The test results show that our approach can improve the efficiency of evacuation time by a big margin compared with traditional methods. For pathway optimization, the learning model provides high accuracy, while the mobile application shows constant response times in critical situations, yielding high user satisfaction scores when it undergoes usability tests. The testing of the system on virtual environments shows the effectiveness for different emergency scenarios, hence proving the applicability of the system in real-world deployment and, in particular, within hazardous industries. Through these results, software-driven evacuation strategies improve the emergency response capability with good feasibility for deployment in developing regions.ExitMatrix, a dynamic evacuation guidance system that uses reinforcement learning algorithms combined with mobile technology to improve performances in emergency responses. This paper will provide an overall solution whereby the pathways for evacuation will be continuously optimized through reinforcement learning, while the guidance will be provided in real-time through a mobile application interface. This includes a web-based tool for creating the building layout and Unity-based simulation environments to perform exhaustive testing. The test results show that our approach can improve the efficiency of evacuation time by a big margin compared with traditional methods. For pathway optimization, the learning model provides high accuracy, while the mobile application shows constant response times in critical situations, yielding high user satisfaction scores when it undergoes usability tests. The testing of the system on virtual environments shows the effectiveness for different emergency scenarios, hence proving the applicability of the system in real-world deployment and, in particular, within hazardous industries. Through these results, software-driven evacuation strategies improve the emergency response capability with good feasibility for deployment in developing regions.

Check out:
https://www.researchgate.net/publication/393983677_Dynamic_Fire_Escape_Pathfinding_System_Using_Machine_Learning_and_Mobile_Technology

Skills: Unity · Flutter · Firebase · PostgreSQL · Node.js