Home
BEST is Beacon-based Indoor Fire Evacuation System. The proposed system offers the safest path based on accurate location with a user-friendly visual supplement.
When designing this system, we were adhering to the following principles:
-
High Accuracy of Indoor Localization: The RSSI values of the beacon signals are filtered with Kalman Filter, and to calculate the distance of the receiver and the transmitters. Machine learning models were used for locating the user.
-
Provideing Optimized Route algorithm in Real-Time: For dynamic individual fire evacuation path, q-learning Algorithm derives route considering the hazards and the components of buildings. The optimal evacuation algorithm avoids congestion by using multiple exits. Finally, the optimized route will be sent to each user in real-time.
-
User-friendly Navigation: The solution of a sight-blocking situation, user-friendly AR technology was adopted. In the 2D map the congestion areas and the users location were viewed and the users will be guide With a 3D Augmented Reality (AR) navigation.
There are 4 sections to illustrate Best; Localization, Server, Algorithm, Nevigation System. Below is a proposed system functional model.
To determine the user's position, beacons and mobile devices were employed. Additionally, the DNN machine learning model and the Kalman filter were applied to increase the localization's accuracy. The user’s smartphone reads the raw RSSI data from the beacon signal. Kalman filtered was used to reduce RSSI fluctuation. Based on the variation of RSSI value in a stationary status, the Kalman filter is re-initialized when a certain amount is over the threshold. As the following below, it filtered the RSSI value recursively and rearranged to yield.
Then the DNN classification model was used to estimate the current position and send it to the server. And also, suitable for fire evacuation systems which has lots of unexpected situations.
The server monitors the environment by using Raspberry Pi and ESP32. Both of this device gather temperature and humidity data from sensors(DHT11, DHT22) to detect the fire. Raspberry Pi also receives signals from beacons and keeps track of the available beacons. When the Raspberry Pi detects a malfunction beacon the cell state will change to a fire cell. The server notifies users using Firebase Cloud Messaging (FCM) once it receives a report that a fire has broken out. Based on the location of each user Sensor-based disaster circumstance and congestion are taken into account in realtime when computing the path. The optimal route for each user is transmitted to the user’s smartphone.
The previous image demonstrates how the data transmission works on each device. To sum up, here are the key facts:
-
The server gets the temperature and humidity data from the sensors
-
With both temperature and humidity data, the algorithm provides the optimal path.
The Q-learning algorithm, a form of Reinforcement learning (RL), is used by the server to provide the evacuation route. The shortest path has typically been the primary consideration when designing evacuation path-finding algorithms. Concidering the congestion and the location of the fire adds the wight of each cell and suggest the safest path to evacuate.
The navigation system employed the following methodology to effectively visualize the optimal path:
-
By displaying a user-friendly 3D AR arrow on the center, it was possible to guide children and foreign users who are not familiar to English.
-
The current position and direction were indicated through a 2D map. On the screen, it informed the evacuee the dangerous area such as congestion or fire occurred by coloring on the map.
We highly commend you to go through each tutorial for the platform that you are building for. The side pane of has it all.