Unified Fire

Wall Collapse Detecting FPGA Hardware
Streamlit app for house visualization, floor plan info, and chat interfact
Streamlit app with accelerometer data, water flow rate calculation, and nearby water bodies distances based on address

Inspiration

The world has had their eyes on the California wildfires for some time now – it got us thinking about the fire response in general. We talked to a firefighter about standard operating procedures when responding to local emergency fire calls, and learned that a significant amount of people and time are devoted to attempting to quickly learn structural information about the house that’s on fire. Combined with the fact that modern homes are increasingly being built with more flammable materials than conventional wooden homes, firefighters face an increasing risk every time they step into a home.

A firefighter’s first priority is safety, for both themselves as well as anyone within a building they are attempting to save. Time and human resources have to be spent surveying the outdoor structure of houses, decreasing the effective direct fire-fighting capacity of crews. The current approach uses technologies in a silo, and educated guesses at best for structural information about load-bearing walls. This leaves a lot of valuable data unused and inaccessible due to the siloing, while also putting firefighters at risk in case of a wall collapse.

In the heat of the moment, firefighters need to operate with maximum knowledge of their situation, from the house’s windows (to allow ventilation of smoke/gases), to the floor plans (to prepare them for which location to target first), to the structural integrity of load-bearing walls (to prevent sudden collapse and further injury), and the ability to know it in a short amount of time (i.e. on the way to the fire site from the fire station). They also need to have one of their most valuable resources, water, in constant supply – this can be challenging in rural environments, where fire hydrants may not be readily available.

We seek to deliver a solution to enhance the fire response with a new technology stack.

What it does

This solution briefs firefighter crews on the fire situation so they don’t learn these details too late.

To reduce the amount of time and resources spent surveying homes for structural components (like presence of windows, etc.) upon arrival, our solution features a 3D model of the home constructed by data stored by the city during the building’s construction. On the tablets that many fire trucks already have ready, crews will be able to study the 3D layout of homes in transit to the scene, so that they won’t lose nearly as much time on scene. Similarly, our solution involves presenting the floor plans of the building being affected to the crew, so that they can have more familiarity with the specific layout before entering the house. One of the most dangerous threats to this safety is the potential of walls to collapse. We used custom-made hardware (FPGA boards) to help firefighters keep tabs on the structural integrity of load-bearing walls, while allowing them to keep their focus on their main priority – safety. As the firefighters move through the building, simply sticking the hardware to the wall initiates a stream of accelerometer data to the firefighting team, providing alerts when the wall is close to breaking down. This enables the firefighters to rescue people while having an acute awareness of the house’s structural integrity. They won’t be caught by surprise in an already stressful situation with this hardware.

Oftentimes in rural situations, fire trucks run the risk of running out of water, since fire hydrants are not always available. To combat this, one fire truck usually stays at the scene with the water it has at its disposal, and another truck makes runs to nearby bodies of water to collect and shuttle more. These bodies of water are not always easily found on applications like Google Maps, however. Our solution allows various water sources to be accessed when refills are too far away.

How we built it

We used StreamLit for the main UI of the solution, which makes it simple for us to display a host of data points and visualizations. It was particularly helpful in illustrating a map of nearby bodies of water for fire trucks to access in situations with no nearby fire hydrants.

We programmed an FPGA board to constantly stream its accelerometer data (which changes value when the board begins to fall) to the StreamLit UI.

We used HuggingFace API to power the chat with the Floor Plan and the relevant information about the house (i.e. the site of the fire response).

Challenges we ran into

We wanted to implement this with an FPGA. It took us some time to learn how to use HDL (hardware description language).

We ran into issues getting the UI to work seamlessly on the Streamlit app. After a good amount of tinkering. We got this to work.

Accomplishments that we're proud of

We are proud to have created a full firefighting stack to make firefighting more effective.. A core focus of our project was to integrate hardware in our design, and we’re proud of the accelerometer we were able to make.

At the end of the day, we wanted to create a product that addressed the problems we learned about from the firefighter we interviewed. By addressing vital concerns surrounding firefighter safety and making an easier way for crews to manage their key resources, we created a technology-enabled workflow that allows firefighters to do their job: fight fires and save lives.

What we learned

We learned how to create a dashboard for multiple urgencies of access using StreamLit and GenAI’s summarization capabilities. We learned custom-hardware engineering via using the FPGA board for wall-collapse monitoring and detection.

Next Steps

One next step is to encourage the accessibility of floor plans and home information relevant for firefighting to homeowners, especially starting with new constructions. We will talk with local governments so that data can be provided to firefighters during a crisis.

Another next step is to do further research on wall-breaking mechanics in a fire to inform firefighters of their risk accurately and early.

Another next step is to integrate existing thermal imaging techniques into our 3D reconstruction of the house on StreamLit to quickly find people trapped in the fire.