Tremor Alert app
All good - no shaking predicted
Mild shaking predicted in 13 seconds
Heavy shaking predicted in 10 seconds
Having bid and lost on the Los Angeles Shake Alert RFP, I felt there could be a way to provide the alerting service in a less expensive and more easily deployable manner. As a mobile developer, cell phones paired with the capabilities of 5G seemed like a possible solution.
What it does
Tremor runs as a background service on Android phones. As an "edge" device it uses algorithms to measure a phone's transition from a "quiet" state, to one identified as experiencing an earthquake. This transition and processed data gets transmitted to the Tremor servers. Normally some percentage of phones are "quiet" in any given region. During an earthquake, there are no phones in the quiet state. This condition is determined on the server and the reported data is quickly analyzed to produce alerts to cell phones in areas that will be affected. Advance warnings of tens of seconds to minutes can be achieved. This is sufficient time to seek cover, and trigger automated responses like stopping trains, opening elevators, securing chemicals, etc. Each affected region will receive a custom alert on expected intensity and time of arrival, based on factors like topography, ground composition, and distance.
How I built it
We built an Android app that tracks the motion data of a phone to determine its "state" - quiet, active, or earthquake. The app is running on a 5G capable phone. It sends its data to a backend we created to analyze the information. The backend condition of all quiet phones going into earthquake state, and generates a notification to affected phones running the Tremor alert app. We wrote an iOS app to receive the notification, generate an audible alert, and present the estimated severity and time of arrival.
Challenges I ran into
Neither of us are Android developers, but we managed to get the essential service working.
Accomplishments that I'm proud of
This is a complex analytical problem running on an edge device, and we managed to get it working well enough to detect earthquake conditions.
What I learned
This is a feasible and cost-effective solution to earthquake detection. While there is still a great deal of work to make it market ready, it could be quickly deployed to all at-risk regions of the world.
What's next for Tremor
The Android app could be a lot more sophisticated, and re-worked by a true Android developer. Calibration can be performed with USGS support in earthquake simulator labs. The backend would greatly benefit from existing USGS algorithms and data, and the alerting mechanisms being developed and open-sourced under the ShakeAlert program.