According to the National Highway Traffic Safety Administration, two in three people will be involved in a drunk driving crash in their lifetime. In 2015, 10,265 people died in drunk driving crashes, one every 51 minutes. Drinking and driving is a major problem in the U.S. and we wanted to take action to mitigate its harmful effects.
What it does
When a person is driving, we closely monitor steering wheel rotation and acceleration so that if the driver is drunk, our algorithm will detect discrepancies in normal driving patterns. The application will first send the driver a notification that he or she is driving drunk, and if the drunk driving persists, the system will send alerts to the drivers' emergency contacts.
How we built it
We collected data through General Motor's Next Generation Infotainment SDK using the iOS steering wheel tool that we developed. Here since we don't have actual drunk driving data, we do our best to mimic the driving patterns of a person. Next the data is used to create a machine learning model using Azure Machine Learning which will then be used to detect if a person is drunk from any data we gather in the future. In our user interface, you are able to indicate your emergency contacts as well as working as a test suite for our steering wheel application. Lastly, when we detect that a person is drunk at the wheel, we create a notification in our panel telling them that they shouldn't be driving, as well as notifying their emergency contacts that they are driving while drunk in order to have them try to reach the driver.
Challenges we ran into
We had trouble developing our drunk-detection algorithm, and it took several different approaches until we found an effective method.
Accomplishments that we're proud of
We are proud of the look of the user interface, and we are proud of the algorithms we developed.
What we learned
What's next for Drunk Detector
We would love to incorporate more sensors in order to improve the accuracy of our application.