After hearing IF talk about this problem, we identified many possible areas to improve their process of cargo transport. Trucks were often taken while the driver left the truck for short periods of time, or stolen from a parking lot over the weekend. The area we chose to focus in was quick notification if the truck was potentially being stolen.
What it does
Our solution is two parts: The truck will not be able to start without a keyfob. If it is, it will send a text message saying a potential theft is in progress. The second part uses R to do statistical analysis.We can figure out the locations of where the cargo truck is- and if it should be there or not. If it ventures out too far from its normal path, a notification will be sent of the truck's position.' As well, we are looking at driving times. If the driver has just finished a shift (ie driving for ~8h), if their truck starts moving again, it is more likely to be suspicious activity. As well, the truck driving during odd times at night (when nobody would be taking deliveries).
How we built it
UI was made using HTML/CSS and Google maps was used for visualization. Flask was used to run our python scripts. The main algorithm is in R which is then wrapped with python.
Challenges we ran into
The data had many issues and anomalies which made processing it difficult, and threw off our algorithim. In fact, we took a significant amount of time just cleaning up the data and removing null points so it was usable. We did not have any information on what path the truck was supposed to take, so it was hard to identify a stolen truck vs one that was not.
Accomplishments that we are proud of
Being able to identify the truck that was stolen Getting the whole thing running. Cleaning up the data and plotting the trucks' movement in real time. Working so well in a team, everything flowed together almost seamlessly.
What we learned
Algorithm design and the importance of data quality.
What's next for Securesy
We can enhance the algorithm, build a working prototype for the keyfob. As well, testing this out on real trucks would provide invaluable data and experience to see what needs to be changed.