In a world where every man and his front door is connected via bluetooth, we see the untapped potential of the technology as a low energy and computationally light alternative to location determination protocols. Being doctoral students, we create the unimaginable every day. We thought we would bring a little of that with us to Hacktrain this year!

What it does

Utilises convolutional artificial neural networks to analyse and predict modes of transport based, upon the accelerometer, gyroscopic and magnetometer data, obtained from smart phone devices. This is then used in combination with bluetooth low energy multilateration, to provide a provable, accurate and pseudoanonymous mechanism, for monitoring passengers travel patterns. We use this data to facilitate retroactive, automated and verifiable ticketing, congestion management, and asset redistribution across the entire UK rail network (and beyond!?)

How we built it

We began by brainstorming a number of potential architectures. I am a strong believer in not touching a single line of code until everyone is on the exact, same page. We then distributed tasks between the four of us, with each individual tackling an area they are particularly fluent within. Ultimately, this culminated into a single, final product, called RailTracer.

Challenges we ran into

We each faced our own challenges, but easily persevered due to our level of expertise. However, one challenge was ensuring communication during all stages, even across seemingly disjoin tasks; as in the end, it all needs to integrate into one beautiful piece.

Accomplishments that we're proud of

We are extremely happy that we were able to achieve an astonishingly high accuracy (with heights of 91%) on an exceptionally small sample set. Our general and most suitable model, settled around 87% accuracy, which is comfortably industry grade.

What we learned

We learned that even if people are working on different parts of the same project, it is still possible to have fun and discuss the ideas we have for our respective components.

What's next for RailTracer

Hopefully, if we win the SilverRail challenge, we would love to work with them on developing the machine learning algorithm further. If we won overall (fingers crossed), we would love to build a strong bridge between the University of Oxford and its scholars, and the UK rail industry.

Built With

Share this project: