So 2014 is here. A time to reflect on some of the research projects we have been doing in the margins of 2013.
In June Mudlark received funding from the TSB Feasibility Study fund to look at tracking user travel behaviour with the view to building game logic around it. This was a bit of parallel research designed to compliment our previous work on Chromaroma by extending the game beyond the constraints of fixed infrastructure in the form of smart cards and RFID terminals in stations and on buses.
We wanted to explore how a user or player could get an enhanced history of their travel in London and beyond by monitoring their movement and transport habits over time.
We were aware of the success of the Moves app and the effortless way it tracks the different modes of travel from walking, cycling, running and transport. It does it with a near miraculous level of accuracy. We wanted to go a step further and see if we could create a more granular level of detail than is provided in the “transport” type of Moves. We believed by developing our own system we could tease out the fine distinction between: train, underground, bus and car.
Working with the School of Electronic Engineering and Computer Science at Queen Mary University we made this happen by utilising and combining sensors and data available on current smartphones – GPS, Accelerometer. We developed an Android App based on a unique algorithm that measures movement patterns and corrects errors in classification through analysis of user movement patterns over small segments of time.
Our feasibility test around East London in December with 5 test subjects using Bus, Train, Underground and Cycle, correctly identified transport mode shift of the user with an overall accuracy of 78.5%. This was an amazing outcome for something that was basically only working with the data being generated by the phone from the testers movements.
We also wanted to augment the sensor algorithm with online look-ups of live transport data. To test this aspect we did quick development that used aspects both of the Moves API and the Transport API (a service which provides user routing on the UK’s transport network’s). We took my Moves data generated on my Iphone and then ran queries on the start and end points of journeys with Transport API’s routing service. This produced remarkably accurate predictions of user journey type down to Bus route or train line.
We ran across some issues with it falsely choosing bus instead of train and vice versa. We discovered accuracy was increased by choosing the fastest journey in any list of possible routes between A and B. This would obviously not always be the case. A user may choose slower journeys and so a future addition would include a method to track likely routes based on a user’s past history of travel and how often they may have travelled that route in the past.
We came to the conclusion that by combining the App with a query of the Transport Api we could reproduce a user’s journey on any transport type in the UK with a high level of accuracy. We hope to explore this more with a future iteration of the app and also integrate some game play in the mix. Watch this space as this develops during 2014.