A startup building a smart navigation device for bicycles.
Beeline provides a navigation device for bicycles. It previously worked broadly as a compass, showing the direction of the target location regardless of roads and on a route recommended by Google Maps. Our aim was to improve the device by incorporating open data sources indicating how safe a given road was and using feedback from cyclists on their preferred route.
We identified open data sets on bus routes, cyclist accident location history, road surface, road gradient, road grades (e.g. A-roads, B-roads), and dedicated cycleways, with up to 14 different characteristics. These were used to build a routing algorithm that would find the route that optimised some characteristics over others according to a few predefined cyclist profiles. For a courier, we placed a greater priority on speed and distance; for a commuter, on distance and safety; and for a tourist, safety, the existence of cycleways and gradient. These priorities were not to the exclusion of others, so, for example, safety was not excluded from the profile for a courier.
We then incorporated data from users of the Beeline app to indicate whether they ‘liked’ or ‘disliked’ a given road. Applying the characteristics in the previous step, according to the profile of the rider, we trained a random forest model to learn whether an individual would prefer one route over another. In addition, these routes were capable of being fine-tuned by the individual.
We built a routing algorithm that successfully navigated the streets of London. Compared with the current technology (Google Maps), the routes suggested were safer (fewer accidents), less congested (less sharing with buses) and quieter (on slower speed limit roads). Typically, the identified roads did not take longer to ride than Google Maps routes. When put into production, the improved Beeline device will make riding bicycles in London significantly safer.