We’ve been working on improving the reliability and quality of the import of street data from the OpenStreetMap (OSM) project.
Two types of traffic signals are read from the OSM data.
- Junctions where cyclists can expect to wait on average 20 seconds.
- Crossings which interrupt journeys less frequently and for a shorter period, and so a 5 seconds average delay is assumed.
The journey planner uses these timings in its calculation of the time to complete a route. The effect will be noticed most strongly in fastest routes, which should now produce a more accurate route than previously.
Currently there are almost 10,000 traffic signals in the system, roughly 10% at crossings and the rest at junctions.
We continue to receive feedback about the quality of the generated routes. In some areas of the country the main problem is the sparsity of map data. That problem is steadily going away as time goes on and OSM contributors fill up the map.
Another problem is that in some areas the OSM ways are not connected. This problem was drawn to my attention in the area of Benfleet, Essex where a circuitous route went a long way round via the ring roads.When we received feedback about this from an OSM user I was able to view that area and discovered the unconnected roads. That user has since joined up the roads in OSM, and following a recent import it is able to produce these much more direct routes.
A more complex problem however is how we interpret the data provided by OSM. The way it works at the moment is to translate way tags that match a certain pattern into an equivalent CycleStreets provision. Currently there are 784 distinct patterns, and although it works, it is becoming unmanageable. We have plans to simplify this and take better account of the transport mode hierarchy.
Until recently the way we have interpreted quietness has been pure. Basically the less busy the road, the quieter it is up to a max of 100%. Cycle Tracks and Park Paths are 100% quiet. But also in this pure sense footpaths (most of which are prohibited to cycling (but its more complex than that)) were also regarded as 100% quiet. The effect of that was that the journey planner was too eager to find quiet route along footpaths, even though it has always made it clear that those bits should be walked.
So we decided to loosen this interpretation of ‘quietness’ to include whether it can be ridden or not. Links that cannot be ridden now only have a maximum quietness of 75%.
See the full list at CycleStreets provision.