[Routing] Crowdsourced costing

[Andy Deakin]

Hi All,
Our company has written vehicle tracking software 
(http://www.motortrace.com), and I have been intending to look at adding 
routing for a while, with an algorithm based on existing journey data.
We have an increasing amount of real data from users (e.g. taxis) which 
includes lat,long,speed and time. From this speed data, I was hoping to 
give an indication of travel times at different times of the day. This 
could then be used in the weighting for the routing to give a quickest time.

I have not had time to spend on this project, and wondered how useful 
this data would be to the routing/mapping community if it could be 
suitably anonomised and permission sought for its use.

Regards,
Andy Deakin

Frederik Ramm wrote:
> Hi,
>
>    all of us involved in routing are probably using more or less clever 
> algorithms to compute the "cost" for traveling along a given road. My 
> own experiements which focused on long-distance routing were quite 
> simple and simply assigned an expected speed to each highway type.
>
> However, this will only ever make us "as good as" existing routing 
> solutions based on proprierary data: we will never become "better" than 
> them (except perhaps the odd detail in our data that they haven't got).
>
> What we should really aim for is to crowd-source the costing, i.e. to 
> give mappers/users control of how likely a certain road will be selected 
> for routing, and how high the penalty is for choosing to route across a 
> given junction (at a given time of day with a given vehicle). There are 
> so many things that locals know ("I would never choose that road on a 
> Tuesday morning because you're more than likely to be stuck behind the 
> garbage collection truck", or "that motorway is only 2% longer but much 
> less likely to have snow", or simply "this is a primary road but the 
> secondary that runs almost in parallel is usually quicker" and so on.
>
> There's still some way to go until we achieve this. Perhaps we need to 
> have a separate tagging scheme or maybe even a separate data base for 
> such information, because it is perhaps a bit more volatile than 
> standard OSM data, and it is also quite likely that those that collect
> costing data are different from those collecting road geometries. But we 
> should really aim for this - a data base where every road has some 
> user-assigned cost would be ideal for routing, and computing the cost 
> based on heuristics is always second-rate.
>
> Bye
> Frederik
>
>