Conceptually, I think it would be possible to improve automotive traffic efficiency by relaying optimized route information to mobile subscribers.
Optimized route information could be easily created by collecting historic data on users commonly used routes to and from work. With these common routes you can extrapolate one to a few alternative possible routes that could be advised to the subscriber which would minimize their work commute.
The way this is accomplished is by collecting and analyzing the entire pool of user routes and forecasting which mix of routes will lead to the least amount of congestion. Trade offs are made between using direct fast access routes and slower alternative routes with more stop and go. Ulitmately the idea is to decrease congestion minimizing wait time and conserving fuel.
Motivation for several routes comes from the ability of single users to remember a few alternate routes.
And elaboration could be to give step by step directions in car, during the commute.
This however poses problems in the area of driver concentration and response times.
With perfect knowledge (impossible) of all traffic conditions and vehicle locations, highly efficient routes can be determined.