From Slashdot: IBM’s Patent-Pending Traffic Lights Stop Car Engines

“I’m sorry Dave, I’m afraid I can’t let your engine idle. The USPTO has just published IBM’s patent application for a ‘System and Method for Controlling Vehicle Engine Running State at Busy Intersections for Increased Fuel Consumption Efficiency.’ Here’s how Big Blue explains the invention: ‘The present disclosure is directed to a method for managing engines in response to a traffic signal. The method may comprise establishing communications with participating vehicles; responding to a stop status indicated by the traffic signal, further comprising: receiving a position data from each participating vehicles; determining a queue of participating vehicles stopped at the traffic signal; determining a remaining duration of the stop status; sending a stop-engine notification to the list of participating vehicles stopped at the traffic signal when the remaining duration is greater than a threshold of time; responding to a proceed status indicated by the traffic signal, further comprising: sending a start-engine notification to a first vehicle in the queue; calculating an optimal time for an engine of a second vehicle in the queue to start; and sending the start-engine notification to the second vehicle at the optimal time.’ IBM notes that ‘traffic signals may include, but are not limited to, traffic lights at intersections, railway crossing signals, or other devices for indicating correct moments to stop and to proceed.'”

As the first commenter says “What could possibly go wrong”.
When will people realize serious Vehicle Infrastructure Integration is essentially doomed from the start? There is no deployment path. We can have smart cars on dumb roads (or as in rail, dumb trains on smart tracks), but trying to do both is asking for failure, the coordination problems are just too deep. The environment could perhaps be more informative, but unless it is informative almost everywhere, it will not be of much use.
I will go on a limb and suggest that in the long run, at all but the most congested intersections we’d be better off with smart cars and no electronic intersection controls (i.e even dumber roads), cars which just detected their environment and navigated appropriately. This can be achieved with roundabouts and similar devices.