Wednesday, May 21, 2008

5-21-08: Inrix

While listening to a podcast from today, I heard about a company called Inrix. It heralds itself as the "Leading Provider of Traffic Information," and is trying to partner up with some GPS companies to take route finding to the next level: using real-time traffic information.

I think this is pretty awesome. I'm looking forward to a time when even local traffic is closely monitored to make finding the quickest route from Point A to Point B a snap. Ideally it would give you two options if one route takes less time but has a longer travel distance and the other route is the opposite. They have that option now in most systems, just not with accurate, real-time traffic information.

I'm pretty sure Inrix utilizes a lot of data mining to provide traffic information. It uses mountains of historical data, then takes into account factors like time of day, day of week, and probably weather and season and God knows what else to calculate the best way to get where you want to go.

I work with algorithms all the time, though I myself don't design all of the algorithms. I mostly analyze and test them. But they do fascinate me. I like to figure out ways to use them to make life a little more efficient. I think the most interesting part is just breaking down different aspects of one's everyday life and seeing what would make things a little easier, or at least a bit more fun. Some things are just plain ol' mundane and could be done with less energy. The morning commute is one of them.

I've posted about this specific topic before, and am reminded of it almost daily. The other day I was heading over to meet a college buddy of mine who is starting an internship at my company. He just found a place near my office. I knew vaguely where it was, but I plugged in the address on my Garmin GPS (mostly out of vanity) just so I wouldn't get lost. Sure enough, it took me down some of the main roads, where I ran into traffic. I then drove by a street I knew full well would have saved me 5-10 minutes of time by cutting off most of the traffic. It wasn't quite a direct route, but if the Garmin had taken into account live traffic, I could have saved a few minutes.

Writing that last paragraph sparked a thought: what if every one had one of these? Would the fact that everyone would likely be looking for the faster route disrupt traffic conditions? You could theoretically have a nearly deserted main street and crowded side streets, with people switching to and from the routes. It could be chaos.

But then I thought that, if you factored in historical data into the equation (in what's called a "voting algorithm" in the business), you could avoid the problem. My other thought was that local traffic might not be as big a problem since most people probably wouldn't use GPS once they learned how to get to and from work. GPS systems are used more frequently for unfamiliar travelers.

I would also hope that the system would have ways to avoid the "game theory" sort of scenario described above, where routes would be changed to address traffic from people using the same system, which would change routes again. It's similar to pilot-induced oscillations, where you would constantly be lagging behind in trying to stabilize the system. Perhaps it would simply take into account a "convenience" factor that would stick to one route even if it weren't the 100% best route if it meant not having to worry about changing routes.

What I envision is a system that would use one of two (or even both) methods. The first would be sort of like the Nielsen Ratings system, in that it would randomly use live data gathered from different vehicles and calculate traffic congestion that would be available to all drivers. I think this has some really cool upsides, but many downsides. First, the amount of processing power required is immense, making it almost not worth it. I'm sure within 10 years this won't be a problem as processing power increases so rapidly, so maybe it won't be an issue. Then there's the invasion of privacy issues people would have. The Big Brother fear of being tracked scares a lot of people, which is understandable. But, it would be pretty cool.

The other method would be one where sensors are installed at intersections and on roadways that detect traffic frequency. It would remain anonymous, which might help with the Big Brother problem, and such sensors are already in place. If the police can track people who run red lights, why not let them track traffic conditions? That's poor logic, but I'm sure there's data just like this being collected right now. My guess would be that this system simply needs to be expanded to roads that are susceptible to high traffic conditions.

I think the first system is the ideal one if it were possible to remain anonymous, since less equipment would need to be installed on roadways. In some places it's simply too difficult to install the system, and maintenance might be a problem. The downside is that some roads might not be worth the effort, since if a road is infrequently used there wouldn't be a point in either installing the sensor equipment or taking a "random sample" of a couple cars. There'd have to be a way to keep individual cars anonymous. The second system would likely be the easier to use right now, since such equipment already exists at many major intersections. I think for now we could just stick to what Inrix is doing and use historical data, but we can always dream.

By the way, these posts are time-stamped with dates and all that, so if you do plan on stealing my idea and getting it patented, good luck. Or, if you have already come up with this idea and have yet to patent it, my services are available for a nominal fee.

No comments: