Are We There Yet?

Are We There Yet?

by Tim Brookes

             I’m still freaked out by the new generation of maps. Mapquest and Google Maps, which know exactly where I live. Anywho.com, which gives not only the location of my house but, at the click of a button, my phone number. And those in-car navigation systems such as Magellan and NeverLost, which show a little arrow proceeding down a road map and periodically murmur “Turn left,” the ghost in my rented machine.

            Do they hire people to fly over every square mile of America? Do they rent camera time on spy satellites? Or just buy lots and lots of town road maps? Surely not: there are nearly 4 million miles of road in the U.S., and heaven knows how many separate intersections. How do they work–and why do they often not work, leaving me lost and fuming?

            The personalized online mapping revolution began with MapQuest. Fifty years ago, road navigation consisted largely of going into gas stations and (a) asking for directions or (b) buying road maps. MapQuest, in fact, grew out of the cartography division of R. R. Donnelley & Sons, who made road maps sold in gas stations. It became the GeoSystems Global Corporation, which launched the MapQuest.com Web site in 1996, and it got a million hits in its first 30 days. GGC changed its name to MapQuest and went public in 1999. A year later, AOL bought the company.

            MapQuest, as most of us know it, is 34 servers housed in the AOL Data Center in Virginia, four to create driving routes, fifteen to generate maps, fifteen to handle geocoding–that is, to match addresses to longitude/latitude coordinates. (MapQuest, it turns out, doesn’t actually know where my house is–it knows where both ends of my road are, and what number my house is, and makes an educated guess.) Every day, these servers generate about 5 million maps and about 7 million sets of driving directions. MapQuest deals with a lot of data — it covers the United States, Canada, France, the United Kingdom, Germany, Italy, Austria, Belgium, Denmark, Luxembourg, the Netherlands, Sweden, Switzerland and Spain to the street level, and it claims to cover “the rest of the mapped world” to the city level.

             It’s not MapQuest that gathers the data, though. In order to know what the world actually looks like, more or less all the new mapping systems depend on two sources: Navteq, based in Silicon Valley, and TeleAtlas, based in the Netherlands. It turns out that, although Navteq and Tele Atlas do use existing aerial photos and official maps, the most up-to-date street-by-street data is gathered the old-fashioned way: by burning fossil fuels and taking notes.

            Navteq teams aim to verify or update every area roughly four times a year, noting down the GPS coordinates of new developments, reroutings and as many as 160 “attributes,” or special features, per section of road: bridges, signs, intersections, new traffic lights, Points of Interest–anything that might give us travellers a clue, or slow us down, or throw us off track. This laborious process is called “ground-truthing.”

            These “attributes” are the key to how these mapping systems know which route to suggest.–after all, the shortest in terms of miles may not be the shortest in terms of time.  Each segment of road is assessed for elements that might slow us down: traffic lights, speed limits, toll booths. These are added up in what is called a “costing,” a summing-up of how much time each possible route will cost us in hold-ups.

            This isn’t an infallible algorithm, though: the Google Maps directions from my house into downtown Burlington, based on data from Navteq and Tele Atlas, follow a route that has statistically fewer traffic lights, but one of them is a notorious five-corner bottleneck made worse by multiple railroad crossings nearby.

            In fact, the new map systems become more interesting once we start to consider and and why they go wrong. MapQuest reports that less than 1 percent of its users contact them for any reason at all, and take that to mean that their directions must be pretty accurate, but in my experience, things do go wrong–on as many as one Mapquest journey in three. A year or so ago I made a late-night drive across Kentucky and thanks to a single wrong direction from Mapquest I ended up, long after midnight, 70 miles away from my destination and heading in the wrong direction. I had to go into a gas station and ask directions.

            Reducing maps to simple directions is great for efficiency until something goes wrong. Much of the information on a map has nothing to do with the particular journey you’re making today, but it enables you to say, “Wait a moment. If I just went over a river I must have missed that turn. I’d better backtrack.”

            In a recent article in the New Yorker, a reporter who had been traveling around New York with a Navteq ground-truthing team decided to leave them in their SUV and take the subway home, but despite being cutting-edge cartographers, they couldn’t drop him off near a subway station—because they didn’t know where one was. Navteq assumes that the car is the be-all and end-all of transportation, and its maps don’t include subway stations. The guy had to go into a gas station and ask directions. By assuming we’re not very bright or not very interested in the world, the new generation of GPS maps makes us passive and rather stupid.

            Above all, when you think about it, any car journey is rather like a skeleton: the individual roads, like long bones, may be where we spend the most time, but they’re also where we need least help. It’s the joints–the intersections, turns, junctions–where we need information accurately and quickly, yet that’s where Mapquest’s directions are most likely to come up short.

            What exactly is the difference between “bear right” and “turn right”? How hard does a turn have to be before it becomes a hard left? If you go straight but the main flow of the traffic bears left, is that a right turn? And the roads at an intersection are often not roads at all–we think of them as ramps or connectors or jug-handles. They’re unlikely to have names, and even if they do, we don’t think of them as being too short and insignificant to have names, so if we’re given a name it may be more confusing than no name at all.

            As I write this I’m in Providence, Rhode Island, and to get to my hotel I had to go straight when the main traffic went right, pass under an interstate down a dark tunnel, turn right into what seemed like a parking lot, turn left down what looked like an alley and then pull a hard right into the hotel forecourt. None of those last stretches of asphalt have a visible street name or number, and I only knew where to go because the first time I stayed here I followed Mapquest’s directions, got lost and drove round and round the area catching intermittent glimpses of the lighted name of the hotel.

            Digitizing the world’s highways sounds fine, but in the end it’s always an old-fashioned analog human who gets lost. Here we are in the twenty-first century, still going into gas stations and asking for directions.

This essay was first published in the U.S. Airways magazine Attache.