E L E C T R O S P H E R E    Issue 1.08 - December 1995

Ground Truth

By Tim Barkow

Never get lost again, as geographic information systems usher in the era of intelligent transportation technology - smart cars.



Here I am on the outskirts of San Francisco, holding a two-page fax of directions, complete with street names and distances down to one-tenth ofa mile, and I miss my turn. Drive right past it. Just keep on rolling through the suburban sprawl of business parks as I peer through the windshield for a sign. Each block offers another array of anonymous façades sprouting from landscaped hedges and manicured lawns, each building is marked with another cheap plastic sign. Everything in Sunnyvale, California, looks the same. After six blocks of backtracking, I find my destination, lock the car, and try to suppress my embarrassment at having lost my way using the most detailed directions I've ever had.

Not 10 minutes later, I'm standing before a great green land yacht, Oldsmobile's model Eighty Eight. This particular Eighty Eight is promising to change navigation forever. Never get lost again - the phrase smells of a deal spun from straw on the back streets of Silicon Valley. But it's true - this quintessential American car has been reborn.

Inside the Eighty Eight, a small LCD is perched in front of the radio on an adjustable metal stalk. It serves as my interface to a digital map that mimics "ground truth" - the exact geodetic lay of the land - with the precision to route me to any destination in any city I choose. My pick: Sebastopol. I've never been there - never even heard of it until about five minutes ago. But in a few seconds, the computer is ready to go. All that's left is to follow the directions and try not to hit anything on the way there.

Buckling into a couple of tons of Detroit's finest and letting a route-guidance computer play backseat driver is a heady prospect. That street back in Sunnyvale? The computer wouldn't have missed it. Audible prompts would have alerted me to watch for the turn, the LCD would have displayed my error as I passed. Let's face it, I am outclassed by a tiny computer the size of a boxed pie - powered by a Motorola 68000 CPU and packing an 85-Mbyte PCMCIA card that carries detailed street maps of the whole of California. After missing that street sign, that one small but critical piece of data, it's now painfully clear that I am not quick, perceptive, or knowledgeable enough to drive in Sunnyvale. And that's exactly what the people at Navigation Technologies want me to think.

Started in 1985 by Russell Shields, Navigation Technologies has grown into one of the premier suppliers of digital-map databases in the world. NavTech, as the company's commonly known, is tackling a daunting task in the age of the high-tech information system: providing content. It is building a database of the road networks across the US and Western Europe that just may be savvy enough to crack open our Euclidean notion of location once and for all.

The database technology that fuels this navigational dingus is the offspring of GIS, or geographic information systems. GIS was originally developed by the US Department of Defense and found a home guiding missiles around the globe. In essence, GIS turns a map into a database, providing the foundation or digitally constructing and manipulating spatially related data sets: Zip codes, forest inventories, buying patterns, development plans. If it's got a location, you can map it. You can add as many different data layers to the map as you like. You can analyse it in a hundred different ways and, in conjunction with a GPS receiver, interact with the data in real time.

In 1984, GIS-based navigation systems were already being used in marine and aeronautic applications when Clifton Haley, the chair of Budget Rent A Car and a private pilot, approached Russell Shields convinced that the mapping technology would soon be used in navigation systems for cars. After conducting some informal studies, Shields also be-came convinced that the technology was right for the road.

But industrial developers of prototype navigation systems, the Motorolas and Rockwells, were having their problems. Before 1985, the only major player in the digital-map business was Etak Inc., now owned by Rupert Murdoch's News Corp. Born a venture-funded organisation, Etak had its fingers in a number of different map products, but many developers didn't feel the company was concentrating on the particular type of map they needed. Notably, Etak was not building map databases with the necessary detail and accuracy to mimic "ground truth."

This was a problem. If a navigation system is supposed to guide you across town, you'd better be sure that every particular of every street is laid out correctly in the map. Otherwise, you might find yourself staring up the business end of a busy highway off-ramp. Without this detail, you can display the map onscreen and overlay the car's GPS-calculated position, but you can't compute a route. What you've got is an extremely expensive novelty, a rich geek's road atlas. But with the depth of an extensive database, that little navigation system isn't just another toy, it's a damn effective tool for getting anywhere.

Heading north from Sunnyvale, I am poring over the offspring of some 40 years of previously top-secret, Cold War technology, the fallout of intercontinental ballistic missiles, global-positioning satellites, and a few billion dollars spent in the national interest, all wrapped up in a neat little package that's going to find me a cash machine and a steaming hot cup of java in Sebastopol, California.

This particular manifestation of technology transfer is Oldsmobile's GuideStar, a stitching together of Rockwell Automotive's Pathmaster hardware, Zexel Corp.'s software, and, at its heart, the NavTech database. GuideStar is no prototype - it's ready to roll. Oldsmobile will offer the system as a US$2,000 (£1,250) option in Eighty Eights nationwide, beginning in 1996. In Germany, BMW has offered the Carin navigation system, powered by the NavTech database, in its 7-series cars for almost a year now. For those who didn't get the database with the car, Philips recently began selling an aftermarket version of the Carin.

The system's design is reminiscent of the Newton and the GameBoy, only more squat and businesslike. A small patch antenna, 2 by 2-inches square and flat like a smashed biscuit, is mounted on the rear deck - the only visible sign that you're packing a global-positioning-system receiver. The GPS signals continually update the car's position on the map. When the signal is blocked by tall buildings, trees, or overpasses, a dead-reckoning system fed by positional and rotation sensors seamlessly takes over, relaying changes to the computer.

Of course, you have to be moving for the gizmo to be much use.

My first hour with GuideStar is less than spectacular. Crammed in midafternoon gridlock on US 101, I bide my time playing around with the interface, switching between the easy-to-read direction screen - a big yellow arrow indicates my next turn, an agonising 17.3 miles away - and the map screen, a bird's-eye view of my location adjustable from 220 yards to 4 miles across. The map view is much better for killing time. The only thing missing is a bit of television. And though GuideStar sticks to providing directions, BMW's Carin system can also provide up-to-the-minute television news on local traffic conditions.

Once released into moving traffic, though, GuideStar does its duty, relaying route information in its simplest terms: distance to the next turn, street name, direction of the turn. This visual data is supplemented with audio prompts that break the silence in synthesised tones, a voice ushering in the age of the intelligent transportation system - the smart car.

The navigation system is deceptively simple. Putting yourself in GuideStar's hands feels as if you've dispensed with depth perception and you're cruising around with one eye closed. Jim Gilmer, NavTech's communications industry director, says that at first most people are confused. "If someone were to describe it to you before you experienced it," says Gilmer, "you might not understand - simply because you have no basis for thinking about it in any other way." But most drivers come around. All it takes, after all, is the admission that a map database knows more about the road than you do.

Asked what it is that excites him about this massive project, NavTech chair Russell Shields answers simply, "To change the way people interact with the geometry of the world. There aren't that many opportunities for those of us who come out of the software and technology industries to make that kind of change."

The 55-year-old Shields seems to be perpetually in the right place for change. In high school, he worked in a bank that had just installed its first computer and ended up going to college already familiar with hardware and software. In graduate school at the University of Chicago, he ran the data-processing operations for the National Opinion Research Center and later the American Hospital Association. He completed all the work for a PhD in computer science, but had a little trouble finishing off. Sounding a little ahead of his time, Shields explains: "The first credit-bureau systems I did had about 10 million customers, while the professor was working on the alumni file, which had about 100,000 people. We never quite saw eye to eye about the theory on large databases."

Shields never finished his thesis.

Instead, he moved on and in 1981, with cellular pioneer Marty Cooper, founded Cellular Business Systems, a firm that provides billing and processing services to the mobile-phone industry. Shields has consulted on numerous database-related projects, including one for the CIA involving storage and retrieval of data.

These days, Shields is a busy man. Standards are being worked out for smart cars, and as those in the computer industry have discovered, helping set those standards can make your company's future. Shields is leaving little to chance on that front: he is a director of the Intelligent Vehicle Highway Society of America, chair of the Transportation Research Board's committee on communications and member of the board's intelligent-vehicle highway systems committee, and chair of the map database standards committee of the Society of Automotive Engineers.

No less impressive than Shields' résumé are his investors. NavTech's backers represent a good portion of the growing roster of international players in the emerging intelligent-transportation systems industry - Philips Media, which holds a majority stake in the company; Motorola; the American Automobile Association and the Automobile Association-UK; Nippondenso, Zexel, and Nichimen Trading Company in Japan; France's Renault; and Italy's EL.DA - each one interested in grabbing a piece of what US Transportation Secretary Federico Pe[^]a expects will grow into a $200 billion (£125 billion) industry by early next century. Though NavTech doesn't expect to break even until 2000, its investors seem willing to plunk down more than $200 million (£125 million) to reach this goal, helping to expand and improve the map database's coverage and accuracy along the way.

Creating the sort of extensive database NavTech has in mind turns out to be less about business savvy and more about brute force - that is, manpower and cash flow. NavTech employs almost 800 people worldwide and has established more than 50 field offices, all working to weave together their separate squares of the database into one massive patchwork model of the road. Half of its employees are toiling on the American database, the other half are working on Europe, most from NavTech's German subsidiary, European Graphic Technologies.

Between the two halves of the operation, NavTech offers a single global standard for automated navigation. Drive from London to Paris - or ship your car to New York for that matter - and all you need for local navigation is to slip the appropriate NavTech database into your car. The field offices, set up throughout Europe and the US, strike up the local relationships which are crucial to gathering accurate roadmaps. But all are also linked into NavTech's global networks, which bring the local maps together into a vast, global patchwork of roads.

Madeline Heiser, a NavTech research analyst, lords over the first stage in the database's development, the research and collection of data. Her library is a maze of white metal cabinets, full of maps and photographs from almost every imaginable source. Even in the 1990s, it seems, it's difficult to get your hands on reliable information. US Geological Survey quadrangle sheets, each charting about 70 square miles, provide the line geometries for the US database, but they're usually old and, because of that, inaccurate. To fill in gaps and update the old information, aerial photos are collected, mostly off-the-shelf to hold down costs. The photos also help provide information on z-levels (the relative height of roadways) that distinguish overpasses.

NavTech rarely uses commercially published maps because of the expensive release fees. So the researchers have to dig. Maps from water and electric utilities are often good sources. Many of the field researchers have connections to their alma maters and former professors who help ferret out hard-to-find information. And, sometimes, when off-the-shelf cartography is the only option, the company buckles down and forks over the cash. NavTech recently purchased 13 maps of areas in New Jersey for $130,000.

Other times, resourcefulness and ready cash are not enough. Heiser recounts the tale of the village of Long Grove, outside Chicago, which wouldn't release any of its maps because, being a well-to-do community, it was worried about its residents' privacy. "You let Aykroyd and Belushi in for The Blues Brothers, you can let us in," Heiser protested. But they disagreed. Madeline Heiser, despite her charm, is no John Belushi. So, NavTech did it the hard way. The company sent out two people to drive the streets and record the entire area. It took two weeks of 10-hour days to finish the job. "Sometimes," Heiser says, "you've got to do the whole thing by hand."

After collection, the maps have to be digitised. And while it might be more efficient to scan the maps, NavTech prefers to control which data it inputs. Scanning the database is quicker, but you can't be sure whether the topology is accurate. If it's not, then roads that actually intersect won't connect, and the computer won't route you correctly. So, again, everything is done by hand.

The digitising is done by teams of geo-coders using Environmental Systems Research Institute's Arc/Info, the leading GIS software package available. Digitising is painstaking work. In San Antonio, one quad sheet took a week to input. Twenty quads may take two to three months. One section of downtown Philadelphia had 22,000 "links," a term used to denote a road length or an address range along a street. That was the biggest quad the geocoders have come across so far.

Once the basic road geometries are input and their topologies checked, the database is handed over to the geocoders. This is where the love is added. Everything from city-, county-, and private-area designations, construction information, street names and aliases, the text of road signage, z-levels, and turn restrictions is entered into the database layer by layer, slowly sedimenting into an intricate data mesh, a virtual reality of the road.

GuideStar demands this kind of detail. It's got to know whether or not there's a concrete divider along a highway, whether two streets cross or one's an overpass. Often, the data is more complicated: Jug handles, left-hand turns that loop to cross traffic from the right lane, are a fairly common feature in New Jersey, for example. Then there's I-66 in Washington, DC, which between 4 and 6 p.m. is legally open only to cars carrying more than one passenger or headed to Dulles airport. Every sign, every painted line along the asphalt, every relevant piece of information has to find its way into the database. Each link normally ends up carrying three to four dozen different pieces of data.

The power to preset "points of interest" - banks, rest stops, restaurants - becomes clear once I arrive in Sebastopol.

In real life, an ordinary mortal is not very likely to know the automated teller closest to any given spot, but with points of interest, the navigation system can figure it out for you.

That's the beauty of a database: being able to quickly and effortlessly reshufße information into its most useful form. With GuideStar, I am able to sort through teller machines by distance, find the nearest one maintained by my bank, follow the steps - turn right, right, left - and in five minutes I am liquid again. But collecting these listings is not effortless. The eight-person, metro points-of-interest team inputs information on categories that include gas stations, cash machines, grocery stores, tourist attractions - the list is enormous. And while the team will probably be overwhelmed in the future with lists of locations sent by the businesses NavTech includes, for now it has only one source - the phone book. The team's typical modus operandi? Wham! Flip to the right section and get typing.

As a section of the database nears completion, when it is as accurate as the digital coding process, sources, and computers can make it, it still may not be accurate enough. There's only one way to be sure the database is going to work: the win one for the Gipper way - get out there on the double yellow line and drive it.

NavTech has sent out 12 teams this year to verify and update sections of the database. Picture an army of ants driving rented cars, fanning out across the country to pick through the maze of alley-ways and off-ramps looking for that one small bit, that crumb, of information they missed.

The company has dispatched crews to check the metro databases of Phoenix, Albuquerque, Seattle, Dallas, Houston, Minneapolis, New Orleans, San Antonio, and Austin. Meanwhile, the separate highway group maintains an ongoing checking process. When NavTech set out to check the Southeast, the company sent 30 people to ride the roads of Miami, Orlando, Atlanta, and the highways in between. That operation took three weeks.

Out on the road, checking the database is a slow, plodding process. In large metropolitan areas, it isn't possible to drive every link in the database, so the metro teams focus on proper addressing, one-ways, and turn-restriction information. "They're after more of the nitty-gritty stuff," says Greg Beaulieau, a geographer in the highways group.

Beaulieau has recently returned from Texas, where he and a co-worker drove 4,000 miles of highway in 11 straight days, working 12 hours a day. Armed with printouts of the database and a digital copy on a laptop, the two wandered along the road, stopping periodically to shoot every sign they saw - using 67 rolls of film - gathering information for the database.

"We went out there and drove it kind of prematurely," Beaulieau says. "We just had raw, naked geometry to go on, and 1 in 20 times we were just looking for recognisable shapes." The process redefines boredom. "We've got a binder full of screen copy in the order that we're going to drive it," he explains. "'Coming up we should have Jones Road.' Sure enough, there it is. We get off, take a photo of the Jones Road off-ramp, and on the off-ramp we take photos of all the signs that tell you where you can branch from that off-ramp. Then we approach it from all possible angles: from the north, hit the cross streets, and then the south. We've got every possible sign for any given interchange off any freeway in the whole country."

Navigation Technologies guarantees its database to be 97 per cent accurate. But it all hinges on data freshness. That's a problem NavTech is not sure how to surmount. The electronics and auto industries sell durable goods: what you bought two years ago works as well today as when you bought it. Their production and distribution systems are tuned for this particular business. A database, however, is a different animal, or rather, vegetable.

Information is perishable. A two-year-old database will not function as well as it did when you bought it. You expect this from an old map, and you work around its errors. But allow a navigation system with an aging database this luxury and it won't get you there. NavTech operates on the assumption that information in the road networks - such as addresses and street names - changes 10 per cent in a year. After two years, you've got only an 80 per cent chance of reaching your destination - 20 per cent of the information is wrong. So how do you get the database to the consumer more quickly?

It's a question the electronics and auto industries have not been able to answer. In most cases, it takes three to six months for a component to move from the factory to store shelves. Throughout that process the database becomes older, and you risk selling a product that is already out of date. Even more difficult, Shields explains, "is that the database is not something you sell as a national product - it's too large. It won't fit on a CD or any other media. And it's regional. Handling the distribution of what are different, regional products is something these industries haven't been able to handle."

Once you've bought a navigation system, the problem becomes more complex, giving NavTech an opportunity to develop lucrative repeat business. Database up-dates have to be administered to customers in the field. Perhaps you'll be able to do this at your local service station: $39.95 (£25) for oil, lube, and download. But for now, NavTech, along with almost everyone else in the industry, is waiting things out, watching for the arrival of the next stage in the communications revolution. The company expects that, in the future, customers will be able to hook the database's hard drive to their PCs and download updates from the Internet or a modem-to-modem connection. But the great connectivity hasn't come yet.

Meanwhile, NavTech should have plenty of time to surmount the larger hurdles that stand in the way of mass-market navigation systems. Shields figures that even 5 per cent penetration is a ways off. "I'd love to see it," Shields says, "but you've got 200 million vehicles on the road in the US and Canada, and an equal number in Europe. Five per cent of that is 10 million cars. Even matching the cellular model, that's a growth rate you won't hit until around 2004." So Navigation Technologies has strapped itself in for the long haul, riding out this uncertain stretch as industry and consumers adapt to and adopt the new digital "everywhere."

The map database is a powerful tool. Although its knowledge extends to every street and address over a large area, each destination a driver chooses, each computed route, generates a new map reduced to solipsistic specificity. Point A to point B and there the map ends. That's all you need to know.

It's pointcasting, if you need a media analog. Old maps, the paper ones, were broadcast, revealing a wide, general subset of the data contained in an area, say, road networks or topography. In the new pointcast media, all destinations exist within the database, but in choosing one, you pull only the information you need. Suddenly you are equally able to get anywhere. The only difference between destinations is time. Who knows what effects this ability will have on our collective sense of community? Navigation systems are far from ubiquitous. But the implications seem clear. You are now the centre of your own mobile universe, you choose your reality through the places you go. The places you avoid effectively cease to exist. Location is dead.

Rolling along the highway in the Olds Eighty Eight, I can sense it. I've been taking directions from a computer for only an hour and my apprehension has subsided. It's comfortable. I have no idea where I am at any instant, and I don't really care. Everything I pass becomes backdrop, even the dirty parts. In an odd sense, it's filler - the meaningless time, the meaningless space I pass between the coffee shop and the deli, the museum and the mall.

When I asked the NavTech staff about charging businesses for inclusion in the points-of-interest layer of the database, they hemmed and hawed, citing a survey done in April by the Electronics Industry Association that ranked Yellow Pages-style information last in features consumers wanted in a navigation system. Nonetheless, you begin to project: once we're all navigating our way around cities and towns by plugging our destination into a computer, will businesses be able to survive if they're not in the database? A consumer looking for a bank or a grocery store is making a decision - and very possibly a purchase - from the driver's seat of his or her car. If your business is not in that database, you're nowhere.

It's hard to believe that NavTech hasn't expended much thought on exploiting this. But maybe their present indifference is warranted. The company, along with the entire industry, must deal with finding its consumers first. So far, only the US rental-car market and European luxury carmakers like BMW and Mercedes have embraced the use of navigation systems.

But this is changing fast. At the Frankfurt Auto Show in September, almost every automaker was at least talking about navigation systems. Beginning next year, Oldsmobile will launch automated navigation into the American market. And these systems will sell: to gadget freaks, backwoods ramblers, the safety conscious, and then the rest of us.

"Do we need navigation systems?" somebody recently asked me. "What's the appeal?" Well, we've all got microwaves, now don't we?

Someday soon, it's very likely that your local auto dealership will ply you with, in addition to bad coffee, pamphlets advertising wireless emergency-alert systems, paging and e-mail messaging services, electronic travel guides, and a host of other applications still in development today. To one degree or another, these apps all hinge on the installation of the hardware required for the navigation system - display, hard drive, CPU. And if navigation systems prove to be a sustainable killer app, well, damn the torpedoes.

But for now, all this is just one more step taken, one more line crossed, one more shift of the scales towards reliance on the information system, the virtual reality. This time, the icon dissected and digitised is the once great open road.

Navigation Technologies' goal is "ground truth," and once the geometry is coded into the database, it's a simple exercise in sampling GPS data and shifting data points a little to the left, or maybe right, to lock in this accuracy. As we struggle to understand the changes technology is wreaking on our lives, perhaps there's some comfort in knowing that, at least in our cars, we're entrusting our well-being to a computer fueled with the truth.

Tim Barkow (tim@wired.com) is production editor at Wired US.