a new world airport and its baggage

Aviation came to Denver mainly in the form of mail. Like much of the inland West, the first airfields served a smattering of passenger flights (both expensive and uncomfortable in the 1930s) and a regular schedule of contract mail carriers. Postal Service air mail contracts built many of the nation's major airlines—and its major airports. One of Denver's simple airstrips, Denver Municipal Airport, transformed from a mail stop to a busy airport when Continental Airlines moved its headquarters to Denver 1937, and then again when a modern passenger terminal was completed in 1946. Along the way, it had been renamed for the mayor who oversaw much of its development: Stapleton.

The Second World War, and then the post-war return, brought both a boom in aviation generally and in Denver specifically. The city's population grew by almost 10,000 a year from 1940 to 1960. During the '60s, Stapleton gained an additional runway and a new terminal as the schedule grew from under forty departures a day to over one hundred. When airline deregulation prompted a total restructuring of the industry, both by consolidation and widespread adoption of a "hub and spoke" model, Denver's central location made it an obvious choice of hub. United, Continental, Western, and Frontier Airlines all made Denver Stapleton a busy node in their growing networks.

With growing traffic, Stapleton showed its age. The passenger terminals were too small, but moreover, the field had become cramped. The runways were too close to each other for more than one simultaneous approach during instrument conditions, a bottleneck so severe that bad weather in Denver would cause flight delays that cascaded nationwide. The airfield at Stapleton could not be expanded: a combination of terrain, adjacent land in private ownership, and legal disputes with nearby residents and Adams County (where Stapleton was located) effectively precluded any runway additions or extensions. By the '80s, Southwest Airlines was rapidly expanding beyond its native Texas—but service to Denver was blocked by a lack of available gates at Stapleton.

By fluke of politics, the City of Denver was abruptly committed to construction of a brand-new airport in 1983: mayoral candidate Monte Pascoe had made a New World Airport the core of his platform. As it happened, he came up third in a race of three, but not before goading the other two candidates—including winner Frederico Peña—into signing a written commitment to see the new airport through. In 1984, planning started for the New Denver Airport. 48.3 square miles of Adams County were annexed to Denver and, in 1989, Denver voters approved a referendum in favor of the new airport by a wide margin. Located northeast of Denver on a huge plot of mostly agricultural land, many of us have been there, at least on a layover: Denver International Airport, DIA .

From the New World Order to contracting scandals to "Blucifer," DIA is an infamous airport. Some level of, well, intrigue was probably guaranteed. DIA was the first major airport project in the United States since Dallas-Fort Worth in 1974. DIA planners, later defending their efforts, noted that they had extensively studied the lessons learned from comparable US projects since the Second World War, of which they had identified only two. One of the largest public works projects to date, DIA's budget started around $3 billion and ended at nearly $5. Along the way, a lot went wrong.

This is not a full history of DIA's awkward early years. That would require a book, and a challenging one to write given that the ensuing lawsuits, political scandal, and media attention mean that many of the facts are disputed. But before we reach our main story, it will be helpful to understand some of the context in which DIA was formed.

By the time primary construction was underway in 1990, DIA was slated to be one of the largest airports in the nation. Despite that, it technically had no airlines. Political details of the project's initiation meant it went ahead without the explicit support of any of its expected tenants. United Airlines would be the largest by far, occupying all of the large B terminal, but negotiations between the City and United did not land a commitment from that airline until over a year into construction. Continental Airlines had agreed earlier, but had just gone bankrupt for a second time in a decade, and Denver came to worry that they would not ultimately be good for the lease payments on the A terminal being built for them. The C terminal, intended for all of the other airlines, remained unoccupied. In the mean time, Denver's booming post-war economy had considerably slowed, turning the airport into an almost completely debt-funded project.

The result was that the airlines, and specifically United Airlines, gained enormous leverage over the design of the airport. That itself is common, but DIA had the unique challenge that the airlines became the dominant stakeholder in its design only after much of the design was complete. While everyone disagrees with everyone else on who was at fault and for what, I think this is a fair napkin sketch: the first two years of serious effort on DIA were led by the City, motivated mostly by factors like tourism and prestige and with little thought for the operational practicalities of a massive airport. In the second two years, United Airlines came on the scene and attempted to reshape the partially-built airport to their aggressive operational requirements.

And those requirements were a formidable challenge. DIA's scale meant that it would be split between four buildings, a ticketing/baggage concourse and the three terminals. This layout allowed room for an ample number of gates. At the same time, congestion on the airfield itself had been a challenge at Stapleton and an issue that the airlines were extremely sensitive to. United Airlines ultimately agreed to operate out of DIA only if the airport could meet United's goal of a breakneck operational tempo with aircraft turned around in under 35 minutes . To keep taxiing aircraft in motion, the terminals were separated widely so that the "alleys" in between could accommodate aircraft at gates, a row of aircraft waiting for gates on each side, and still have room for two taxiing jets to pass each other. It is, in effect, something like six taxiways between each terminal. The measurements work out such that terminals A, B, and C are about 0.5, 1, and 1.5 km from the ticketing building.

The buildings are not connected at the surface (except, today, terminal A and ticketing). Instead, DIA used an innovative design in which all movement between the buildings happens underground. This allowed a far more open airfield design than at large airports with above-ground halls (consider, for example, Phoenix), enabling a traffic concept where aircraft would land on one side of the field, taxi to a gate, continue away from the gate in the same direction, and take off from the other side of the field.

The building layout and connecting tunnel became one of the project's first weaknesses. Airlines, making their late entry into the project, quickly accumulated design changes that they made hard requirements to their leases. Underground construction is notoriously difficult to modify, though, leading to an expensive set of changes that sometimes required almost complete demolition and reconstruction of parts of buildings. By the time United Airlines signed an agreement, at the end of 1991, the project was already behind schedule and over budget.

Among DIA's most interesting facets are the conspiracy theories. The decision to name the commission that organized its grand opening the "New World Airport Commission" didn't help, on account of its close resemblance to the New World Order which was a more prominent vein of conspiracizing in the early '90s. Early '90s aesthetic sensibilities were factor as well: Michael Singer's officially untitled "Interior Garden" at Terminal C , and Luis Jiménez's "Mustang" placed prominently at the approach of Peña Boulevard, have a certain eerie presence. Leo Tanguma's murals in the baggage claim area are rooted in the Mexican Socialist tradition of murals that look to the future with equal parts optimism and uncertainty. They depict themes of war and peace, of harmony and discord, and moreover the tension between them—in a way that is perhaps a bit too visceral for jet-lagged travelers.

As with all conspiracy theories, there is of course a kernel of truth. There really were strange things happening underground at DIA. The project really did have its secrets, and it really did envision a future very different from its present. On the other hand, these quiet happenings were all more pedestrian, more municipal, than the conspiracy theorists like to think. This was a new world airport, not an order. DIA's expansive basement was not intended, as some contend, as a logistical center for hundreds of thousands of bodies. It corrals something more everyday, and in even larger numbers: bags.

One of the most arduous parts of an aircraft's turnaround is the loading and unloading of baggage. The problem is particularly acute at a large field like DIA, where bags have to cover considerable distances. Consider the case of a passenger checking in for a flight out of Terminal C: the passenger will need to travel over a mile to their gate, a problem solved by the $84 million Westinghouse APM100 automated guideway transit system. 26 autonomous vehicles run a never-ending loop of the underground connector tunnel, serving stops at each terminal every 90 seconds. This system has not been trouble-free, but the APM100 was a descendant of the 1970s techno-optimist "transit plowshares program" that saw defense and aerospace contractors apply their expertise to urban transit. While this program mostly failed to improve urban transit, it did a lot for airports, and by the time DIA was underway the APM100 was a well-proven design.

The larger problem was baggage. At most airports of the 1990s, and in fact at most airports today, baggage handling is mostly manual. Belts take bags from the ticketing counters to a loading area where workers sort them onto carts. Tugs pull those carts to gates, where bags are loaded onto the airplane as its last set of bags are sorted onto a new set of carts to be towed to other gates or to the baggage claims. DIA was going to be different: airport planners worried that baggage tugs would have to drive multi-mile routes around the terminal buildings to reach ticketing and the baggage claims, and that was after the time and labor involved in sorting. United Airlines wanted transfer bags moved from one gate to another in under ten minutes to meet their turnaround goals, and in part to assuage concerns about such a large airport being a nuisance for travelers, a common theme of DIA boosterism was speed and convenience. Travelers were promised that their bags would arrive at the claim before they did.

As soon as these words were said it was known that achieving them would require an automated baggage handling system of unprecedented scale. There was just one problem: there was no such system, and no one was even planning one.

Airports are a strange form of real estate. They are almost invariably owned by a state or municipal government (or a corporation chartered by one of the two). Airlines pay fees for use of the airfield, but they also pay to lease the gates. In this regard, an airport is a bit like a shopping mall, a landlord that provides extensive services but ultimately exists to collect the rent. Airlines usually do much of the logistical work themselves, or contract service providers to do so. In the early '90s, as it is today, the norm at airports is that airlines handle their own baggage. When you fly Southwest, Southwest employees load your bags into Southwest carts that are pulled around by Southwest tugs. The airport's involvement is limited to a fairly basic set of belts, rooms, and pathways that bring bags in and out of the building.

This was the plan at DIA, to the extent that there was any plan, when United Airlines signed their gate lease agreement for the entirety of Terminal B. For the largest airline, operating from the largest terminal building, and moving bags to and from the biggest section of the ticketing and bag claim areas, the baggage system would be a huge project. United is said to have quickly observed that the airport project was already behind schedule and that little if any thought had been devoted to baggage handling. In line with norms and the City of Denver's expectations, United took on the problem themselves.

Over a decade earlier, in 1978, United Airlines had driven down its baggage handling time at San Francisco's international terminal by engaging a contractor called the Docutel Corporation. Today, Docutel is best remembered as one of the companies with a strong claim to the invention of the ATM. The company's history is a bit stranger: originally a loose spinoff of Texas Instruments called Recognition Equipment, Docutel's pioneering achievement was the invention of optical character recognition (OCR). Their principal client for this technology was none other than United Airlines, which used it to automate validation of boarding passes . The relationship between United and Recognition Equipment was strong enough that, by the 1970s, Docutel was both a pioneer in ATMs and a major manufacturer of airport automation equipment.

Docutel's solution at San Francisco involved a device they called a "Telecar." Telecars were small electric vehicles that traveled on rails carrying a plastic tub. A belt could push a bag onto a Telecar, which traveled to a destination where it tipped the bag off onto another belt. The San Francisco system was not without trouble: Docutel went bankrupt during its development, and in large part to save the project, United is said to have been involved in brokering an agreement that saw Docutel's airport division bought out by Boeing. This created the confusingly named Boeing Airport Equipment, or BAE. This BAE was unrelated to the British Aerospace BAE, a fact that has evaded a surprising number of histories of the DIA baggage scandal—although I am inclined to cut these confused writers a bit of slack. By the time of DIA's construction, BAE was in fact under British ownership by another three letter acronym, BTR (British Tire and Rubber). Boeing Airport Equipment BAE still exists to some tiny degree as a portion of Schneider Electric. In any case, despite the eventful corporate history, the SFO international terminal automation had developed into a satisfactory system that United thought was an ideal model for DIA.

United's deal with BAE specified a $20 million system that would connect all of the terminal B gates with the ticketing counters and baggage claims, all to be designed and built over the span of 2.5 years to meet DIA's opening date target. This was an ambitious program, and all parties involved admitted considerable schedule risk, but at the time BAE was by far the dominant supplier of baggage handling equipment and enjoyed an excellent reputation. If anyone could make it happen, it was BAE, especially given their similar projects at SFO and Munich's Franz Joseph Strauss airport. What was perhaps understated as a problem is that DIA's installation would be orders of magnitude larger and more complex than either of those systems, and that meeting its performance requirements would require a complete redesign of key components.

These might have already presented an insurmountable challenge even if the project had not, in 1992, been restarted. United's strict requirements for baggage handling, and the trouble they were clearly having fielding such a system on schedule, seemed to have made the City of Denver nervous. Even that was reassuring compared to the status of terminals A and C, where just a couple of years before the airport was slated to open, there was still no progress whatsoever on baggage handling. Financially struggling Continental seemed to be expecting to use whatever the Airport Authority came up with or to improvise something last minute, and it was still up in the air who (if anyone) would use terminal C. As a result of 1980s economic trends, traffic at Stapleton airport had actually begun to decline rather than grow, softening the demand airlines had for an improved facility. Ironically, this lack of interest from airlines made the City view the project as even more urgent and critical, since they would have to effectively "win over" the airlines with a state-of-the-art facility to ensure success. Maybe, then, the decision not to provide a unified baggage handling system as a service to tenants was a mistake.

In early 1992, just a couple of months after United awarded the contract to BAE for their terminal, the City of Denver developed specifications for an airport-wide automated baggage handling system and put them out to bid. Of the sixteen companies the City identified as prospective suppliers, only three responded with proposals. Ironically, BAE was not among them: in part because it was already scrambling to design the system for United, BAE concluded that it was simply not possible to complete an airport-wide system on the timeline the City required. By that point, there were not much more than 18 months before the airport's announced opening date.

Of the three bids received, I have found historical documentation of only one. A coalition of industrial automation companies, including Canadian transit-industrial-complex giant UTDC, proposed a design that was similar to the conventional system of belts, handlers, and tugs, but with extensive use of computers to speed sorting and complete automation of the tugs. UTDC described a system of autonomous electric bag carts they called Destination Coded Vehicles, or DCVs. A bit like London's erstwhile Mail Rail, baggage handlers would load bags onto a DCV and then hit a button that would send it along tracks, through the tunnels, to another sorting area near the departure gate. Each DCV would carry dozens of bags, and they were large enough to also accommodate a few oversized items. At this phase of Denver's tourist development, oversized luggage was on everyone's mind: among the new airport's amenities would have to be first-class handling of skis.

The City rejected all three bids. As best I can find recorded, the major problem was that all three involved more manual labor for sorting, loading, and unloading than the City considered desirable, and that the City doubted the ability of the systems to meet the time goals (despite vendor assurance that they could). The City of Denver strongly favored a design without any manual loading, and so far the closest proposal by far was BAE's contract for United.

Calling off the competitive bidding process, the City entered three days of reportedly heated negotiations with BAE that culminated in a sole-source agreement to extend the United system across the entire airport for $175.6 million. This messy contracting process, with competitive bidding hastily abandoned in favor of a deal negotiated behind closed doors, compounded the quickly discovered problem that BAE was highly integrated and intended to manufacture everything off-site at its extensive plant in Texas. This meant that BAE wouldn't obviously meet Denver's contracting preference rules for local labor, minority-owned businesses, etc. On top of that, a dispute between BAE and labor unions over payrates for mechanics (BAE's contract with the City stipulated a rate of $12 an hour when the union's negotiated rate for City projects was $20) led to a two-day strike by key industrial electricians and mechanics. BAE's contract went through multiple rounds of revisions, amendments, and scope changes, and the lawsuits did not all settle for many years after the airport opened.

The schedule would be extremely tight. BAE had, reportedly, been open with the City that they believed the project would take a full year longer than planned. They agreed to the contract only on certain terms: that requirements and designs would be frozen very early on, that BAE would receive first priority over other contractors on scheduling work at the construction site, and that BAE would receive almost carte blanche authority to modify the building designs to house the system.

Of course, in practice, no part of this agreement held: terminal A was only partly leased by the questionably solvent Continental Airlines and terminal C wasn't leased at all, so the City specified an extremely minimal installation in those buildings. As other airlines finally signed onto the project, they immediately demanded expansion and changes. Much of the construction was underway or even complete, so as a practical matter, it was rarely possible to modify building designs and the baggage system had to be shoehorned into every nook and cranny. A combination of poorly coordinated project management and political maneuvering meant that BAE's "priority one" status was far from agreed, and much of the installation schedule was determined by posturing and squabbling between the contractors, all of whom by this stage were desperate to finish their own part while happy to blame delays on anyone else.

Some members of the City's project team had initially been skeptical of the BAE proposal. BAE had won them over, in large part, by building a demonstration system at their headquarters in Texas that included most of the mechanical features of the proposed DIA solution. What went unnoticed by even BAE's skeptics is that the small scale of the demonstration system didn't just minimize the mechanical and reliability problems. It also required none of the complex computer logic that would become one of the project's biggest problems.

As happens in any contracting debacle, there are many places to put blame, and a lot more people looking for somewhere to put it. There is certainly some fault all around: histories of the project always note the constant change orders coming from the City, often either at the behest of the airlines or driven by last-minute efforts to control costs. An oversize ski retrieval belt in the baggage claim area was lengthened for more capacity, shortened to cut costs, lengthened after airline objections, then shortened again. An entire arterial loop was eliminated from the project for cost control, but then airlines still expected its capacity to be available, leading to modifications to mitigate the loss. BAE contends that the project would have gone much more smoothly if the design had actually been frozen as the City had originally agreed. At the same time, given the political and business considerations, the City may have been well-founded in its fear that delivering anything less than the airlines demanded would set Denver up for an inter-airport dispute not unlike that between Dallas-Fort Worth and Dallas Love Field.

Another problem was access. BAE expected to be able to go anywhere, and do anything, at any time. They believed this to be a hard requirement to complete the project. At the actual worksite, things worked differently: each of the four major buildings was under the leadership of a separate on-site construction manager, and the four managers reached four different positions on BAE's involvement in the project. Better-connected local contractors often got their way over BAE, and besides, some of the places that BAE intended to start installation work hadn't been built yet anyway. BAE identifies the problems with project management, on-site leadership, and coordination between contractors as the major cause of delays in testing. On the other hand, the development of the control software, which was completely under BAE's control back at their offices in Texas, had reportedly not even begun when the countdown to opening day ticked under one year.

Knowing that the installation proved so troubling, let's consider what, exactly, BAE was trying to install. The baggage handling system for DIA was based on the Telecar installation at SFO, but there would be many enhancements, particularly for reasons of delivery time.

When a passenger checked into DIA, a ticketing agent would affix a barcoded label to their bag and place it on a belt. The belt terminated at something like a miniature "train station," where it simply ended at the side of the Telecar tracks. When the bag reached the end of the belt, a barcode scanner read the label, which was associated in a database with the RFID tag on an arriving Telecar. To meet the deadlines involved, the Telecars really moved, with a top speed of 19 to 24 MPH depending on who you ask (probably revised downwards over time for reasons that will become clear). What made the system truly sporty, though, is that the Telecars did not stop. Empty Telecars slowed to about 4 MPH as they passed a loading belt, which sort of threw the luggage into the Telecar's bin before it tipped upwards to make sure the bag settled into place.

Because it was a project of the early '90s, BAE opted for a distributed control system based on microcomputers rather than a design based around a large system. This was a quintessential application of real-time computing, after all, an area in which microcomputers were increasingly dominating. Over a hundred 486 machines ran custom control software on OS/2, with Windows-based workstations for operators and maintenance staff in multiple control rooms.

The Telecars were propelled along their tracks by linear induction motors—a fin descending from the bottom of each Telecar acted as the "rotor" while the powered stator coils were fixed in long strips under the tracks. Much like the famed Tomorrowland Transit Authority and its airport compatriot in the parking garages of Houston's George Bush Intercontinental Airport, the linear induction drive system had the advantage of simplifying the Telecars (which, in contrast to other proposed designs, were not self-powered at all but more like carts that were pushed along the rails by the LIMs), but the disadvantage of requiring a relatively new and unproven motor technology.

As each Telecar rolled along with the magnetic fields of the LIMs, RFID readers detected the golf-puck tags mounted to their sides and a computer checked the tag ID against the database to determine which bag was in the Telecar. Based on another database lookup to determine the bag's destination, the computer decided whether or not to operate switches to direct the Telecar onto various different tracks. A main "bus" of Telecar tracks in the central underground tunnel moved bags between buildings at high speed, where they were switched onto one or more rings that took them around the perimeter of the terminal building.

When a Telecar reached the gate where its bag was expected, the plastic tub on top tipped high up on its side at just the right moment so that the bag would fall out onto a waiting belt. The Telecar never stopped, it only slowed down, to about 6 MPH when unloading. The belt carried the bag to a waiting baggage handler for loading onto the plane, while the departing Telecar joined a pool of spares that circulated around the rings until there was a bag to be picked up.

I put a lot of time into writing this, and I hope that you enjoy reading it. If you can spare a few dollars, consider supporting me on ko-fi. You'll receive an occasional extra, subscribers-only post, and defray the costs of providing artisanal, hand-built world wide web directly from Albuquerque, New Mexico.

If you think a bit about the mathematics of such a system, you can imagine the challenges involved in just the management of the empty Telecars. Loading was done by belts that quite literally ended at a cliff at the edge of the track, such that if a bag was waiting at the loading station the belt could not advance until a Telecar passed by to toss the bag into. This meant that any delay in getting empty Telecars to a ticketing counter, for example, caused the bags to back up on the belt until the agents had nowhere to put them. The entire system was like this: later analyses liken it to a deeply complex graph of queues feeding into other queues, in which a delay at any point could cause a rapid cascade through the rest of the system.

Compared to SFO, for example, the DIA baggage system was not only far larger (with nearly 20 miles of track compared to just a couple) but vastly more complex, with hundreds of gates at which bags could be either dropped off or picked up, bags moving between gates either within a terminal building or across terminal buildings, and a steady but very large flow of bags to and from the ticketing counters and baggage claims. To meet the demand, BAE provided almost 4,000 Telecars. Most of them were "standard," but hundreds were special oversized units designed specifically to handle skis while still negotiating the tight turns of the underground tracks.

Besides coordinating the movement of all the bags, the computer system had to coordinate all of the empty Telecars... a big enough part of the problem that BAE seems to have referred to the main software package as the "empty cart management system." The scale of the system also made maintenance complex, something evidently not adequately considered since one of the many last-minute change orders was the addition of a maintenance facility directly connected to the rail network so that Telecars could be repaired without having to be hoisted from the track to a dolly.

The algorithmic complexity was not unexpected. Munich's airport used a similar system based on automated carts, and it reportedly took nearly two years of testing, monitoring, and tuning the algorithm that dispatched carts to achieve reliable performance. That news either didn't make it to Denver or was ignored, since the DIA system was expected to go from zero to production on a shorter timeline than just that testing period. In the early '90s analytical solutions to the problem were unknown, and they remain very difficult today. In retrospect it is obvious that a computer simulation should have been developed to evaluate the control system. It does seem like BAE had intended to do so, but they simply ran out of time, and the software was completed so late that virtually no testing was performed before attempts at commissioning the system.

DIA had been slated to open on October 31st, 1993. At this point, only a bit over a year and a half had passed since the start of work on the airport-wide BAE system. By the time that date passed, the opening of the airport had already been postponed twice due to schedule slippage in the baggage handling system. The new date, March 9th of 1994, flew by as well. Testing of parts of the system had begun, but were plagued by electrical problems with the linear induction motors. The somewhat experimental design of the LIMs was certainly a suspect, but since most of the problems were related to tripping protection circuitry (due to voltage surges, phase imbalances, etc.), most of the blame landed on the City due to its ownership of the electrical distribution system. Filtering equipment was ordered to protect the LIMs but, having been ordered after the airport was already supposed to be open, the several month lead time on the specialty electrical equipment guaranteed even more severe delays.

All of these challenges came to a head in April of 1994. By this time, BAE was testing the full-scale system by sending demonstration luggage between scores of imaginary flights operated by major names like "Antarctic Airlines." In one of the more incredible stories in the history of engineering debacles, the City of Denver decided, without informing BAE, to invite the media.

The ensuing "open house" led directly to open season on the project. Reporters were taken into the tunnels below the terminals to observe bags flung by belts into open space, landing with a thud on the concrete floor below. These were, it turned out, some of the better-off suitcases: others were partially caught by Telecars and flung into posts at speed. Telecars rounding tight bends, which were far more common than BAE had intended due to the need to cram the tracks into already-built utility tunnels, tended to toss their contents into the corners of the tunnels where they piled up. The most unfortunate bags were "loaded" directly onto empty tracks or fell out of moving Telecars, until the next Telecar ran them over and tore them open.

The media fiasco was incredibly embarrassing to the project and to the City of Denver. Television stations ran footage of bags that were taking flight well before they reached an airplane. Newspaper reporters described loose clothing piling up like dirty laundry under the tracks. Some lucky reporters witnessed a full-speed collision of two Telecars, crushing what might have been someone's luggage in between them. Occasionally, bad sensors on the Telecars led them to abandon a bag on the side of a tunnel as if they were tired of carrying it.

These mechanical problems were, it turned out, the tip of the iceberg. Less visible to the reporters were the engineering findings of the testing, which concluded that fixing the loading and unloading process and teaching the Telecars to be more cautious drivers would take months and leave them with the bigger problem: software.

The software that controlled the system was, well, buggy. Empty cart management was a huge problem: there were obvious bugs in the software that gave way to deeper systemic problems. Logic errors around tracking Telecars and updating the database meant that empty Telecars in one terminal building were sent to another where they were needed, but by the time they arrived the computer system would forget why they were there and send them back where they came from. Moreover, BAE was far behind on their plans to integrate awareness of airline schedules and passenger volumes into the dispatching system. Instead of planning ahead, the system was reactive, and each landing plane would cause a brief crisis as empty carts were sent as quickly as possible while the bags stacked up on a belt that couldn't move.

The concept of scanning barcodes and associating them with RFID tags on the carts for routing was a familiar one in industrial environments, but it didn't work very well in the chaos of an airport—or at least in the chaos of this particular airport. Between the loading and unloading problems and software bugs, the computer system often lost track of which bag was in which Telecar, leading bags to be delivered to the wrong place. When the system was already running behind, it was difficult to get a bag picked up again for rerouting. The barcode readers themselves turned out to be a problem, they would fail to read the labels on the bags at all, forcing the system to send the bag to an unloading station to be relabeled. This happened to so many bags that the small staff, intended to handle a few exceptional situations, couldn't keep up.

The press demonstration had such an impact that DIA's opening date was delayed once again, this time indefinitely. For Denver's mayor, by that time Wellington Webb, it was a political disaster. The worst part of the whole thing was a single, simple mistake that dated, one could argue, to the original conception of the airport: there was no plan B. The automated system was just supposed to work.

By this point in the project, the City of Denver had commissioned multiple consulting firms that invariable concluded that the automated system would not, in fact, work, at least not on any reasonable timeline for the opening of the airport. It took the April press demonstration, though, to finally spur action. The next month, the City of Denver engaged German consulting firm Logplan to analyze the state of the system. Logplan intended to take a section of the installed Telecar track and run it 24/7 for an exhaustive test, intending to collect statistics to understand what problems were most acute and how to prioritize improvements. Instead, the test track performed so poorly, with so many jammed cars and collisions, that Logplan abandoned the test plan. Instead, they recommended that the City install a complete second baggage handling system of a more conventional design so that the airport could open.

On the advice of Logplan, the City signed a contract a few months later with German-American conveyor belt firm Rapistan Demag. Rapistan installed a very typical design of belts that bridged the gaps between passengers, aircraft, and a large fleet of carts pulled by human-operated tugs. A computer system provided some convenience and speed, but it was, fundamentally, the standard approach to airport luggage that United and Denver had completely rejected earlier in the project. Because the installation of the "backup" luggage system required new electrical supply arrangements and modifications to buildings, it cost about $50 million.

The New World Airport opened on February 28th, 1995, a sixteenth month delay due to the lack of baggage handling. When the airport did open, almost all baggage handling was manual, pulled by tugs both above ground and through tunnels that had been hastily fitted with railroad-style signals to work around the inability of two baggage tugs to pass in the narrow tunnels. Ever since, the story of DIA and its baggage has become one of the most famous case studies in the failure of an ambitious systems engineering project. DIA and the ABE Telecar system are the subject of so many term papers, masters theses, and textbook sidebars that it is perhaps rivaled only by the Therac-25 and the Tacoma Narrows Bridge.

Along the way, certain myths have accumulated. For example, there has been a tendency to cast the story of DIA as that of an AI ran amok. "The computer prioritized delivering bags," people say, "so it just delivered them anywhere they could go." As with most such myths, there is a kernel of truth: the routing algorithm was complex and had numerous bugs that caused nonsensical routing decisions, and mechanical failures caused bags to be deposited in all kinds of places that were not intended. Certain overload conditions caused "unidentified" bags that had lost their barcode association to accumulate, so that they were all rapidly delivered to a troubleshooting station that turned into a big pile. There was none of the intention, though, that people attribute to this supposed rogue AI. No decisions were being made: rather, a series of happenstances meant that the bags never reached the decision points. The places that they piled up had less to do with the whims of an intelligent machine than it did with the geometry of the track, which was only made to fit by the use of hairpin turns. The Telecars used rollercoaster-style grip wheels that kept them firmly on the rails, but the bags had no such adhesion and sometimes kept going in their preferred direction as the Telecar took off the other way.

Among the conspiracy theories surrounding DIA, the most prominent involve extensive underground construction. Whether DIA was built as a cover story for a deep underground military base, as a concentration camp for political enemies of the New World Order, or just in alignment with the ley lines to leave us all a little unsettled, it was impossible not to notice that throughout construction there was something strange happening underground. Once again, the kernel of truth: there was extensive underground construction, and it was controversial, complicated, and as the scandal grew it even became secretive.

DIA's automated baggage handling system never achieved full operational capability. In the years shortly following the airport's opening, United Airlines worked with BAE to complete a scoped-down version of their original airline-specific baggage handling system. The Terminal A Telecar system was simplified and isolated into a separate system that operated only between gates, which never saw much use. The Terminal B system, that exclusively served United, was used only to deliver outbound bags from the ticketing counters to the gates. Terminal C's Telecar tracks and stations, some never completed, were abandoned.

While it would be pleasing to say that everything strange about DIA is a consequence of an early-'90s industrial automation project gone wrong, that would leave out so many other stories, struggles, and scandals. The baggage handling system is an incredible story, but it is also mundane. Airports are enormous infrastructure projects, and even smaller efforts seldom go smoothly. DIA's baggage woes are joined by innumerable other problems with the airport's construction ranging from the quality of the runway paving to financial improprieties. They say that you can't make an omelet without breaking some eggs, and I very much doubt that you can build an airport without breaking some plans. I hesitate to fault a city government for swinging for the fences.

In 2005, the last operating part of the automated baggage handling system, United's outbound luggage deliveries, shut down. The Telecar system was shut down. United said that they were spending over a million dollars a month maintaining the troublesome Telecars, and it was cheaper to go back to traditional tugs. Most of the equipment was abandoned in place.

DIA was, perhaps, ahead of its time. The kind of technology required by the Telecar system, from robotics to demand-responsive control algorithms, have advanced tremendously over the last 30 years. Orlando Airport's new Terminal C, for example, successfully operates an "individual carrier system" that is substantially similar to the Telecar design. Amusingly, a smaller-scale automated baggage handling has operated at DIA for the last several years, limited to moving bags from the ticketing counters through the security screening process.

Even now, automated baggage handling is rare. The majority of airports still use carts, tugs, and a whole lot of manual labor. DIA's planned system, if built today, would still rank among the few largest in the world—but decisively behind the Siemens-built system at Dubai International Airport, at least on a volume basis. DIA might still be ahead of its time today: Dubai's state-of-the-art baggage network still doesn't meet the original delivery time targets at DIA. But, then, neither did BAE.

If DIA does harbor underground secrets, they won't be secret much longer. This year, the airport announced plans to remove the last of the Telecar equipment from the connecting tunnel and refit the space it used to occupy as a pedestrian walkway. The $300-700 million project is scheduled to start early next year, but from what I can tell they haven't announced an opening date. That's probably a wise decision.

1. IATA letters DEN. DIA is an unusual case of a major airport that is commonly known by a three letter acronym that does not match its IATA designation, presumably because the letters DEN had belonged to Stapleton until its closure, leaving them ambiguous as to which airport when DIA was new.
2. Some accountings give a value of 30 minutes, and even a lower value of 20 minutes for narrowbody jets. The exact target probably changed over the course of the project.
3. Untitled (Interior Garden) is not exactly land art, in that it is indoors, although it certainly shares some common traits. That's not coincidental, Singer was very much associated with the Land Art movement and several of his projects fall comfortably within that category. It should also be noted, though, that he had an affinity for airports: other prominent works include "Uplifted Ground" on a parking garage access bridge at Austin International. As often happens with Land Art, long-term upkeep of Singer's work has sometimes been a challenge, and DIA's Untitled (Interior Garden) was nearly removed several years ago.
4. This was the beginning of a long thread of technology and standards that is still seen in the machine-readable text at the bottom of passports and some national identification cards.