Union Conflict Leaves FAA Seeing STARS

The FAA’s Standard Terminal Automation Replacement System (STARS) project echoes those corporate ERP war stories in which the benefits of a packaged software purchase prove elusive because of the need to customize it to match the organization’s existing operations.

STARS is the new computer system for one of the air traffic system’s most intense choke points, approach and departure control. The facilities dedicated to that function are called TRACONs (for Terminal Radar Approach Control), as distinguished from the En Route facilities that control traffic outside the vicinity of a major airport.

STARS is intended to replace the Automated Radar Terminal System (ARTS), developed by the former IBM Federal Systems unit that’s now part of Lockheed Martin. Featuring open, Unix-based interfaces, STARS aims to give the FAA a system that will be more practical to maintain and a platform for future enhancements.

STARS also features an independently developed backup system, derived from a competing software package Raytheon got from its 1997 merger with Hughes Aircraft. The idea there is to provide a high degree of reliability and resiliency, since a software fault that causes the primary system to crash would never be duplicated in the backup server.

The FAA started out with the goal of buying what in government jargon is known as a COTS (commercial off-the-shelf) system originally developed for Norway that was starting to be adopted elsewhere in Europe and other parts of the world. The contract with Raytheon was signed in 1996.

But after controllers objected to the user interface—and in 1998 persuaded the FAA to agree to a series of changes—STARS became a substantial software development project, requiring more than 100,000 lines of new code, on top of 900,000 in the product Raytheon originally delivered. The original $940 million contract grew to a projected $1.4 billion.

There are two basic interpretations of why this happened, and there is probably some truth to both. George Donohue, the associate administrator for research and acquisitions who led the STARS selection process, argues that the union held the system hostage and generally decided to be a pain in the neck for reasons that had as much to do with contract negotiations as with flight safety. This forced Raytheon to go back and develop custom software to accommodate a user interface hardware more like that of the ARTS system STARS was selected to replace.

The other viewpoint, which is now the party line at FAA, is that it was a mistake not to get more input from the people who would actually have to use the system and that, despite its success overseas, the initial COTS product was inadequate for the demands of U.S. air traffic. FAA Administrator Jane Garvey has also been working hard to get past years of antagonism and forge a more collaborative relationship with the controllers union, the National Air Traffic Controllers Association (NACTA).

Whatever their merits, the changes that the FAA and its union settled on busted the budget and set STARS back about 4 years. Originally slated to begin its rollout in Boston at the end of 1998, STARS today is still only deployed in preliminary form at a couple of relatively low-volume FAA sites, the El Paso and Syracuse TRACONs.

However, judged by the revised schedule and goals agreed to in 1998, STARS is back on track—although it’s still ramping up for the nationwide rollout that originally was supposed to begin in December 1998. Deployment of the final product is now scheduled to begin this November in Philadelphia and continue until 2008, when all 173 TRACONS are to have been upgraded. However, the Department of Transportation Inspector General’s Office has warned that schedule could still prove too optimistic, particularly if a related project for deploying digital radar systems continues to lag.

The project still operates under a cloud and is the subject of regular hearings before the U.S. House Aviation Subcommittee, where skeptics, including Chairman John Mica, have pressed the FAA to prepare a contingency plan to deploy an updated version of ARTS.

At the same time, the project is winning praise. After analyzing the schedule and budget metrics, CrossTalk magazine, which is the official Department of Defense software engineering journal, found that the project was running within 3% of plan, overall. CrossTalk named STARS to its list of the top 5 government software projects, as judged by software quality. CrossTalk even found room to praise the decision to make STARS look more like ARTS, saying it will reduce the training required to implement the new system.

The irony there is that DOD decided to accept the original Raytheon workstation design, QWERTY keyboard and all. DOD reasoned that the young people it is constantly recruiting were more likely to be confused by an ARTS-style ABC keyboard than by the standard computer layout. The military co-sponsored the STARS acquisition and currently has it in use at Eglin Air Force Base, with plans to deploy to 199 facilities.

Donohue points to this as evidence that much of the last several years of STARS development work was unnecessary. He particularly takes exception to NATCA’s claim that the original design was unsafe. In addition to working fine for DOD, the original STARS configuration was successful in several European nations, including Germany. “It was manifestly not unsafe,” Donohue grumbles. “I don’t recall planes falling out of the sky over Frankfurt.”

But Steven Zaidman, the current associate administrator for research and acquisitions, says STARS has been struggling to recover from a COTS acquisition strategy that turned out to be inappropriate. He dismisses what Raytheon originally delivered as “a Norwegian system that wouldn’t work in our airspace.” In the more crowded U.S. skies, the original STARS workstations simply didn’t give controllers the tools they needed to keep up with their heavy workload, he says. As a result, he says, “we had a commercial product that we wound up turning into a developmental project, and that is a prescription for trouble.”

Raytheon is essentially through with the software-development phase, except for the code required to fix problems uncovered during testing. A “red team” evaluation performed by Mitre CAASD found that Raytheon’s historical productivity rate for C++ coding should allow it to resolve the key bugs and be ready to begin deployment on schedule.