Twitter Facebook Google RSS
 
ARTICLE 

Modern Trainers on the Way For USAF Traffic Controllers 

11  2,002 

by Christian B. Sheehy 

A new training system for U.S. Air Force air-traffic controllers will help reduce procedural errors and expedite student qualification, said industry and military officials.

The service will spend up to $72.5 million on 94 simulators.

The so-called tower simulation system is designed to improve voice command features and increase automation. Current technology relies on pseudo-pilot mediation between controller and aircraft. The new trainer’s voice recognition technology proposes to eliminate a reliance on potentially fallible person-to-person communication links in favor of more streamlined, code-structured operations.

A growing demand for qualified air-traffic controllers is driving the Air Force to employ fully automated ATC simulation systems capable of preparing controllers to handle any airframe, at any location, under any conditions.

Regular adjustments to on-base air-wing strength—to meet changing readiness requirement—has presented challenges in maintaining sufficient levels of qualified controllers without available simulation technology on site, officials said. There is an excessively large time gap, these officials said, between initial technical schooling and actual field qualification.

Without the capability to train seamlessly from school to flight line, qualification time and safety risks are increased, as controllers can train only periodically, relying on real, as opposed to simulated, air traffic, explained Thomas L. Harris III, program manager for the tower simulation system at the Air Force Training Systems Product Group, Wright-Patterson Air Force Base, in Dayton, Ohio.

“The key is to provide a physical interface, complete with real-time sensory cues, that will ensure an ongoing training environment without the need for live air traffic,” he told National Defense. “This system will provide students with their own ‘airspace’ to gain proficiency and certification on multiple types of airframes in multiple types of scenarios.”

The trainer’s manufacturer is Adacel Systems Inc., in Montreal, Canada—a subsidiary of Adacel Technologies Ltd., of Melbourne, Australia.

Using voice recognition technology in place of current pseudo-pilot methodology, the simulator will enable student air-traffic controllers to hone their skills without the supervisory or location-specific requirements of live traffic training. Substituting software technology for human mediation between controller and aircraft, the system is expected to speed up the time necessary to qualify student controllers—in a simulation-based environment capable of replicating any live traffic scenario.

“With the current reliance on a controller to pseudo-pilot link in live and simulated air traffic scenarios, the potential for procedural error caused by human-to-human miscommunication is real,” said Gary Pearson, program director for ATC tower simulation technology at Adacel. “With voice recognition, we remove the need for a continuous human presence, and thus human error, through the integration of a sound processing system that can interpret a command much faster and more precisely than the human ear.”

In the system’s primary phase, a voice recognition engine attempts to match an ATC vocal command against a text-speech module or gram-o-file of acceptable commands. Once recognized, the command is sent to the pilot behavior system, which compares the command input with radio transmissions being received from the aircraft.

The radio transmissions are then matched with the ATC commands via the text-speech module and routed back through the pilot behavior system, which sends the appropriately matched command to the aircraft.

“This technology does in four steps what pseudo-pilots do in one,” remarks Pearson. “Controller inputs are filtered from voice recognition engine to text-speech module to pilot behavior system and then back through text-speech module before they are sent to the aircraft. Where pseudo-pilots have only one chance to hear and process a controller command, the voice recognition system has four chances to catch an inconsistency before a command is sent out.”

The software, though automated, can also receive inputs from an instructor who may want to override a controller command. By simply requesting to “disregard” a transmission, the system will delete a command, whether or not it has been received by the aircraft, and simply re-issue a new one.

Working from a database of standard ATC commands, the Adacel system comprises a series of training modules that offer controllers a variety of scenarios—from routine to highly complex. The visuals are realistic enough that they are comparable to actual flight operations.

“In keeping with the unpredictabilities of everyday air-traffic management, success in the automation of any manually-operated function depends on the software’s ability to recognize unusual variations in commonly-seen patterns,” Pearson noted. “In response to this reality, real-life variables such as weather, communications and ground conditions were built into the system to be added or subtracted per instructional needs.

“Though not a new concept, voice recognition technology has yet to become fully self-sufficient in the face of ever-changing input complexities,” Pearson said. Whether an unrecognized command or a series of recognizable commands in the wrong order, voice recognition software can only process the data that it is programmed to understand.

Acknowledging this, the Adacel approach is built around an open architecture of standard air-traffic control terminology that can be manually overridden at any time. “If an instructor wants to change a scenario in midstream to simulate a more complex situation, he or she can do so without losing the real-time flow of events,” Pearson added.

Current Training
Currently, Air Force ATC simulator training is conducted solely at the formal technical school and radar approach-control stations at. The technical training school for air-traffic controllers is at Keesler Air Force Base, in Biloxi, Miss. The radar-control stations used for simulation training are located at various installations around the world.

Once graduated from technical school, personnel manning flight-line control tower posts do not have access to simulation equipment, working only with live traffic situations.

“A gap between technical school ATC training and on-site qualification has existed, with no fill aside from removing needed tower controllers for further instruction,” said Kip Spurio, director of resources and requirements, of the Air Force Flight Standards Agency, in Capitol Heights, Md. “Without continual live traffic, controllers are often idle, unable to complete their training cycles on time.”

Under the auspices of the Air Force tower simulation systems (TSS) program, Adacel technology will recreate a virtual tower environment consisting of ground controller, local controller, flight data controller and station coordinator stations. MaxSim version 3.2 system software will accurately replicate voice communications between controller and simulated aircraft and ground support equipment. A high-fidelity visual display, 270 degrees horizontal and 34 degrees vertical, will offer a simulated, three-dimensional view of the specific airbase as seen from the control tower. An exercise instructor position also is provided, allowing for operational control and interaction with a specific scenario as required.

One of the challenges facing on-base ATC training has been maintaining a qualified core of personnel in the wake of constant trainee turnover. “With duty station reassignments occurring regularly and ATC qualifications being specific to each location due to differences in aircraft, ground equipment, and runway layout, each new trainee arriving needs to be qualified for that particular station,” said Spurio. “Without simulation available, delays in training without live traffic as well as the risks associated with live traffic training have to be confronted. Using the Adacel system, trainees will be able to qualify in advance for their new duty stations, reducing both certification time and live-traffic safety risks.”

Another source of discontent among air-traffic controller instructors has been the need for manual instruction at almost every level of the training cycle. Currently, a minimum of four people are needed to run a simulation scenario. Adacel technology proposes to decrease that number to two, requiring only a trainee/controller and coordinator/instructor to run the software.

“Technically, the student could run the system alone, however, an instructor needs to be there to upgrade variations within a given scenario,” Spurio noted.

As ATC training has gotten more complex in response to advances in airframe technologies, the demand for greater system standardization to ensure effective controller-aircraft communication has increased. “Streamlining controller command phraseology has become a necessity in accommodating both higher numbers and types of aircraft,” Spurio pointed out. “When trying to land an F-16, followed by a C-130, a controller had better make sure that each pilot receives the specific command designated for that aircraft.”

To achieve this goal, voice recognition technology was built around a limited set of recognizable commands. Once programmed, the software cannot misinterpret a command, as might be the case with a live pseudo-pilot. “An often-occurring problem in communications between controller and pseudo-pilot is the use of non-standard terminology,” Spurio says. “As soon as the controller gets away from standard code, the risk that the pseudo-pilot will misinterpret increases greatly. With a voice recognition system, there is no risk, because it will not accept what it does not recognize.”

Beyond simple command recognition is the intelligence needed to determine if a command is the most appropriate for a given scenario. “Adacel’s software is constructed around a stored database of integrated commands that work in conjunction with one another, Pearson said. “For every one command, there are a given number of matching responses. Using a built-in reasoning capability, the software is able to choose the most logical command-response match and then relay the information accordingly.”

As the foundation for the successful transmission of controller commands, Adacel’s text speech module—programmed to know what sounds make up which words—breaks down each spoken word into its component sounds, according to elements such as pitch and speed. It then recreates the words by matching them against a gram-o-file of stored commands. With this capability, the software can eliminate any extraneous sound that may be detrimental to clear message transmission.

In the case of aircraft-radio transmissions, all static noise is removed prior to controller reception, because the controller does not actually receive the original transmission but a recreated version of the same.

“In the same way that individual sounds are associated to create words by the text speech module, individual words are associated to create proper commands by the pilot behavior system,” Pearson said. “The system, knowing what commands go with which responses, takes inputs, breaks them down into their component words, and then reconstitutes the message, matching it against radio responses from the aircraft. If the system finds the transmissions do not match, it will not transmit the controller command.”

Spoken commands may not be recognized for one of two reasons: either the words themselves are not clear or the message content does not make sense for a given situation. In either case, the command will be rejected unless close enough to a recognizable match. “In the case of poor diction by the controller, the words can often be deciphered without command re-issuance,” Pearson indicated. “If the wrong command is issued by the controller, the software will attempt to find the appropriate one using the aircraft radio transmission, however, this is more unlikely due to the greater complexity of the match.”

Adacel’s technology also offers advances in screen display and visual resolution. Using a series of mounted mirrors in the rear of the simulation room, images are reflected to screen displays in front of the room without interruption by personnel movement or positioning. High image resolution is accomplished using an advanced projection system, capable of four times the pixel output of other systems. “Since each pixel only holds one color, the more you have to work with the higher the definition of each image,” Pearson notes.

In recreating a virtual control tower environment, Adacel software is also capable of producing unexpected factors that are often encountered in live traffic situations. Conditions such as inclement weather, additional runway traffic, or even aircraft malfunctions can be replicated in real-time. “As in the real world, at least half of the conditions that a controller must deal with are less than optimum,” Pearson said. “Though nothing prepares a controller for the real thing like the real thing, the software enables training for adverse scenarios that can be interjected or removed at any time.”

In recent years, Air Force traffic-control centers have been under mounting pressure to streamline both their information acquisition and dissemination procedures, to produce qualified controllers. “A big challenge is simply fulfilling the logistical supportability requirements that must be met in order to keep air traffic moving safely and efficiently,” Spurio noted. “This will help shorten the time it takes to get a controller from the classroom to ... where they are needed most.”

Along with the 94 simulation systems to be delivered under the current contract, Adacel will provide the Air Force with training system support for maintaining and updating hardware and software, including creating, updating and maintaining a host of imagery databases.

“The plan for this system is to do a thorough factory acceptance test only on the first three systems being delivered,” Harris said. “Once the testing is complete and all the discrepancies are corrected, this will become the system baseline for all other systems delivered.

The initial contract with Adacel is worth $2.4 million for three simulators and for the development of a computer-based instruction course on the operation of the simulator. The contract has annual production options through 2007. The Air Force currently is planning on procuring approximately 94 simulators from 2002 through 2007. Also, the contract contains annual options through 2011 for contractor logistic support and for a training system center, to support the fielded simulators. Estimated contract value is approximately $72.5 million if all the options are exercised by the Air Force.

  Bookmark and Share