US military hunts mobile targets in Korea with 4 TeraOps

By Marcus Boone · August 9, 2026

High-value targets escaped detection during Desert Storm, sparking a drive for automated sensors that now track 10 priority targets in Korea from a fragile, non-air-mobile trailer.

It is a 30x8x8 foot, triple-axle, commercially available trailer. To the casual observer, it looks like a piece of industrial shipping equipment, but inside, it is the singular brain for a Predator UAV. This trailer—the Ground Control Station, or GCS—contains the pilot workstations, the cooling systems, and the synthetic aperture radar (SAR) terminals. It is an awkward piece of hardware; the FAS Intelligence Resource Program's record of the system notes that it is not configured for air mobility and requires "special handling" to load into C-130 or C-141 aircraft.

There is a profound fragility to this arrangement. For all the mythology surrounding the Predator as an autonomous eye in the sky, the aircraft itself is a shell. The GCS record is explicit: "All mission imagery recording is located in the GCS since the Predator has no onboard recording capability." The drone does not remember what it sees; it merely transmits. If the trailer fails, or if the link to the TROJAN SPIRIT II satellite terminal is severed, the vision vanishes.

A 30x8x8 Foot Trailer

The GCS is designed for a narrow, linear workflow. One trailer controls one aircraft. While original plans envisioned a single station managing two Predators—one via line-of-sight and one via Ku Band link—the record admits that technical feasibility for such a concept was not even included in "near term plans."

The human element is reduced to the Pilot and Payload Operator (PPO) workstations. These are the "primary means for providing direct and responsive" control. The PPO does not just fly the plane; they manage the sensor payload, turning the Predator into a remote-access telescope. The imagery is processed through DEMPC terminals for mission planning and data exploitation, turning raw visual data into actionable intelligence. It is a clinical process, removed from the geography of the target, conducted from within a cooled trailer powered by dual 35 kw generators.

Automatic Target Recognition

By the mid-1990s, the military sought to remove more of the human from the loop. This is where the STARLOS Program enters the record. The objective was the development of "real-time Automatic Target Recognition (ATR) technology." The goal was not merely to see a target, but to have the system identify it automatically using Synthetic Aperture Radar.

The impetus for this was the perceived failure of manual identification during Operation Desert Storm. The military struggled to engage "high value mobile targets" such as Scud-B launchers and other Transporter Erector Launchers (TELs) because these systems are typically hidden in protected sites and exposed only for a "very short period of time."

To solve this, the government pursued a system that could locate and identify targets with a "low false alarm rate" and pass that data to a weapons system within that tiny window of exposure. The STARLOS system utilizes fine resolution SAR to provide imaging across wide areas in most weather conditions, day or night. This imagery is then fed into "state-of-the-art algorithms" operating within a high-performance processing architecture.

By 1994, the program had demonstrated the ability to detect and identify six high-value mobile target types simultaneously. This wasn't done with a drone, but with a Twin Otter manned aircraft acting as a "surrogate UAV." The computing power required for this automation was significant: 4 TeraOps of power packed into five standard airborne VME chassis. The algorithms were specifically trained to counter "camouflage, concealment, and obscuration effects," essentially teaching the machine how to see through the tricks humans use to hide.

Deployment to Korea

The evolution of the GCS moved quickly from testing to operational requirements. The record notes that PM TESAR began soliciting the development of two Tactical Endurance (TESAR) ATR systems based on non-developmental or commercial off-the-shelf (NDI/COTS) algorithms.

The ambition grew from identifying six targets to ten. The list of priority targets became specific: the 240 mm MRL (Multiple Rocket Launcher) and the 170 mm Howitzer. The goal was to integrate these capabilities directly into the Ground Control Station of a Predator UAV system.

One of these integrated systems was earmarked for a specific purpose: deployment to Korea. The record lists this as part of the "counter MRL ACTD" (Advanced Concept Technology Demonstration), marking the transition of the Predator from a reconnaissance tool into a precision-strike enabler in the Pacific theater. Other related units involved in this ecosystem included the 11th and 15th Reconnaissance Squadrons and the Air Force UAV BattleLab.