How Radio-Frequency Detection and Visual Identification Strengthen Counter-Drone Defense.

The U.S. Army has awarded Xtremis a $46.8 million research and development contract to build an artificial intelligence-enabled system that detects frequency-agile drones through their activity in the electromagnetic spectrum. The effort could strengthen Army defenses against unmanned aircraft designed to evade conventional radio-frequency surveillance.

Known as Augur Shield, the program will develop a radio-frequency DevOps pipeline capable of drawing data from a network of multi-domain sensors. The system is intended to use artificial intelligence and machine learning to detect, track, and characterize unmanned aerial systems that change communication frequencies during flight, a technique that can complicate interception and electronic attack. The contract includes two option years, according to available award information.

Related News: Belgium’s IDDEA Unveils Offline AI System for Real-Time Battlefield Military Equipment Identification

A U.S. Army soldier monitors the electromagnetic spectrum during training in Germany, illustrating the type of RF awareness underpinning the Army’s Augur Shield research. (Picture source: US DoD)

At this stage, Augur Shield does not involve the procurement of a finalized, off-the-shelf counter-drone system. It is an applied research effort intended to develop, train, and test new electromagnetic detection methods. The US Army is examining how RF sensor networks combined with artificial intelligence models can improve the detection of drones whose communication methods change rapidly.

Xtremis specializes in monitoring, analyzing, and managing the electromagnetic spectrum. The company develops architectures that combine distributed radio-frequency sensors, computing resources, and algorithms to produce a real-time picture of electromagnetic activity within a defined area. Xtremis presents Augur Shield as a combination of hardware and AI-enabled software designed to monitor the RF spectrum, detect devices operating within it, characterize their activity, and exploit the resulting information. This approach builds on technologies previously developed by the company, including Advanced Dynamic Spectrum Reconnaissance, or ADSR.

ADSR is a spectrum reconnaissance system intended to help wireless networks observe and adapt to their electromagnetic environment. It relies on data collected by sensors distributed across the field rather than on a single centralized monitoring point. These observations can be compared to detect new signals, monitor changes in the RF environment, and characterize different forms of electromagnetic activity. The US Army has previously tested ADSR with soldiers from the 101st Airborne Division. In that context, the system was intended to help military communication networks detect and avoid jamming while reducing RF emissions that could be identified and exploited by an adversary. Through Augur Shield, this experience is now being redirected toward the drone threat.

Xtremis is therefore developing methods that allow sensors to collect electromagnetic signatures and models to be trained to recognize activity associated with unmanned systems. Artificial intelligence does not replace the RF sensor. It processes observations, identifies recurring patterns, and distinguishes relevant signals within a dense electromagnetic environment. The funding of Augur Shield indicates that the US Army is not only investing in additional sensors. It is also seeking to automate the interpretation of the growing volume of electromagnetic data generated by those sensors. This reflects changes in the threat environment. Drones may use different protocols, change frequencies, adapt their behavior, or be assembled from commercially available components. Research must therefore focus on methods that can evolve alongside the systems being observed.

MEGA enables fully offline, AI-powered visual identification of distant aerial threats, delivering rapid classification and key technical data directly to the operator. (IDDEA)

A comparable approach can be found in visual identification through MEGA, a technology developed by Belgian company IDDEA.

MEGA was initially created to recognize military land equipment from images or video feeds. The system combines a visual recognition algorithm with a structured database of military platforms. Its role is not limited to indicating that a vehicle is present. It is intended to identify the type of equipment, as well as its model or variant, and then provide related technical and operational information, including data on its combat capabilities, armament, and protection.

As with Augur Shield, MEGA relies on model training. In this case, the algorithm learns to recognize visual features such as turret shape, chassis configuration, wheel or track layout, armament, sensors, and the overall architecture of the observed system. The two technologies therefore process different forms of data, electromagnetic information for Augur Shield and visual information for MEGA, while following a comparable principle: converting a raw observation into information that can be used by an operator.

One of MEGA’s main characteristics is its ability to operate entirely offline. The database and algorithm are loaded onto the hardware in advance, allowing identification to be performed locally without sending imagery to a remote server or relying on internet access or cloud infrastructure. IDDEA has also developed an embedded version based on a dedicated artificial intelligence module. This architecture allows MEGA to be integrated directly into a combat vehicle, observation station, air-defense system, remote weapon station, or drone. Imagery from the platform’s sensors can then be processed locally, providing an initial identification of the detected equipment together with the associated technical information.

Building on its work involving land equipment, IDDEA is now extending MEGA to the identification of aircraft and drones. The objective is not only to confirm the presence of an airborne object, but also to determine whether it is, for example, a reconnaissance drone, loitering munition, MALE unmanned aircraft, attack drone, or another type of aerial system. The system may attempt to identify the exact model. This task becomes more difficult with small FPV drones assembled locally or manufactured from commercially available components. Two aircraft may have a similar configuration while having been modified in the field or adapted for different missions.

In these cases, recognizing the payload may be more relevant than determining the precise model. The system can be trained to identify the presence of a mine, grenade, mortar round, improvised explosive charge, or weapon fitted to the platform. This information can help distinguish a drone being used for observation from one prepared for an immediate attack, allowing the operator to adapt the response to the assessed threat level.

US Army soldiers from the 3rd Infantry Division operate a counter-UAS system during Combined Resolve 25-02 to detect, identify, and counter unmanned aerial systems in an operational environment. ( US DoD) 

The comparison between Augur Shield and MEGA therefore highlights two complementary requirements: detecting that a drone is approaching, and then determining which type of drone is approaching and what it is carrying.

An electromagnetic capability such as Augur Shield can provide an initial warning and direct other sensors toward a particular area. A visual system such as MEGA can then process imagery from an electro-optical camera, infrared sensor, observation drone, or surveillance station to classify the target in greater detail. An additional layer can subsequently support target tracking and neutralization. An RF sensor depends on the presence of detectable electromagnetic activity. A camera can provide more detailed information, but its performance depends on range, visibility, lighting conditions, and sensor resolution. Radar can detect an object independently of its radio emissions, but the small size and low-altitude flight profile of some drones can make them difficult to distinguish from background clutter.

Combining several sensors and levels of analysis therefore appears better suited to the diversity of current drone threats. Artificial intelligence can help connect the different stages of this chain: detection, classification, identification, prioritization, tracking, and response selection. Through Augur Shield, the US Army is funding research into electromagnetic detection and characterization. Through MEGA, IDDEA is developing a capability centered on visual identification. Against drones that differ widely in size, control method, mission, and payload, combining several levels of detection and identification appears more suitable than relying on a single sensor.

The same principle applies to neutralization. Upper-tier air and missile defense systems such as Patriot or THAAD may be assigned to complex or high-value aerial and ballistic threats, but lower and less costly defensive layers are required to respond proportionately to drones and loitering munitions. A layered counter-UAS architecture must therefore combine detection and identification tools with effectors adapted to the type, value, and level of threat posed by each target.

Discover more about IDDEA and MEGA here. 

Written By Erwan Halna du Fretay - Defense Analyst, Army Recognition Group
Erwan Halna du Fretay holds a Master’s degree in International Relations and has experience studying conflicts and global arms transfers. His research interests lie in Security and strategic studies, particularly the dynamics of the defense industry, the evolution of military technologies, and the strategic transformation of armed forces.