Case Studies in Autonomous Operations

• Introduction
• Chapter 1 Swarm Tactics in Contested Airspace
• Chapter 2 Loitering Munitions: Precision, Persistence, and Blowback
• Chapter 3 Edge AI for ISR: Finding Needles in Noisy Haystacks
• Chapter 4 Undersea Autonomy: UUVs in Mine Countermeasures and Reconnaissance
• Chapter 5 Human–Machine Teaming at the Brigade: Decision Aids Under Fire
• Chapter 6 Autonomous Convoys: Surviving IEDs and Ambushes
• Chapter 7 Counter-UAS: Electronic Warfare Meets Kinetic Intercepts
• Chapter 8 Target Recognition in Cities: False Positives, False Confidence
• Chapter 9 Accelerating the Kill Chain: Fusion Cells and AI Orchestration
• Chapter 10 Mission Autonomy When Comms Go Dark
• Chapter 11 Adversarial Deception: Spoofing, Jamming, and Sensor Tricks
• Chapter 12 Embedding Rules of Engagement: Constraints, Overrides, and Fail-Safes
• Chapter 13 Spectrum as a Battlefield: EMS Control and Cognitive Radios
• Chapter 14 Securing the Robot: Cyber-Physical Attacks and Resilience
• Chapter 15 Training the Machine: Data Pipelines and the Sim-to-Real Gap
• Chapter 16 Fighting Together: Coalition Interoperability and Data Sharing
• Chapter 17 Persistent Eyes: Satellites, HAPS, and Autonomous Tasking
• Chapter 18 Uncrewed Surface Patrols: USVs for Harbor and Littoral Defense
• Chapter 19 Deconflicting Fires: AI for Targeting and Fratricide Avoidance
• Chapter 20 Protecting Civilians: Geofencing, Risk Models, and Verification
• Chapter 21 Sustainment at Scale: Predictive Maintenance and Readiness
• Chapter 22 From Lab to Battlefield: Acquisition, Prototyping, and Iteration
• Chapter 23 Information Operations: Deepfakes, Detection, and Counter-Narratives
• Chapter 24 Counter-Autonomy: Decoys, Camouflage, and Deception Campaigns
• Chapter 25 Wargaming the Machine: Red Teaming and Scenario Planning

This book examines how artificial intelligence and autonomous systems have actually performed in modern military operations—what worked, what failed, and why. The promise of autonomy has been advertised for decades: faster decisions, reduced risk to personnel, and effects delivered at lower cost and greater scale. Reality at the edge is more complicated. Harsh environments, adversarial interference, incomplete data, and organizational friction test even the most elegant designs. Our goal is to move beyond hype and fear, providing grounded analyses that practitioners can use to make better choices in planning, procurement, training, and command.

Each chapter presents a curated case study built from open-source reporting, operational lessons, technical documentation, and practitioner perspectives. We define “autonomous operations” broadly, spanning uncrewed platforms, decision-support systems, automated sense–make–decide loops, and human–machine teams operating across air, land, sea, space, and cyberspace. Rather than focusing on platforms as ends in themselves, we trace end-to-end mission threads: tasking, sensing, fusion, target identification, engagement authority, battle damage assessment, and post-mission learning. This holistic view reveals where autonomy adds real value, where it introduces new failure modes, and how context determines outcomes.

A recurring theme is that performance is path-dependent. Data quality, communications architecture, and doctrine shape what autonomy can do more than any single algorithm. Systems that excel in permissive conditions may collapse under jamming, spoofing, or rapid concept-of-operations shifts. Conversely, modest tools succeed when paired with disciplined TTPs, resilient networks, and clear authorities for human oversight. We therefore emphasize measurable outcomes—mission success rates, time-to-effect, false-positive/false-negative balances, survivability, and cost-per-effect—while also assessing second-order effects such as escalation dynamics, coalition interoperability, and civilian harm mitigation.

The book also surfaces the human dimension. Autonomy changes workload, trust, and accountability. Operators must understand when to rely on recommendations, when to challenge them, and how to recover from automation surprises. Commanders wrestle with compressed decision timelines and the delegation of authorities. Engineers confront the sim-to-real gap, model drift, and adversarial manipulation. Policymakers must align legal frameworks, rules of engagement, and export controls with technical realities. By extracting doctrinal implications and policy takeaways from each case, we aim to bridge communities that too often talk past one another.

Readers will notice that many “failures” are productive. Near-misses, false alarms, and aborted missions frequently teach more than clean successes. We treat them not as indictments of autonomy but as design signals: where to harden against spoofing, how to structure human-on-the-loop interventions, what telemetry to log for auditability, and which guardrails most effectively reduce operational risk. Equally, standout successes are rarely magic—they are the visible tip of investments in data stewardship, open architectures, disciplined testing, and iterative fielding with user feedback.

This is a practitioner’s book. It is written for commanders, operators, engineers, analysts, acquisition professionals, and policymakers who must make consequential decisions under uncertainty. By presenting comparative, real-world case studies and distilling cross-cutting lessons, we hope to sharpen judgment about when and how to employ AI-enabled autonomy responsibly. The stakes are high: lives, stability, and strategic credibility depend on getting these decisions right.

The skies above modern battlefields are becoming increasingly crowded and complex. Gone are the days when a single, high-value aircraft could dominate a sector with relative impunity. Today, adversaries deploy sophisticated integrated air defense systems, electronic warfare capabilities, and a growing array of unmanned aerial vehicles (UAVs). In this challenging environment, the concept of "swarm tactics" has emerged as a compelling, if still nascent, solution, leveraging the power of distributed autonomy to overwhelm, confuse, and ultimately defeat opposing forces.

A military swarm isn't just a bunch of drones flying together. It's a coordinated, synchronized group of relatively small autonomous agents, often UAVs, designed to achieve a mission objective faster and more effectively than a single, larger platform could. The underlying principle is simple: maximize target saturation to overwhelm an opponent's defenses. Think of it like a flock of birds evading a predator; individual movements are simple, but the collective behavior is remarkably agile and difficult to predict.

The appeal of swarm tactics is particularly pronounced in contested airspace, where communication and GPS signals are likely to be jammed or denied. A centralized command structure, reliant on constant communication with a single control station, becomes a significant vulnerability. If that central link is severed, the entire operation can collapse. Swarms, however, are designed with resilience at their core. Each drone within the swarm possesses a degree of autonomy, enabling it to operate even when communication with a central controller is interrupted. This decentralized control is a key differentiator, allowing the swarm to adapt dynamically to changing conditions and continue its mission even if some units are lost.

One of the primary advantages of autonomous swarms is their ability to navigate and operate in GPS-denied environments. By fusing data from inertial navigation systems with visual terrain matching, individual drones can maintain their bearings without relying on external satellite signals. This is crucial in an environment where an adversary will almost certainly attempt to disrupt or spoof GPS. Furthermore, the sheer number of units in a swarm means that even if some are jammed or shot down, the overall mission can still proceed. This redundancy contributes significantly to their survivability.

The development of AI-driven autonomy software is at the heart of these advanced swarm capabilities. These systems utilize machine learning, including reinforcement learning and neural networks, to enable drones to process sensor data, predict environmental changes, and make optimal decisions in real-time. This allows individual drones to act as autonomous agents while contributing to the collective intelligence of the swarm. For example, if part of a swarm encounters enemy radar activity, those drones can quickly adjust their course while others continue their intelligence-gathering tasks in different areas.

The operational scenarios for swarm tactics are diverse. In reconnaissance and surveillance missions, swarms can quickly cover vast areas, gathering intelligence more efficiently than a single drone. Imagine a swarm simultaneously capturing photos of various points of interest, significantly reducing the time needed to map a widespread location. This capability was demonstrated, for instance, in the Israel Defense Forces' operations against Hamas, where drones detected firing locations, enabling targeted strikes.

Beyond intelligence gathering, swarms can also be employed in offensive operations. Electronic warfare-resistant drones can deliver precision strike capabilities even when conventional systems are compromised. They can act as decoys, drawing fire and influencing adversaries to expend valuable missiles, while simultaneously identifying air defense locations. The goal is to overwhelm anti-aircraft defenses through sheer numbers, opening pathways for other assets or delivering their own payloads. This "saturation" approach is particularly effective against static or less agile defense systems.

The U.S. Defense Advanced Research Projects Agency (DARPA) has been a significant player in exploring the potential of swarm tactics. Their OFFensive Swarm-Enabled Tactics (OFFSET) program envisions infantry forces using swarms of up to 250 small unmanned aircraft and ground systems in complex urban environments. The program aims to create an ecosystem for rapidly developing, evaluating, and integrating swarm tactics into field operations, complete with an advanced human-swarm interface for real-time monitoring and direction. This highlights the continued emphasis on human oversight, even as autonomy increases.

The concept of a single operator managing a large swarm is a critical aspect of reducing human workload and enabling operations at scale. DARPA demonstrated in 2022 a swarm of over 150 drones controlled by a single operator, with projections for AI-enabled swarms of up to 1,000 drones within five years. This shift from "one operator, one drone" to "one operator, many drones" fundamentally alters the human-machine teaming paradigm, demanding intuitive interfaces and robust AI decision-making.

However, the deployment of swarm tactics is not without its challenges. One of the most significant hurdles is maintaining communication and coordination within the swarm, especially in the face of sophisticated electronic warfare. While individual drones may have autonomous navigation capabilities, effective swarm behavior often relies on inter-drone communication to share information and synchronize actions. Developing resilient communication networks, such as mesh networks, that can withstand jamming is paramount.

Another practical limitation for many drone swarms remains battery life. Maintaining a large number of drones in the air for extended periods, especially when performing complex maneuvers or carrying payloads, can quickly drain power. This necessitates careful mission planning, potential in-flight recharging solutions, or the acceptance of expendable platforms. The trade-off between payload, endurance, and cost is a constant design consideration.

From a defensive perspective, the rise of swarm tactics has spurred the development of "counter-swarming" measures. These often involve a layered approach, combining early detection systems like advanced radar and RF sensors with electronic warfare and kinetic countermeasures. For example, the U.S. Air Force Research Laboratory developed the Tactical High Power Operational Responder (THOR), an electromagnetic weapon designed to overwhelm the circuitry of multiple drones simultaneously, causing them to crash. Other solutions include high-energy lasers and high-power microwaves, which can disable or destroy drones in a swarm.

The ongoing conflict in Ukraine has provided a stark reminder of the "nasty, brutish, and short" lifespan of many drones in a contested electromagnetic spectrum. This reality underscores the importance of resilience in swarm design, emphasizing the ability to operate effectively despite losses and disruptions. The rapid evolution of electronic warfare measures and countermeasures in such conflicts drives continuous innovation in autonomous navigation, hardened communications, and adaptive swarm behaviors.

The ethical and policy implications of autonomous swarms are also subjects of intense debate. While the current focus remains on "human-in-the-loop" or "human-on-the-loop" systems, where human operators maintain ultimate control and decision-making authority, the increasing autonomy of these systems raises questions about accountability and the delegation of lethal authority. Operators must understand when to trust the AI's recommendations, when to intervene, and how to recover from unexpected automation behavior. Clear rules of engagement and robust fail-safes are essential for responsible deployment.

The future of swarm tactics in contested airspace will likely involve further integration of AI for more sophisticated decision-making and adaptive behaviors. This includes the ability of AI to anticipate adversary actions and optimize swarm strategies in real-time, learning from past missions to improve performance. Moreover, the convergence of diverse autonomous systems – air, ground, and even maritime – into multi-domain swarms presents an even more formidable challenge to traditional defense strategies. This evolving landscape demands continuous innovation, rigorous testing, and careful consideration of the operational, ethical, and policy dimensions of autonomous swarm operations.