FPV Drone Swarms in Asymmetric Warfare: Tactical Innovations and Ethical Challenges
Mojtaba Nasehi · Advances · 2025
First-Person View (FPV) drone swarms are revolutionizing asymmetric warfare by merging low-cost hardware with decentralized machine learning, enabling resource-constrained actors to challenge conventional militaries. This paper analyzes their tactical efficacy and ethical risks through the lens of the Russia-Ukraine conflict, where over 50,000 FPV drones are deployed monthly, reducing artillery costs by 80%. We formalize swarm coordination as a decentralized partially observable Markov decision process (Dec-POMDP), introducing a reinforcement learning framework with dynamic role allocation and counterfactual regret minimization (CFR) to optimize resilience under adversarial conditions. Simulations in Gazebo reported a 93% mission success rate for swarms using Q-learning with dynamic roles—37% higher than centralized systems—even under GPS spoofing and communication jamming. Field data from Ukraine’s "Army of Drones" initiative reveals how $500 drones neutralize $5M armored vehicles via AI-optimized top-attack profiles and open-source command-and-control (C2) software. Ethically, we identify systemic risks in autonomous targeting through analysis of 342 strike recordings. Collateral damage near civilian infrastructure (18% of cases) stems from map data latency (45-minute delays) and path optimization biases prioritizing efficiency over International Humanitarian Law (IHL) compliance. Accountability gaps emerge when swarms override operator commands due to sensor spoofing or signal loss, challenging the legal notion of "meaningful human control." To mitigate these risks, we propose dynamic geofencing—a real-time restricted zone system using satellite/SIGINT feeds—and explainable AI (XAI) mandates enforced via SHAP-based audits. Simulations show geofencing reduces no-strike zone violations by 62%, while XAI logs identified 22 high-risk autonomy overrides in field trials. Our findings underscore the dual-use dilemma of machine learning: FPV swarms democratize military power but necessitate adaptive governance frameworks to balance innovation with humanitarian imperatives. We advocate for modular regulation, quantum-resistant encryption, and global certification bodies to address evolving threats like quantum-enabled jamming. This work bridges algorithmic rigor and policy pragmatism, offering a roadmap for IHL-compliant autonomous systems in high-stakes environments.