New AI Framework PL-MARL Keeps Drone Swarms Connected with Minimal Bandwidth

Coordination collapse is a critical risk for drone swarms operating in bandwidth-constrained environments. In sparse-signaling aerial networks, limited communication and information aging can cause drones to lose track of each other, degrading coverage and mission reliability. To address this, researchers Chuan-Chi Lai and Ang-Hsun Tsai from the IEEE community propose the Predictive Lightweight Multi-Agent Reinforcement Learning (PL-MARL) framework. The core innovation is a Kinematic-Aware Inference Engine that proactively reconstructs neighbor trajectories using physical priors—without relying on frequent signal exchanges. This enables a computation-for-communication trade-off that decouples structural resilience from signaling frequency, reducing control overhead while maintaining robust coordination.

The approach was validated through simulations that replicate extreme signaling scarcity and node failures. Results demonstrate that PL-MARL maintains superior coverage performance and mission continuity compared to traditional methods, even when bandwidth is severely limited. By minimizing control overhead, the framework preserves spectrum for payload services (e.g., sensor data or video feeds) while ensuring resilience against interference and node loss. The paper has been accepted for publication in IEEE Wireless Communications Letters (July 2026), marking a scalable, low-latency solution for robust aerial coordination in real-world applications like disaster response, surveillance, and communications relay.