A Multi-Layered Model Predictive Control Framework for Collision-Aware UAV Swarms in 3-D Intelligent Transportation Systems

The increasing integration of uncrewed aerial vehicles (UAVs) into intelligent transportation systems (ITS) presents significant challenges in ensuring safe and efficient navigation in dense, shared airspace. This paper proposes a novel, multi-layered Model Predictive Control (MPC) framework that enables drones to coexist safely in congested three-dimensional environments. The proposed approach sequentially integrates three complementary modules: (i) Communication-Adaptive MPC (CA-MPC) that dynamically adjusts prediction horizons and safety buffers based on real-time communication quality; (ii) Uncertainty-Field MPC (UF-MPC) that constructs continuous probabilistic risk fields from sensor, prediction, and environmental uncertainties to guide collision-free trajectories; and (iii) Intent-Predictive MPC (IP-MPC) that incorporates predicted maneuver intentions of neighboring drones to enhance cooperative motion planning. The combined system is evaluated through high-fidelity 3D simulations of 50 (to 500) autonomous drones operating in free-space swarm environments. Performance metrics include collision rate, minimum separation distance, trajectory tracking accuracy, intent prediction effectiveness, and computational feasibility. Results demonstrate that the proposed framework significantly reduces collision risk, improves corridor throughput, and maintains robust operation under communication loss and sensor uncertainty. The study highlights the potential of multi-layered, risk-aware, and intent-informed MPC for enabling scalable and safe aerial traffic management in future urban air mobility systems. Comprehensive simulations, conducted in a MATLAB R2024b environment with a high-performance 50-500 drone high-fidelity software-in-the-loop (SIL) testbed, validate the effectiveness of the proposed multi-layered MPC framework under dense swarm traffic conditions.