Flight Test of Multi-Agent Aircraft Swarm for Polar Research

This work presents a model predictive formation-flight control framework for multi-agent uncrewed aerial systems (UAS) designed for high-resolution remote sensing in polar environments. Unlike existing approaches based on proportional–integral (PI) control, consensus protocols, or potential-field coordination, which generally lack predictive constraint handling and degrade in tightly spaced formations while suffering from slow and out-of-phase convergence, the proposed method leverages a model predictive control (MPC) architecture to explicitly regulate inter-vehicle spacing while each aircraft tracks predefined survey trajectories over rapidly changing glacier topography and environmental disturbances. The proposed controller leverages a model predictive control (MPC) architecture capable of explicitly enforcing the stringent geometric constraints demanded by multi-baseline IPR coherence. To ensure tight formation flight as the swarm follows prescribed survey lines over rapidly evolving glacier topography, the MPC computes optimal velocity perturbations that regulate longitudinal spacing while ensuring that commanded airspeeds remain within safe aerodynamic limits, thereby preventing stall or overspeed conditions and maintaining stable flight. Because the nonlinear MPC computational load increases approximately linearly with the prediction horizon N, a detailed sensitivity analysis is conducted to assess scalability and determine real-time feasible settings for formations of varying size. Comparative evaluations with a baseline PI controller demonstrate improved formation preservation, path-tracking accuracy, and robustness to wind disturbances as formation size grows. A high-fidelity case study over Helheim Glacier in East Greenland, incorporating realistic digital elevation models and representative aircraft performance, validates the proposed method and confirms its suitability for coordinated multi-UAS IPR surveys in highly dynamic polar environments.