Toward robust UAV swarm coordination in low-altitude networks: Online learning control under system and channel uncertainty

Robust tracking control of UAV swarms has attracted significant attention, yet most existing works assume reliable leader–follower communication and accurate system models. In practice, uncertain dynamics and unreliable wireless links can severely degrade swarm coordination. In this paper, we develop an online robust tracking control framework for UAV swarms operating over unreliable multiple-input multiple-output (MIMO) channels. We first establish a dynamic model that captures coupled uncertainties arising from inter-UAV coordination and communication imperfections. Building on this model, we formulate the tracking problem and cast it as an ergodic optimal control problem for an auxiliary system that unifies system and channel variations. To address the resulting computational challenges, we derive an equivalent lower-dimensional representation that significantly reduces complexity. Feasibility and stability under this representation are guaranteed via controllability cone decomposition, contractive system analysis, and Lyapunov drift arguments. Based on the reduced-dimensional robust control solution, we further propose an online stochastic-approximation-based learning algorithm, whose almost sure convergence is established using the ordinary differential equation (ODE) method. Simulation results demonstrate that the proposed approach achieves improved tracking stability, faster convergence, and higher computational efficiency than benchmark schemes, particularly under stochastic channel conditions.