Robust Networked MPC for Quadrotor Trajectory Tracking with Adaptive Horizon and Delay Estimation

This paper presents a networked control architecture for quadrotor unmanned aerial vehicles (UAVs) operating under communication constraints, specifically non deterministic delays and stochastic packet loss. To mitigate the adverse effects of network induced latencies in both wired and wireless channels, a model predictive control (MPC) framework leveraging a buffering mechanism is proposed. The proposed framework combines packetized predictive control through command buffering, online delay estimation, and adaptive prediction-horizon selection to preserve tracking performance in the presence of communication uncertainties. The primary objective is to align the performance of the networked system with that of an ideal, delay-free environment. To alleviate the computational burden inherent in long horizon optimization, an enhanced adaptive version of the controller is introduced. In this formulation, the predictive horizon, and consequently the buffer length, is dynamically adjusted based on real time delay estimations. The efficacy of the proposed adaptive networked MPC is validated through comprehensive numerical simulations and a comparative analysis against conventional MPC schemes employing standard hold input mechanisms using LHS and Monte Carlo techniques. Furthermore, sufficient conditions guaranteeing local input-to-state practical stability are derived. Results demonstrate that the proposed adaptive buffer-based approach achieves comparable tracking performance while substantially reducing computational effort, ensuring tracking precision, increased robustness against packet losses and communication delays, and a more favorable trade-off between control performance and computational complexity.