Evaluation of Guidance, Navigation, and Control Algorithms for Hydrogen-Powered Multi-Aircraft Systems

This paper presents a modular digital-twin framework to compare inner-loop attitude controllers: geometric PD (“PID”), discrete Linear-Quadratic Regulator (LQR), and a move-penalized linear Model Predictive Controller (MPC), for a heavy-lift T30-class quadcopter intended for hydrogen propulsion study. The twin couples 6-DoF rigid-body dynamics, actuator mixing, motor/ESC lag, a bus-level electrical model, and stochastic wind (Ornstein–Uhlenbeck) with look-ahead guidance on a sharp lawnmower survey. To isolate controller effects, task difficulty is equalized by autotuning a single scalar so that the achieved cross-track Root-Mean-Squared Error (RMSE) lies in a 2.6 ± 0.25 m band. The tuned controllers then run identical 600 s simulations under the same wind seed and retimed speeds. On the equalized run, PID and LQR achieve 2.45 m and 2.55 m RMSE, respectively, while MPC settles at 3.13 m due to its move penalty and finite horizon. All three deliver survey-class performance with mean bus power ≈ 4.8 kW and peaks in the 7–8 kW range, but MPC reduces energy per meter by approximately 2.3% at the cost of relaxed lateral accuracy. A 10-seed Monte Carlo confirms this trade: PID/LQR remain in-band for 90%/80% of seeds, while MPC consistently lowers energy per meter with similar mean power but gentler peaks. For hydrogen-electric UAVs, these metrics map directly to propulsion co-design, where energy per meter informs hydrogen mass and range, peak power sets stack/buffer sizing, and actuator smoothness affects balance-of-plant transients. The results show that controller selection is not only a matter of tracking accuracy but also an energy-management lever: PID/LQR suit survey tolerance, while MPC-style penalization favours endurance and balance-of-plant stability. Keywords: Digital twin; Guidance, Navigation, and Control; Heavy-lift Multirotor; Hydrogen-electric propulsion; Predictive Control Model.