Answers are grounded in this paper and cite their sources.
Real-time embedded implementation can produce substantial differences between theterminal performance predicted by numerical simulation and that realized by aphysical guidance-and-control processor. This study characterizes that gap througha controlled Model-in-the-Loop (MIL) and Hardware-in-the-Loop (HIL) evaluation ofterminal guidance for a high-dynamic flying vehicle. The two environments use thesame nonlinear six-degree-of-freedom plant, aerodynamic database, actuator model andconstraints, fixed-step plant solver, controller update rate, and paired seededvon K\'arm\'an turbulence realizations. True Proportional Navigation Guidance (TPNG)and a gain-scheduled modified PI autopilot with direct angular-rate damping areexecuted on an STM32F429 ARM Cortex-M4 microcontroller at 50 Hz, while the nonlinearvehicle and disturbance models are propagated on a real-time target computer. Forthe four investigated disturbance realizations, the MIL configuration yields planarterminal errors of 0.027--0.460 m, whereas the HIL experiments yield137.6--1865.6 m. The embedded guidance-and-control computation requires an averageof 0.356 ms within the 20 ms control interval, corresponding to 1.78% execution-timeoccupancy. Thus, computational overload is excluded as the direct source of theobserved degradation. The results demonstrate that end-to-end implementation effects,including communication, task scheduling, sample alignment, and data representation,must be treated as part of the closed-loop guidance design. The current experimentsquantify their aggregate effect; causal attribution to individual mechanisms requirescontrolled delay and interface-ablation tests.
No takes yet. Share an insight, caveat, or question.
Khalifa et al. (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: