Event-Triggered Multi-Model Predictive Control for Hypersonic Flight Vehicles

Abstract

The large-envelope velocity and altitude tracking of hypersonic flight vehicles is hindered by a combination of dynamic and practical constraints. Dynamically, it suffers from strong nonlinearities and varying operating points; practically, it contends with unavailable states, actuator limits, and the heavy computational burden imposed by periodic predictive optimization. This study proposes a tracking differentiator-based event-triggered state-feedback model predictive control (TD-ETSF-MPC) framework with multiple local prediction models. Three local linear models are built at representative equilibrium points and are activated based on the current operating condition. Meanwhile, a tracking differentiator provides smoothed reference commands, and a Luenberger observer estimates the unmeasured states for feedback. The proposed predictive controller directly accounts for actuator constraints within the finite-horizon framework. To reduce computational cost, online optimization is executed only when the estimated-state error exceeds a preset bound or when the time since the last update reaches the maximum interval. Outside these conditions, the controller keeps the active model and the last computed control command. An event-triggered augmented error model is formulated to explicitly account for model switching, observer estimation errors, input holding, and bounded inter-update intervals. For the nominal closed-loop system, mode-dependent Lyapunov inequalities yield sufficient conditions for local exponential stability; this analysis is further extended to establish uniform ultimate boundedness in the presence of bounded disturbances. Simulation results demonstrate that TD-ETSF-MPC achieves tracking performance comparable to TD-SF-MPC with satisfactory robustness, while its update/hold statistics further confirm that the optimization problem is solved only intermittently, thus significantly reducing redundant computations.