Robust adaptive H_∞ fault-tolerant predictive control for air-ground integrated highway emergency self-organizing...

Road traffic safety remains a critical challenge for public safety governance, and highway emergencies demand rapid, reliable communication and decision support. For air-ground integrated wireless ad-hoc networks in such scenarios, existing fault-tolerant control schemes rarely address the coupled effects of highly time-varying channels, packet loss, actuator/sensor degradation, and accident risk prediction within a unified framework. This paper proposes a robust adaptive H_∞ fault-tolerant predictive control method incorporating RBF-DNN (Radial Basis Function-Deep Neural Network) approximation for a two-layer heterogeneous "UAV-road" network. A networked control model integrating link quality, queue evolution, and resource execution constraints is constructed, and an RBF-DNN is employed for online approximation of unknown nonlinearities and unmodeled dynamics. An adaptive law with projection and σ-modification is developed to suppress parameter drift, and an event-triggered communication mechanism is introduced to reduce overhead and avoid Zeno behavior. Within a Lyapunov framework, we rigorously establish bounded stability and an H_∞ performance upper bound for the closed-loop system under link failures, random packet loss, and partial actuator/sensor faults. Simulation results show that, under sudden link interruptions, 40-50 % packet loss and 75 % actuator performance degradation, the proposed controller maintains QoS and prescribed γ-level H_∞ performance with asymptotic state convergence. On a real S103 highway dataset (372 samples, 10.8 % positives), the associated accident risk prediction module achieves 70-75 % test accuracy with PR-AUC is 0.683 and ROC-AUC is 0.715, demonstrating practically useful high-risk segment discrimination under severe class imbalance and confirming the engineering feasibility of the proposed control prediction framework.