System identification provides the data-to-model link for analysis, prediction, diagnosis, and control of dynamical systems. As engineering and scientific systems become increasingly nonlinear, high-dimensional, networked, time-varying, and partially observed, identification methods must handle noisy data, incomplete prior knowledge, safety constraints, and downstream control requirements. This review surveys system identification for complex dynamical systems through a unified framework that connects model structure, operating mode, estimation target, and computational setting. We first revisit single-system identification, including regularized regression, sparse Bayesian learning, Kalman-type filtering, and recent structured recursive methods such as auxiliary-model, multi-innovation, hierarchical, filtering-based, and coupled identification. We then review multi-system identification for families of interacting or related subsystems, covering deep neural architectures, sparse equation discovery for ordinary and partial differential equations, hybrid and topology-aware modeling, and meta-learning-based fast adaptation. Finally, we discuss large language model-assisted workflows for symbolic regression, equation discovery, feature generation, and scientific modeling. By comparing these methodological streams, this review highlights their strengths, limitations, and applicability to prediction, synchronization, diagnosis, and control. We identify key open challenges, including robust identification under biased or limited data, uncertainty-aware modeling, interpretable learning, reproducible benchmarking, and tighter integration between identification, experiment design, and control synthesis.

Low-altitude unmanned aerial vehicle (UAV) logistics networks are subject to multiple operational disturbances that can severely degrade delivery performance. This paper proposes a Kriging-based resilience assessment framework that efficiently evaluates the reliability and resilience of such networks under multi-dimensional disturbances. A discrete-event simulation environment is developed that incorporates Voronoi-based airspace topology, A* path planning, and dynamic re-planning. Three disturbance types—structural (no-fly zones), functional (adverse weather), and informational (communication delays)—are modelled and screened via Sobol sensitivity analysis. Two complementary Kriging surrogates, trained on Latin Hypercube samples, replace expensive Monte Carlo simulations with a 25-fold speed-up: one predicting endpoint cumulative orders for failure probability estimation, and the other predicting the integral resilience triangle index for resilience surface characterisation. Results show that structural disruption and weather degradation contribute nearly equally to performance variance (52.4% vs. 47.6%). The resilience surface reveals that only 10.9% of the disturbance parameter space sustains high resilience (R ≥ 0.8), while 26.4% falls into a low-resilience regime (R < 0.6) where severe throughput loss and incomplete recovery coexist. The proposed framework provides a computationally efficient tool for resilience-informed management of low-altitude logistics networks.

This paper studies an average-cost maintenance optimization problem for a single-component system subject to continuous degradation and random shocks. The degradation process is modeled by a Gamma process, while random shocks arrive according to a Poisson process. A system failure occurs when either the degradation level reaches or exceeds its failure threshold, or the accumulated number of shocks reaches the shock failure threshold. Since failures are non-self-announcing, both the failure state and the system condition can only be revealed through inspections. To reduce the long-run operating cost, we propose a state-dependent non-periodic inspection and replacement policy, where the next inspection interval and the replacement action are jointly optimized according to the current degradation level and accumulated shock state. A numerical algorithm based on state-space discretization and bisection search is developed to compute the optimal policy. Numerical results show that the proposed policy consistently outperforms the optimized periodic inspection and replacement benchmark, with average-cost-rate reductions ranging from 8.60% to 29.33% across the tested parameter settings. Sensitivity analysis further demonstrates that the proposed policy consistently outperforms the benchmark policy under different cost, shock, and degradation parameter settings. The results confirm that dynamically adjusting inspection intervals according to the system state can effectively balance inspection cost, downtime cost, and replacement cost.