Disturbed Dynamic Analysis and Robust Decision-Making Control of Complex Networks

Abstract

: Complex networks in cyber–physical, transportation, and information infrastructures operate under topology variations, unmeasured disturbances, and limited actuation. This paper proposes robust disturbance-aware data-driven decision control (R-D3C), which couples a sliding-window graph-regularized estimator, disturbance-envelope adaptation, sparse intervention allocation, receding-horizon optimization, and a robust safety projection. The theory directly bounds the dynamic prediction regret of the implemented sliding-window estimator. A checkable sufficient condition for safety-filter feasibility is coupled with an explicit slack-and-backup fallback when the strict projection is infeasible. Practical input-to-state stability and sparse-allocation risk reduction are established. The nominal comparison uses 30 paired runs with standard deviations and significance tests, while the separate deployment-impairment benchmark uses 20 runs; evaluation also includes the IEEE 39-bus topology and timing tests up to 1000 nodes. R-D3C achieves the lowest constraint-violation rate; its residual error is statistically comparable with adaptive MPC while preserving an explicit safety layer.