Dynamic Resilience Assessment and Collaborative Reconfiguration of UAV Kill Webs Based on Physical-Logical-Temporal Coupling

Abstract

In highly contested environments, the survivability of Unmanned Aerial Vehicle (UAV) Kill Webs faces severe challenges. Existing evaluations over-rely on static topological connectivity, ignoring physical constraints (distance, energy, latency), often causing a fatal misjudgment: “structural preservation alongside operational paralysis.” To address this, we propose a dynamic resilience assessment and collaborative reconfiguration framework integrated with a “physical-logical-temporal” coupling. First, an effective model based on heterogeneous closed-loop cycles is constructed. A Time-aware Accumulated Normalized Operational Capability (T-ANOC) metric is proposed to rigorously quantify the realistic penalties of reconfiguration latency and spatial attenuation on System-of-Systems (SoS) effectiveness. Second, under resource constraints, a many-to-one collaborative reconfiguration mechanism is defined, deeply comparing an Energy-Distance Collaborative Strategy (EDCS) with a Deep Reinforcement Learning (DRL) global optimization strategy. Large-scale Monte Carlo simulations reveal three profound insights: (1) It quantitatively verifies the nonlinear decoupling characteristic where operational collapse precedes topological disintegration; (2) Although DRL approaches short-term recovery limits, EDCS untangles the “energy-latency-effectiveness” trilemma. By reducing energy consumption by 45% and latency by 35%, EDCS achieves long-term sustainable resilience; (3) Microscopic node contribution analysis (peak contribution 238.9%) mechanistically quantifies the nonlinear emergent gains triggered by heterogeneous functional reorganization. This study provides rigorous mathematical support for designing high-survivability, self-adaptive UAV Kill Webs. Compared with topology-centered robustness metrics and purely learning-driven reconfiguration policies, the proposed framework jointly models physical feasibility, heterogeneous functional substitution, and temporal recovery loss within a unified resilience-analysis process.