This paper proposes a complex network-based evaluation method for assessing the architecture of intelligent connected vehicle cyber-physical systems (IVCPS). The approach converts the IVCPS architecture into a complex network model by defining nodes, edges, and their weights, while introducing metrics such as degree centrality and edge betweenness centrality to quantify network characteristics. Further analysis of optimization strategies through modeling enhances network robustness. Using a “vehicle-cloud collaborative hazardous materials transportation scenario” as an example, this paper analyzes architectural network metrics to identify core nodes. Optimization results demonstrate that the MLDF strategy stably enhances robustness with minimal edge additions. Conclusions emphasize the importance of strengthening core node connections and redundant paths to optimize IVCPS architectural performance.