Optimized Design of Multi-Layer LEO Satellite Constellation for Regional Traffic Demand in Non-Terrestrial Networks

Non-terrestrial networks (NTNs) leverage low Earth orbit (LEO) satellite constellation (LSC) to provide wide-area and seamless connectivity. However, regional traffic demand is highly differentiated, and a conventional single-layer constellation cannot satisfy on-demand coverage and high uplink-capacity requirements under deployment-cost constraints, thereby limiting NTN performance. This paper proposes an optimization framework for multi-layer LEO constellation design tailored to differentiated regional traffic. A spatio-temporal graph (STG) representation is introduced to capture the time-varying connectivity of ground-to-satellite links (GSLs) and inter-satellite links (ISLs), and an invulnerability model is developed to characterize the reliability and service continuity of the satellite network under disturbances. In addition, a terrain-aware terminal distribution and regional traffic model is incorporated to analyze the average uplink capacity. Building on these models, a four-objective constellation design problem is formulated to maximize the average multiple coverage ratio, uplink capacity, and invulnerability, while minimizing deployment cost. To solve this problem, an FCNN-surrogate-assisted Stage-Aware Dynamic Adaptive NSGA-III (FCNN-SADANSGA-III) optimization method is developed. Simulation results demonstrate that the proposed framework achieves improved multi-objective trade-offs and enables constellation configurations that better matches regional traffic demand.