Hierarchical Optimization Based Multi-objective Dynamic Regulation Scheme for VANET Topology

TL;DR

The paper tackles VANETs' highly dynamic topologies that inflate $L_{\text{avg}}$ and $E[T_{sd}]$ while constraining $W_{\text{total}}$. It introduces a hierarchical, two-layer topology regulation framework: a local layer for rapid adaptation via feature extraction and neighbor-aware fusion, and a global layer for coordinated multi-objective optimization using dynamic normalization and a dual-mode solver on $S(t)$. Key contributions include a dynamic multi-objective model, a three-stage feature fusion mechanism, and oscillation suppression through a performance improvement threshold and topology validity checks, demonstrated on real urban maps with SUMO where $L_{\text{avg}}\approx 4$ hops, $E[T_{sd}]\approx 0.01$ s, and markedly higher throughput than baselines. This approach enables real-time, distributed VANET optimization in high-mobility environments, enhancing reliability and efficiency for safety and cooperative applications.

Abstract

As a core technology of intelligent transportation systems, vehicular ad-hoc networks support latency-sensitive services such as safety warning and cooperative perception via vehicle-to-everything communications. However, their highly dynamic topology increases average path length, raises latency, and reduces throughput, severely limiting communication performance. Existing topology optimization methods lack capabilities in multi-objective coordination, dynamic adaptation, and global-local synergy. To address this, this paper proposes a two-layer dynamic topology regulation scheme combining local feature aggregation and global adjustment. The scheme constructs a dynamic multi-objective optimization model integrating average path length, end-to-end latency, and network throughput, and achieves multi-index coordination via link adaptability metrics and a dynamic normalization mechanism. it quickly responds to local link changes via feature fusion of local node feature extraction and dynamic neighborhood sensing, and balances optimization accuracy and real-time performance using a dual-mode adaptive solving strategy for global topology adjustment. It reduces network oscillation risks by introducing a performance improvement threshold and a topology validity verification mechanism. Simulation results on real urban road networks via the SUMO platform show that the proposed scheme outperforms traditional methods in average path length (stabilizing at ~4 hops), end-to-end latency (remaining ~0.01 s), and network throughput.

Figures (6)

• Figure 1: Road Network of a Real-World Urban Traffic Scenario.
• Figure 2: Comparison of Average Path Length Across Different Algorithms.
• Figure 3: Comparison of End-to-End Delay Under Different Algorithms.
• Figure 4: Comparison of Network Throughput Under Different Algorithms.
• Figure 5: Boxplots of the Three Performance Metrics Across Different Algorithms.