A novel fuzzy system-based genetic algorithm for trajectory segment generation in urban global positioning system

The widespread adoption of Global Positioning System (GPS) technology has rendered trajectory data pivotal for urban traffic planning and travel pattern analysis. However, the traditional trajectory generation methods are constrained by manual cluster number settings, impeding both automation and optimality.

To address this issue, an enhanced Fuzzy System-based Genetic Algorithm (FGA) is proposed for automated trajectory segment generation.

Based on the angle-based partitioning and cosine-constrained segmentation strategy, the proposed method integrates a fuzzy system into the genetic algorithm to dynamically adjust crossover and mutation probabilities. This enables automatic cluster number determination and sub-trajectory generation without human intervention. Moreover, the algorithm incorporates global search capability to avoid local optima. The FGA is further combined with least squares regression and applied to real-world taxi GPS data for trajectory reconstruction.

The experiment results demonstrate that FGA, when combine with different clustering algorithms (K-means, K-median, FCM), consistently identifies appropriate cluster numbers and produces globally optimal, smooth trajectory representations. The method improves clustering quality, trajectory continuity, and stability across multiple clustering strategies.

The proposed FGA offers an effective and adaptive solution for trajectory segment generation in urban GPS system. Future work will explore enhancing its scalability, robustness to noise, and rule generalization across diverse datasets.