<p>This paper addresses the critical challenge of balancing dynamic spatiotemporal adaptability with multi-task collaborative optimization in wireless sensor systems and mobile device-generated trajectory clustering for smart cities. To tackle the geometric insensitivity of traditional distribution alignment and the representation degradation due to gradient conflicts, we propose a Progressive Multi-task Deep Trajectory Clustering (PMDTC) framework. Our approach features an Entropy-Regularized Optimal Transport Clustering algorithm that enhances geometric awareness through an adaptive cost matrix and iterative row-column normalization, alongside a Progressive Hierarchical Extraction Architecture that integrates task-specific expert systems with shared knowledge routing to mitigate multi-objective interference. Extensive experiments on real-world urban datasets demonstrate that PMDTC significantly outperforms state-of-the-art baselines in three metrics, providing an effective solution for trajectory pattern analysis in complex urban scenarios.</p>