TrajMamba: An Efficient and Semantic-rich Vehicle Trajectory Pre-training Model

2025-10-20

AI Summary

Balancing semantic richness and computational efficiency in vehicle GPS trajectory modeling remains challenging: POI/address text introduces substantial computational overhead, while raw trajectories contain redundant points that degrade embedding quality. To address this, we propose Traj-Mamba Encoder—a novel architecture integrating GPS motion patterns with road topology awareness. It introduces the first trip-purpose-aware lightweight pretraining mechanism, requiring no additional annotations, and a knowledge-distillation-based learnable keypoint compression module that jointly optimizes trajectory redundancy removal and semantic enhancement. Leveraging the Mamba architecture, it efficiently captures long-range temporal dependencies and seamlessly integrates POI semantic encoding. Evaluated on two real-world datasets across three downstream tasks—trajectory classification, stop detection, and destination prediction—our method achieves an average accuracy improvement of 2.1–4.8 percentage points over state-of-the-art methods, while accelerating inference by 37%–52%.

Technology Category

• Planning, Routing, and Scheduling: Planning with Language Models
• Data Mining & Knowledge Management: Representing, Reasoning, and Using Provenance, Trust
• Machine Learning: Mixture of Experts (MoE)

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Vehicle GPS trajectories record how vehicles move over time, storing valuable travel semantics, including movement patterns and travel purposes. Learning travel semantics effectively and efficiently is crucial for real-world applications of trajectory data, which is hindered by two major challenges. First, travel purposes are tied to the functions of the roads and points-of-interest (POIs) involved in a trip. Such information is encoded in textual addresses and descriptions and introduces heavy computational burden to modeling. Second, real-world trajectories often contain redundant points, which harm both computational efficiency and trajectory embedding quality. To address these challenges, we propose TrajMamba, a novel approach for efficient and semantically rich vehicle trajectory learning. TrajMamba introduces a Traj-Mamba Encoder that captures movement patterns by jointly modeling both GPS and road perspectives of trajectories, enabling robust representations of continuous travel behaviors. It also incorporates a Travel Purpose-aware Pre-training procedure to integrate travel purposes into the learned embeddings without introducing extra overhead to embedding calculation. To reduce redundancy in trajectories, TrajMamba features a Knowledge Distillation Pre-training scheme to identify key trajectory points through a learnable mask generator and obtain effective compressed trajectory embeddings. Extensive experiments on two real-world datasets and three downstream tasks show that TrajMamba outperforms state-of-the-art baselines in both efficiency and accuracy.

Problem

• Modeling travel purposes from textual addresses without computational overhead
• Handling redundant GPS points to improve trajectory embedding quality
• Learning movement patterns from both GPS and road perspectives efficiently

Innovation

• Jointly models GPS and road perspectives for movement patterns
• Integrates travel purposes into embeddings without extra overhead
• Uses knowledge distillation to compress trajectories and reduce redundancy