Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Micro-Macro Spatial-Temporal Graph-based Encoder-Decoder for Map-Constrained Trajectory Recovery

Tonglong Wei, Youfang Lin, Yan Lin, Shengnan Guo, Lan Zhang, Huaiyu Wan

TL;DR

The paper tackles map-constrained trajectory recovery from sparse GPS data by proposing MM-STGED, a graph-based encoder–decoder that jointly models micro-semantics of individual trajectories and macro-semantics across trajectory groups. The micro module encodes trajectories as a fully connected spatial-temporal graph that fuses node and edge information via a novel graph convolution, while the macro module builds a macro trajectory flow graph and road-condition representation to guide recovery. The model is trained with a multi-task loss balancing road-segment prediction and moving-ratio estimation, and experiments on two real-world datasets show MM-STGED consistently outperforms state-of-the-art baselines, especially at larger sampling intervals. The approach provides more accurate reconstructions and better captures human travel preferences and traffic context, offering significant value for ITS tasks such as route analysis and mobility modeling.

Abstract

Recovering intermediate missing GPS points in a sparse trajectory, while adhering to the constraints of the road network, could offer deep insights into users' moving behaviors in intelligent transportation systems. Although recent studies have demonstrated the advantages of achieving map-constrained trajectory recovery via an end-to-end manner, they still face two significant challenges. Firstly, existing methods are mostly sequence-based models. It is extremely hard for them to comprehensively capture the micro-semantics of individual trajectory, including the information of each GPS point and the movement between two GPS points. Secondly, existing approaches ignore the impact of the macro-semantics, i.e., the road conditions and the people's shared travel preferences reflected by a group of trajectories. To address the above challenges, we propose a Micro-Macro Spatial-Temporal Graph-based Encoder-Decoder (MM-STGED). Specifically, we model each trajectory as a graph to efficiently describe the micro-semantics of trajectory and design a novel message-passing mechanism to learn trajectory representations. Additionally, we extract the macro-semantics of trajectories and further incorporate them into a well-designed graph-based decoder to guide trajectory recovery. Extensive experiments conducted on sparse trajectories with three different sampling intervals that are respectively constructed from two real-world trajectory datasets demonstrate the superiority of our proposed model.

Micro-Macro Spatial-Temporal Graph-based Encoder-Decoder for Map-Constrained Trajectory Recovery

TL;DR

The paper tackles map-constrained trajectory recovery from sparse GPS data by proposing MM-STGED, a graph-based encoder–decoder that jointly models micro-semantics of individual trajectories and macro-semantics across trajectory groups. The micro module encodes trajectories as a fully connected spatial-temporal graph that fuses node and edge information via a novel graph convolution, while the macro module builds a macro trajectory flow graph and road-condition representation to guide recovery. The model is trained with a multi-task loss balancing road-segment prediction and moving-ratio estimation, and experiments on two real-world datasets show MM-STGED consistently outperforms state-of-the-art baselines, especially at larger sampling intervals. The approach provides more accurate reconstructions and better captures human travel preferences and traffic context, offering significant value for ITS tasks such as route analysis and mobility modeling.

Abstract

Recovering intermediate missing GPS points in a sparse trajectory, while adhering to the constraints of the road network, could offer deep insights into users' moving behaviors in intelligent transportation systems. Although recent studies have demonstrated the advantages of achieving map-constrained trajectory recovery via an end-to-end manner, they still face two significant challenges. Firstly, existing methods are mostly sequence-based models. It is extremely hard for them to comprehensively capture the micro-semantics of individual trajectory, including the information of each GPS point and the movement between two GPS points. Secondly, existing approaches ignore the impact of the macro-semantics, i.e., the road conditions and the people's shared travel preferences reflected by a group of trajectories. To address the above challenges, we propose a Micro-Macro Spatial-Temporal Graph-based Encoder-Decoder (MM-STGED). Specifically, we model each trajectory as a graph to efficiently describe the micro-semantics of trajectory and design a novel message-passing mechanism to learn trajectory representations. Additionally, we extract the macro-semantics of trajectories and further incorporate them into a well-designed graph-based decoder to guide trajectory recovery. Extensive experiments conducted on sparse trajectories with three different sampling intervals that are respectively constructed from two real-world trajectory datasets demonstrate the superiority of our proposed model.
Paper Structure (33 sections, 22 equations, 7 figures, 5 tables)

This paper contains 33 sections, 22 equations, 7 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Map-constrained Trajectory Recovery
  4. Trajectory Modeling Technology
  5. Preliminaries
  6. Basic Definition
  7. Problem Definition
  8. Methodology
  9. Model Overview
  10. Graph-based Trajectory Micro-Semantics Encoder
  11. Micro Spatial-Temporal Trajectory Graph Construction
  12. Trajectory Encoder
  13. Road Network-aware Layer
  14. Macro-Semantics Extraction Module
  15. Macro Trajectory Flow Graph Construction
  16. ...and 18 more sections

Figures (7)

  • Figure 1: Illustrations of map-constrained trajectory recovery, road segment, and moving ratio.
  • Figure 2: Illustrations of the trajectory’s micro-semantic information and different trajectory encoders. (a) the micro-semantic information of a trajectory $\tau_2 = \langle p_1, \cdots, p_4 \rangle$ includes the explicit absolute information of single GPS points and the implicit relative information between GPS points. (b) Sequence-based encoders. (c) Graph-based encoder.
  • Figure 3: The overall framework of MM-STGED. It consists of a graph-based trajectory micro-semantics encoder, a macro-semantics extraction module, and a graph-based trajectory recovery decoder.
  • Figure 4: Illustration of macro-semantics extraction.
  • Figure 5: Key hyper-parameters experiment on the Chengdu dataset.
  • ...and 2 more figures

Theorems & Definitions (4)

  • Definition 1: Trajectory
  • Definition 2: Road Network
  • Definition 3: Map-constrained Trajectory Point
  • Definition 4: $\epsilon$-Sampling Interval Map-constrained Trajectory