Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

MFTraj: Map-Free, Behavior-Driven Trajectory Prediction for Autonomous Driving

Haicheng Liao, Zhenning Li, Chengyue Wang, Huanming Shen, Bonan Wang, Dongping Liao, Guofa Li, Chengzhong Xu

TL;DR

MFTraj addresses map-free trajectory prediction in autonomous driving by leveraging historical trajectories of the target and nearby agents to forecast future paths without HD maps. It introduces a four-module encoder–decoder architecture comprising a behavior-aware module with a VRNN–GRU-based behavioral encoder, a position-aware module with an LSTM-based encoder, an adaptive structure-aware GCN-based interaction module (coupled with Linformer for efficient attention), and a residual decoder to produce the target trajectory over a horizon $t_f$. The approach emphasizes continuous behavioral attributes via centrality-inspired criteria and dynamic graphs to capture spatio-temporal interactions, achieving robustness under substantial data loss while maintaining competitive accuracy and reduced parameter requirements. Evaluations on Argoverse, NGSIM, HighD, and MoCAD show MFTraj outperforms many state-of-the-art baselines in complete and missing-data scenarios, highlighting its potential to reduce map-dependency and data demands in real-world autonomous driving tasks.

Abstract

This paper introduces a trajectory prediction model tailored for autonomous driving, focusing on capturing complex interactions in dynamic traffic scenarios without reliance on high-definition maps. The model, termed MFTraj, harnesses historical trajectory data combined with a novel dynamic geometric graph-based behavior-aware module. At its core, an adaptive structure-aware interactive graph convolutional network captures both positional and behavioral features of road users, preserving spatial-temporal intricacies. Enhanced by a linear attention mechanism, the model achieves computational efficiency and reduced parameter overhead. Evaluations on the Argoverse, NGSIM, HighD, and MoCAD datasets underscore MFTraj's robustness and adaptability, outperforming numerous benchmarks even in data-challenged scenarios without the need for additional information such as HD maps or vectorized maps. Importantly, it maintains competitive performance even in scenarios with substantial missing data, on par with most existing state-of-the-art models. The results and methodology suggest a significant advancement in autonomous driving trajectory prediction, paving the way for safer and more efficient autonomous systems.

MFTraj: Map-Free, Behavior-Driven Trajectory Prediction for Autonomous Driving

TL;DR

MFTraj addresses map-free trajectory prediction in autonomous driving by leveraging historical trajectories of the target and nearby agents to forecast future paths without HD maps. It introduces a four-module encoder–decoder architecture comprising a behavior-aware module with a VRNN–GRU-based behavioral encoder, a position-aware module with an LSTM-based encoder, an adaptive structure-aware GCN-based interaction module (coupled with Linformer for efficient attention), and a residual decoder to produce the target trajectory over a horizon . The approach emphasizes continuous behavioral attributes via centrality-inspired criteria and dynamic graphs to capture spatio-temporal interactions, achieving robustness under substantial data loss while maintaining competitive accuracy and reduced parameter requirements. Evaluations on Argoverse, NGSIM, HighD, and MoCAD show MFTraj outperforms many state-of-the-art baselines in complete and missing-data scenarios, highlighting its potential to reduce map-dependency and data demands in real-world autonomous driving tasks.

Abstract

This paper introduces a trajectory prediction model tailored for autonomous driving, focusing on capturing complex interactions in dynamic traffic scenarios without reliance on high-definition maps. The model, termed MFTraj, harnesses historical trajectory data combined with a novel dynamic geometric graph-based behavior-aware module. At its core, an adaptive structure-aware interactive graph convolutional network captures both positional and behavioral features of road users, preserving spatial-temporal intricacies. Enhanced by a linear attention mechanism, the model achieves computational efficiency and reduced parameter overhead. Evaluations on the Argoverse, NGSIM, HighD, and MoCAD datasets underscore MFTraj's robustness and adaptability, outperforming numerous benchmarks even in data-challenged scenarios without the need for additional information such as HD maps or vectorized maps. Importantly, it maintains competitive performance even in scenarios with substantial missing data, on par with most existing state-of-the-art models. The results and methodology suggest a significant advancement in autonomous driving trajectory prediction, paving the way for safer and more efficient autonomous systems.
Paper Structure (14 sections, 22 equations, 4 figures, 5 tables)

This paper contains 14 sections, 22 equations, 4 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Methodologies
  4. Inputs and Outputs
  5. Behavior-aware Module
  6. Position-aware Module
  7. Interaction-aware Module
  8. Residual Decoder
  9. Experiments
  10. Experimental Setup
  11. Experimental Results
  12. Ablation Studies
  13. Qualitative Results
  14. Conclusion

Figures (4)

  • Figure 1: Architecture of the proposed trajectory prediction model.
  • Figure 2: Overview of our adaptive structure-aware GCN. The real-time trajectories of the target and observed agents are captured using a topology graph to form a feature matrix. This matrix undergoes aggregation, updating, and iteration within the GCN. As new agents are observed in real-time, the GCN dynamically adjusts its topology, updating features for the added nodes.
  • Figure 3: Qualitative results of MFTraj and HiVT on Agroverse.
  • Figure 4: Qualitative results of MFTraj on NGSIM. Target vehicle is depicted in red, while its surrounding agents are shown in blue.