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.

What problem does this paper attempt to address?

The main objective of this paper is to propose a new trajectory prediction model named MFTraj (Map-Free, Behavior-Driven Trajectory Prediction), specifically designed for the application of autonomous vehicles (AVs) in complex dynamic traffic scenarios. The distinguishing feature of MFTraj is that it does not rely on high-definition map (HD maps) data for effective trajectory prediction. Specifically, the MFTraj model addresses the problem in the following ways: 1. **Model Architecture**: MFTraj is a map-free model, meaning it does not depend on high-definition map information. It utilizes historical trajectory data combined with a novel behavior-driven module to predict future trajectories. This design reduces the need for resource-intensive high-definition maps and enhances the model's generality and adaptability. 2. **Behavior-Aware Module**: This module captures the continuous changes in driver behavior through dynamic geometric graphs (DGG) and various centrality measures (such as degree centrality, closeness centrality, etc.), thereby better understanding the interactive behavior of all participants on the road. 3. **Position-Aware Module**: Unlike traditional methods, this module emphasizes relative positions rather than absolute positions to better capture the spatial dynamics of individuals and groups. 4. **Interaction-Aware Module**: A novel adaptive structure-aware graph convolutional network (GCN) is employed, which can dynamically adjust its topology based on the real-time observed number of participants, thus more flexibly responding to changes in traffic conditions. 5. **Residual Decoder**: Processes node vectors to predict the future trajectory of the target vehicle. Experimental results show that MFTraj not only performs excellently on the complete dataset but also exhibits strong robustness and adaptability in the case of missing data. Additionally, compared to existing state-of-the-art (SOTA) models, MFTraj is more efficient in terms of the number of parameters while achieving higher levels of prediction accuracy. This indicates that MFTraj represents a significant advancement in trajectory prediction for the field of autonomous driving.

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

A Cognitive-Driven Trajectory Prediction Model for Autonomous Driving in Mixed Autonomy Environment

Enhanced Multimodal Trajectory Prediction for Autonomous Vehicles Using Advanced Diffusion Model Techniques

Characterized Diffusion and Spatial-Temporal Interaction Network for Trajectory Prediction in Autonomous Driving

A Cognitive-Based Trajectory Prediction Approach for Autonomous Driving

Self-Aware Trajectory Prediction for Safe Autonomous Driving

A Probabilistic Architecture of Long-Term Vehicle Trajectory Prediction for Autonomous Driving

Human Observation-Inspired Trajectory Prediction for Autonomous Driving in Mixed-Autonomy Traffic Environments

Trajectory Prediction for Autonomous Driving Based on Multiscale Spatial‐temporal Graph

DACR-AMTP: Adaptive Multi-Modal Vehicle Trajectory Prediction for Dynamic Drivable Areas Based on Collision Risk

Characterized Diffusion Networks for Enhanced Autonomous Driving Trajectory Prediction

Less is More: Efficient Brain-Inspired Learning for Autonomous Driving Trajectory Prediction

AI-TP: Attention-based Interaction-aware Trajectory Prediction for Autonomous Driving

What Truly Matters in Trajectory Prediction for Autonomous Driving?

Advancing Autonomous Driving Safety Through LLM Enhanced Trajectory Prediction

Trajectory Prediction for Autonomous Driving with Topometric Map

Trajectory Prediction in the Complex Junction Scene Using Information Fusion and Behavior Pattern Recognition

BAT: Behavior-Aware Human-Like Trajectory Prediction for Autonomous Driving

MDSTF: a multi-dimensional spatio-temporal feature fusion trajectory prediction model for autonomous driving

Rapid Vehicle Trajectory Prediction Based on Multi-Attention Mechanism for Fusing Multimodal Information

An Integrating Comprehensive Trajectory Prediction with Risk Potential Field Method for Autonomous Driving