SGNetPose+: Stepwise Goal-Driven Networks with Pose Information for Trajectory Prediction in Autonomous Driving

2025-03-11

AI Summary

To address insufficient pedestrian trajectory prediction accuracy and underutilization of pose information in autonomous driving, this paper proposes a stepwise goal-driven network integrating skeletal keypoints and joint angles. Methodologically, it introduces— for the first time—the fusion of human pose angular features with a target-oriented, stepwise modeling mechanism, augmented by horizontal temporal flipping to enhance few-shot generalization. Skeletal sequences are extracted via pose estimation, and inter-segment joint angles are computed; the SGNet architecture is enhanced to enable multimodal spatiotemporal feature fusion and end-to-end trajectory regression. Evaluated on the JAAD and PIE datasets, the method achieves state-of-the-art performance, significantly outperforming the original SGNet. It delivers substantial improvements in both trajectory prediction accuracy and timeliness of collision warning, demonstrating superior practical utility for safety-critical autonomous driving applications.

Technology Category

• Humans and AI: Human-Aware Planning and Behavior Prediction
• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Search and Optimization: Learning to Search

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web query analysis, representation and understanding
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Predicting pedestrian trajectories is essential for autonomous driving systems, as it significantly enhances safety and supports informed decision-making. Accurate predictions enable the prevention of collisions, anticipation of crossing intent, and improved overall system efficiency. In this study, we present SGNetPose+, an enhancement of the SGNet architecture designed to integrate skeleton information or body segment angles with bounding boxes to predict pedestrian trajectories from video data to avoid hazards in autonomous driving. Skeleton information was extracted using a pose estimation model, and joint angles were computed based on the extracted joint data. We also apply temporal data augmentation by horizontally flipping video frames to increase the dataset size and improve performance. Our approach achieves state-of-the-art results on the JAAD and PIE datasets using pose data with the bounding boxes, outperforming the SGNet model. Code is available on Github: SGNetPose+.

Problem

• Predict pedestrian trajectories for autonomous driving safety.
• Integrate skeleton and pose data for accurate trajectory prediction.
• Enhance SGNet architecture to improve collision prevention and efficiency.

Innovation

• Integrates skeleton data with bounding boxes
• Uses pose estimation for trajectory prediction
• Applies temporal augmentation to video frames