Abstract:Vehicle trajectory prediction is crucial for advancing autonomous driving and advanced driver assistance systems (ADAS). Although deep learning-based approaches - especially those utilizing transformer-based and generative models - have markedly improved prediction accuracy by capturing complex, non-linear patterns in vehicle dynamics and traffic interactions, they frequently overlook detailed car-following behaviors and the inter-vehicle interactions critical for real-world driving applications, particularly in fully autonomous or mixed traffic scenarios. To address the issue, this study introduces a scaled noise conditional diffusion model for car-following trajectory prediction, which integrates detailed inter-vehicular interactions and car-following dynamics into a generative framework, improving both the accuracy and plausibility of predicted trajectories. The model utilizes a novel pipeline to capture historical vehicle dynamics by scaling noise with encoded historical features within the diffusion process. Particularly, it employs a cross-attention-based transformer architecture to model intricate inter-vehicle dependencies, effectively guiding the denoising process and enhancing prediction accuracy. Experimental results on diverse real-world driving scenarios demonstrate the state-of-the-art performance and robustness of the proposed method.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to improve the accuracy of vehicle trajectory prediction in autonomous driving and advanced driver - assistance systems (ADAS), with a particular focus on vehicle following behavior and vehicle - to - vehicle interactions. Although deep - learning - based methods, especially those using the Transformer architecture and generative models, have significantly improved prediction accuracy in capturing complex nonlinear patterns in vehicle dynamics and traffic interactions, they often overlook detailed following behavior and vehicle - to - vehicle interactions that are crucial for practical driving applications, especially in fully autonomous or mixed - traffic scenarios. To address this challenge, this study introduces a Scaled - Noise - Conditioned Diffusion Model for following - trajectory prediction (FollowGen), which integrates detailed vehicle - to - vehicle interactions and following dynamics into a generative framework, thereby improving the accuracy and rationality of the predicted trajectories. Specifically, the main contributions of the paper include: - Proposing FollowGen, a new generative framework for predicting vehicle trajectories in following - scenarios, incorporating detailed vehicle - to - vehicle interactions. - Developing a temporal - feature - encoding pipeline, consisting of GRU layers, a position - based attention mechanism, and Fourier embeddings, to effectively extract the temporal features of historical vehicle trajectories. - Proposing a noise - scaling strategy that conditions isotropic Gaussian noise on the historical motion - feature encodings of vehicles. The scaled noise replaces the isotropic noise in the diffusion process. - Modeling the dynamics between following vehicles through a cross - attention - based Transformer architecture, guiding the extracted interaction embeddings into the denoising network to direct the trajectory - generation process. - Verifying the robustness and generality of FollowGen in multiple real - world scenarios through comparative and ablation studies, including cases of HV following HV, HV following AV, and AV following HV. Through these methods, FollowGen aims to more accurately capture micro - interactions in following - scenarios, especially in dense - traffic scenarios, which are crucial for maintaining safe and efficient traffic flow.

FollowGen: A Scaled Noise Conditional Diffusion Model for Car-Following Trajectory Prediction

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

Crossfusor: A Cross-Attention Transformer Enhanced Conditional Diffusion Model for Car-Following Trajectory Prediction

A multi-modal vehicle trajectory prediction framework via conditional diffusion model: A coarse-to-fine approach

Trajectory Prediction and Risk Assessment in Car-Following Scenarios Using a Noise-Enhanced Generative Adversarial Network

TransFollower: Long-Sequence Car-Following Trajectory Prediction through Transformer

Hierarchical vector transformer vehicle trajectories prediction with diffusion convolutional neural networks

A Car-Following Model Based on Trajectory Data for Connected and Automated Vehicles to Predict Trajectory of Human-Driven Vehicles

Capturing Car-Following Behaviors by Deep Learning

IDM-Follower: A Model-Informed Deep Learning Method for Long-Sequence Car-Following Trajectory Prediction

Characterized Diffusion Networks for Enhanced Autonomous Driving Trajectory Prediction

Car-Following Model Based on Deep Learning and Markov Theory

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

EquiDiff: A Conditional Equivariant Diffusion Model For Trajectory Prediction

Car-following Informed Neural Networks for Real-Time Vehicle Trajectory Imputation and Prediction

DiffTAD: Denoising diffusion probabilistic models for vehicle trajectory anomaly detection

Intention-aware Denoising Diffusion Model for Trajectory Prediction

DeepCF: A Deep Feature Learning-Based Car-Following Model Using Online Ride-Hailing Trajectory Data

An Integrated Car-Following and Lane Changing Vehicle Trajectory Prediction Algorithm Based on a Deep Neural Network

Controllable Diverse Sampling for Diffusion Based Motion Behavior Forecasting