Awesome-Interaction-Aware-Trajectory-Prediction

Awesome Interaction-Aware Trajectory Prediction

The Awesome Interaction-Aware Trajectory Prediction project serves as a comprehensive resource for those interested in the latest advancements in trajectory prediction technology. The focus of this project is to aggregate and curate the latest research materials, which include datasets, blogs, research papers, and public codebases. These resources are beneficial for both academic researchers and industry professionals interested in trajectory prediction.

Project Maintainers

The project is maintained by distinguished individuals from renowned institutions:

• Jiachen Li from Stanford University
• Hengbo Ma and Jinning Li from the University of California, Berkeley

They are open to collaboration and invite others to contribute by pulling requests to add new resources or by initiating discussions through email.

Citing the Project

The project encourages others to cite their work if they find the repository useful. Key publications by the maintainers include:

• "EvolveGraph: Multi-Agent Trajectory Prediction with Dynamic Relational Reasoning" published in NeurIPS 2020.
• "Conditional Generative Neural System for Probabilistic Trajectory Prediction" presented at IROS 2019.

Table of Contents

The repository is well-structured and divided into various sections for ease of navigation:

• Datasets: This section includes a wide range of datasets categorized into vehicles and traffic, pedestrians, and sport players.
• Literature and Codes: This section offers survey papers and studies on various aspects like intelligent vehicles, pedestrians, and sport player interactions.

Key Sections

Datasets

1. Vehicles and Traffic: Datasets in this category provide comprehensive data on various agents such as vehicles, cyclists, and pedestrians captured in urban, highway, or mixed scenarios using sensors like LiDAR, cameras, and radar.

2. Pedestrians: This includes datasets focused on people in various environments such as urban streets, shopping centers, and stations. Sensors primarily include cameras.

3. Sport Players: Datasets featuring sports environments, such as football fields, basketball halls, and American football settings, captured mostly using cameras.

Literature and Codes

1. Survey Papers: This section compiles extensive research surveys covering topics like machine learning for trajectory prediction in autonomous vehicles, deep learning models, social interactions in autonomous driving, and motion prediction for pedestrians and vehicles.

2. Physics Systems with Interaction: The listed papers explore interaction networks and systems that model physical dynamics, relations, and interactions using neural networks and other AI techniques.

3. Intelligent Vehicles & Traffic & Pedestrians: A collection of cutting-edge research that offers innovations in multi-agent motion prediction and trajectory prediction for autonomous vehicles and pedestrian dynamics.

This project provides a structured approach to understanding and exploring the domain of trajectory prediction. Whether you are an academic looking to delve into state-of-the-art research or an industry professional seeking practical datasets and code, the Awesome Interaction-Aware Trajectory Prediction project is an invaluable resource hub.