The recent advancements in crowd dynamics and pedestrian simulation research are significantly enhancing our ability to model and predict human behavior in complex environments. A notable trend is the integration of real-time data with digital twin frameworks, which allows for more accurate and context-specific simulations, particularly in high-risk scenarios like air travel. This approach addresses the limitations of traditional models by incorporating data assimilation techniques and advanced machine learning algorithms to predict and manage crowd movements more effectively. Additionally, the development of enhanced driving risk fields and multimodal trajectory prediction models is improving the safety and efficiency of both autonomous and human-driven vehicles by quantifying and mitigating risks in real-time. Another innovative direction is the use of geometric graph neural networks to model human interactions in crowded spaces, leveraging domain-specific knowledge to improve trajectory predictions. These advancements collectively promise to revolutionize crowd management and safety in various public and transportation settings.