Abstract: Pedestrian motion modeling has become one of the hottest research topics in virtual reality field because of its important application potential in game industry, public safety and intelligent transport. This paper classifies current pedestrian motion modeling methods into four categories according to two dimensions: rule-based or data-driven, macro level or micro level. Firstly, typical models of each category are introduced. The latest research progresses and characteristics of these methods are emphasized. Secondly, the data sets and evaluation indicators commonly used in pedestrian motion modeling are introduced, and, different methods are further evaluated in collision avoidance, motion variety, scene generalization and simulation precision. Finally, challenges and future development of pedestrian motion modeling are analyzed in the background of new techniques like digital twin and meta universe.