Pedestrian Trajectory Simulation on University Campus Using Deep Inverse Reinforcement Learning: A Case Study of Southeast University Wuxi Campus

Campus users’ daily activities rely on on-campus facilities, making campus environmental organization crucial to their quality of life. Consequently, campus space optimization is a key research focus. However, existing methods, using pedestrian flow simulation tools based on spatial topology or deep-learning-based prediction models, often neglect environmental factors’ influence on behavior. Campus users’ behavior is goal-oriented, and ignoring environmental elements may lead to design misjudgments. Inverse reinforcement learning(IRL) can model human-environment interactions in real-world settings, offering a novel approach to pedestrian trajectory simulation. Thus, this study uses Southeast University Wuxi Campus as a case study, collecting real pedestrian trajectories and environmental data. A deep maximum entropy IRL method constructs a pedestrian trajectory simulation model. After validation, the model effectively simulates behavior under environmental influences and quantifies their impact, providing a useful tool for campus space optimization.