Generative Modeling of Pedestrian Behavior: A Receding Horizon Optimization-Based Trajectory Planning Approach

Abstract

Urbanization is bringing together various modes of transport, and with that, there are challenges to maintaining the safety of all road users, especially vulnerable road users (VRUs). There is a need for street designs that encourages cooperation between road users. Shared space is a street design approach that softens the demarcation of vehicles and pedestrian traffic by reducing traffic rules, traffic signals, road marking, and regulations. Understanding the interactions and trajectory formations of various VRUs will facilitate the design of safer shared spaces. In line with this goal, this paper aims to develop a methodology for generating VRUs trajectories that accounts for behaviors and social interactions. We develop a receding horizon optimization-based pedestrian trajectory planning algorithm capable of modeling pedestrian trajectories in a variety of shared space scenarios. Focusing on three scenarios - group interactions, unidirectional interaction, and fixed obstacle interaction - case studies are performed to demonstrate the strengths of the resulting generative model. Additionally, generated trajectories are validated using two benchmark datasets – DUT and TrajNet++. The three case studies are shown to yield low or near-zero Mean Euclidean Distance and Final Displacement Error values supporting the performance validity of the models. We also analyze gait parameters (step length and step frequency) to further demonstrate the model’s capability at generating realistic pedestrian trajectories.