DeCoR: Design and Control Co-Optimization for Urban Streets Using Reinforcement Learning

2026-05-20

AI Summary

This study addresses the inefficiency of pedestrian and vehicular traffic in urban streets by proposing a two-stage reinforcement learning framework that enables end-to-end co-optimization of crosswalk layout generation and signal control strategies for the first time. The approach models pedestrian networks using graph structures and employs a Gaussian mixture model to generate crosswalk layouts. A shared policy network facilitates adaptive signal control that generalizes across diverse layouts without requiring retraining. Evaluated on a real-world urban corridor, the method reduces pedestrians’ travel time to the nearest crosswalk by 23%, decreases waiting times for pedestrians and vehicles by 79% and 65% respectively, and achieves these improvements with fewer crosswalks, thereby significantly enhancing overall network efficiency.

Abstract

Modern vision systems can detect, track, and forecast urban actors at scale, yet translating perception outputs to urban design remains limited. We introduce DeCoR, a two-stage reinforcement learning framework that leverages flow observations to co-optimize crosswalk layout and network-level signal control. The design stage encodes the pedestrian network as a graph and learns a generative policy that parameterizes a Gaussian mixture model over crosswalk location and width, from which new crosswalks are sampled. For each layout, a shared control policy learns adaptive signal timings to minimize joint pedestrian and vehicle delay. On a 750 m real-world urban corridor with demand sensed from video and Wi-Fi logs, DeCoR learns a layout that reduces pedestrian arrival time to their nearest crosswalk by 23% while using fewer crosswalks than existing configurations. On the control side, DeCoR reduces pedestrian and vehicle wait time by 79% and 65%, respectively, relative to fixed-time signalization. Further, the control policy generalizes to demands outside of training and is robust to layout changes without retraining.

Problem

• urban design
• traffic signal control
• crosswalk layout
• reinforcement learning
• co-optimization

Innovation

• reinforcement learning
• urban design
• traffic signal control
• co-optimization
• generative policy