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

2026-05-20 · Source: Takara TLDR - Daily AI Papers · Field: Transportation & Mobility — Autonomous Vehicles & Smart Transportation, Transportation Infrastructure, Artificial Intelligence & Machine Learning · Depth: Expert, medium

Summary

DeCoR is a two-stage reinforcement learning framework designed to co-optimize urban street crosswalk layouts and network-level signal control. It addresses the challenge of translating modern vision system outputs into effective urban design. The framework's design stage encodes the pedestrian network as a graph, learning a generative policy that parameterizes a Gaussian mixture model for crosswalk location and width. Subsequently, a shared control policy learns adaptive signal timings for each layout, aiming to minimize joint pedestrian and vehicle delay. Evaluated on a 750 m real-world urban corridor using demand data from video and Wi-Fi logs, DeCoR achieved significant improvements. It reduced pedestrian arrival time to their nearest crosswalk by 23% using fewer crosswalks than existing configurations. Furthermore, the control policy decreased pedestrian and vehicle wait times by 79% and 65% respectively, compared to fixed-time signalization, demonstrating robustness to new demands and layout changes without retraining.

Key takeaway

For urban planners and traffic engineers aiming to modernize city infrastructure, DeCoR presents a compelling reinforcement learning framework. You should consider adopting this two-stage co-optimization approach to simultaneously enhance pedestrian safety and vehicle flow. By leveraging real-time demand data, your teams can move beyond static designs and fixed-time signalization, potentially reducing pedestrian arrival times by 23% and overall wait times by up to 79%. This method offers a robust, adaptive solution for dynamic urban management.

Key insights

Reinforcement learning can co-optimize urban street design and traffic control for improved efficiency and safety.

Principles

• Co-optimization improves urban efficiency.
• Generative policies can design infrastructure.
• RL control policies generalize well.

Method

DeCoR uses a two-stage RL framework: a generative policy for crosswalk layout (Gaussian mixture model) followed by an adaptive signal control policy to minimize delays.

In practice

• Apply RL for urban infrastructure design.
• Integrate vision/Wi-Fi data for demand sensing.
• Develop adaptive signal timing systems.

Topics

• Reinforcement Learning
• Urban Planning
• Traffic Management
• Crosswalk Design
• Signal Control
• Smart Cities

Best for: AI Scientist, Research Scientist, AI Engineer

Related on AIssential

• Fusing Urban Structure and Semantics: A Conditional Diffusion Model for Cross-City OD Matrix Generation — Fusing urban semantics and spatial constraints via a conditional diffusion model enhances cross-city OD matrix generation accuracy and generalizability.
• How Cities Use AI to Improve Playground Design — AI and commercial-grade equipment are transforming urban playgrounds to enhance safety, development, and community engagement.
• Hybrid AI model combines graphs and transformers for real-time traffic forecasts — Urban congestion is a critical global issue, leading to commuter delays, economic inefficiencies, and tragically, a million deaths annually worldwide.
• Multi-Agent Reinforcement Learning for Safe Autonomous Driving Under Pedestrian Behavioral Uncertainty — Co-training SDCs with personality-driven MARL pedestrians significantly improves safety by exposing vehicles to realistic, uncertain human behaviors.
• Modelling Customer Trajectories with Reinforcement Learning for Practical Retail Insights — Reinforcement learning accurately models customer trajectories, outperforming heuristics for retail layout optimization.
• Momentum Based Reward Design for Low Emission Traffic Signal Control — A Momentum-Based Reward Function (MBRF) for DRL traffic control improves throughput-emission trade-offs and learning stability by prioritizing continuous vehicle movement.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Takara TLDR - Daily AI Papers.