A Unified Multivariable Controller for Integrated Traffic Light and Speed Limit Regulation in Urban Systems via Adaptive Intelligence
Authors:
- Cristian IzaCICristian Iza
- David IzaDIDavid Iza
- Juan-Luis Posadas-YagüeJuan-Luis Posadas-Yagüe
- Jose-Luis Poza-LujanJPJose-Luis Poza-Lujan
Preprint
- 11 June 2026
- preprint
- Published by Elsevier in SSRN Electronic Journal
Abstract
Current urban traffic management systems present a critical limitation: the operational dissociation between infrastructure control (traffic signal synchronization) and vehicle dynamics (speed profiles). This work addresses this deficiency by introducing a unified multivariable controller that integrates traffic signal synchronization and dynamic speed regulation within a high-dimensional control framework. Unlike conventional systems, in which vehicles passively adapt to static infrastructure, this architecture applies a dual-action control policy in which a Deep Neural Network (DNN) simultaneously determines lane-level speed profiles and real-time traffic signal phase adjustments. Through this coordinated mechanism, traffic signals maintain authority over the temporal structure of the network, while speed regulation governs vehicle flow, consolidating both control functions into a single system, a capability generally absent in previous works. Our approach incorporates a richer set of input variables and controloriented metrics, enabling a more precise characterization of the system. This architecture leverages this enriched information to modify traffic regulations in real time. The methodology compares autonomous traffic signal control, independent speed limit regulation, and the proposed integrated controller. Implemented in Python and the Simulation of Urban MObility (SUMO) platform, the architecture employs an adaptive feedback loop to synchronize traffic lights and speed limits in real time. Experimental results demonstrate substantial improvements over traditional strategies, achieving a 25.55% reduction in queue lengths, a 24.98% improvement in traffic flow, and a 27.89% reduction in carbon dioxide (CO2 ). By consolidating heterogeneous control actions into a single computational engine, this framework provides a potential scalable foundation for future connected mobility.Keywords
This publication has 19 references indexed in Scilit:
- Regression machine learning-based highly efficient dual band MIMO antenna design for mm-Wave 5G application and gain predictionScientific Reports, 2025
- Parameterized-action based deep reinforcement learning for intelligent traffic signal controlEngineering Applications of Artificial Intelligence, 2025
- Multi-Agent Optimizing Traffic Light Signals Using Deep Reinforcement LearningIEEE Access, 2025
- Resonance frequency prediction of dielectric antennas for liquid sensing via support vector regressionScientific Reports, 2024
- MARVEL: Bringing Multi-Agent Reinforcement- Learning Based Variable Speed Limit Controllers Closer to DeploymentIEEE Access, 2024
- Stabilizing reinforcement learning control: A modular framework for optimizing over all stable behaviorAutomatica, 2024
- A Graph-Based Framework for Traffic Forecasting and Congestion Detection Using Online Images From Multiple CamerasIEEE Access, 2024
- Multibranch Machine Learning-Assisted Optimization and Its Application to Antenna DesignIEEE Transactions on Antennas and Propagation, 2022
- Short-Term Traffic Prediction With Deep Neural Networks: A SurveyIEEE Access, 2021
- Antenna Clustering for Simultaneous Wireless Information and Power Transfer in a MIMO Full-Duplex System: A Deep Reinforcement Learning-Based DesignIEEE Transactions on Communications, 2021