traffic management

This study addresses key challenges in urban traffic congestion management—namely, the high cost of traditional modeling, data sparsity, and the difficulty of enforcing hard constraints—by proposing a novel traffic signal control approach that integrates behavioral systems theory with data-driven predictive control. The method operates without requiring an explicit system model, effectively handles sparse observational data, and rigorously enforces physical and operational constraints. Validated through microscopic traffic simulations in a high-fidelity closed-loop environment replicating Zurich’s road network, the approach significantly reduces total travel time and CO₂ emissions. These results demonstrate its potential as a new paradigm for intelligent traffic control.

This work addresses the challenges posed by heterogeneous road networks—such as diverse intersection layouts, varying signal coverage, and unknown topologies—by proposing a decoupled graph neural network (GNN) approach for traffic signal control. The method externalizes phase definition and timing logic from the learning module, employing a shared GNN to score traffic movements and mapping these scores to valid signal phases via a deterministic association matrix. This design decouples model parameters from both the action space of intersections and the graph size. Integrated with typed mean aggregation and proximal policy optimization (PPO), the proposed strategy demonstrates effectiveness on unseen synthetic grids and five real-world urban road networks, enabling transfer across heterogeneous networks, though it exhibits sensitivity under shifts in signal coverage distribution.

To address traffic congestion on signal-free suburban freeways exacerbated by growing commuter demand and constrained infrastructure, this paper proposes a vehicle-mounted distributed speed control protocol based on physics-informed reinforcement learning (PIL-RL). Methodologically, we introduce a novel corridor-level abstraction modeling framework that integrates macroscopic traffic flow dynamics, gap-acceptance theory, and high-fidelity microscopic simulation (PTV Vissim), embedding emergent traffic phenomena directly into the proximal policy optimization (PPO) algorithm. This enables fully decentralized, real-time velocity optimization with strong generalization across heterogeneous traffic conditions. Experimental evaluation on realistic network simulations demonstrates a 5% increase in total throughput, a 13% reduction in average delay, and a 3% decrease in stop occurrences—substantially improving traffic flow smoothness and robustness against congestion onset—without requiring any new physical infrastructure.

To address congestion at highway on-ramp merging zones, this paper proposes a cooperative ramp metering method integrating reinforcement learning (RL) and model predictive control (MPC). The approach embeds policy-gradient RL algorithms—specifically PPO and SAC—into a differentiable MPC framework built upon the Cell Transmission Model (CTM) of traffic flow. A stage cost function is designed to jointly optimize traffic states, ensure control smoothness, and enforce hard queue-length constraints. Crucially, the MPC optimization problem itself serves as a differentiable functional approximator for the RL policy, enabling end-to-end policy optimization while guaranteeing strict constraint satisfaction. Evaluated on a benchmark highway network, the method significantly reduces mainline congestion, achieves 100% compliance with queue constraints, and outperforms conventional MPC and state-of-the-art adaptive controllers across all metrics. Notably, it demonstrates superior robustness and adaptability under model mismatch and dynamic traffic demand fluctuations.

Traditional traffic signal control neglects high-order spatiotemporal dependencies among intersections, hindering real-time, network-wide optimization. To address this, we propose a multi-agent traffic signal control system for urban road networks: (1) an edge-cooperative perception architecture enables dynamic, multi-intersection data acquisition; (2) hypergraph learning is innovatively embedded into the critic network of a multi-agent soft Actor-Critic (MA-SAC) framework to explicitly model high-order spatiotemporal interdependencies among intersections; and (3) a spatiotemporal hypergraph encoder jointly encodes dynamic topology and temporal evolution. Evaluated on multiple real-world and simulated road network datasets, our method achieves up to a 32.7% reduction in average vehicle travel time, alongside significant improvements in throughput and system stability. The source code and training environment are publicly released.

This study addresses the limitations of prevailing urban logistics traffic models, which predominantly focus on macroscopic traffic flow and rely on black-box AI predictions, thereby hindering interpretable research with open parameters in supply chain contexts. Departing from the vehicle-flow-centric paradigm, this work adopts an individual-vehicle perspective and introduces a kinetic parameter framework grounded in Art.Kinema to characterize driving cycle features. By integrating high-frequency empirical vehicle speed data, the authors employ factor analysis and generalized linear models to quantify the impacts of exogenous factors—such as time of day, road type, and weather—on driving behavior. The resulting context-driven predictive model demonstrates strong goodness-of-fit and robustness, offering a theoretically sound foundation for transparent route planning and logistics decision-making.

This work addresses the limited generalization of existing methods in complex scenarios by proposing a novel architecture based on adaptive feature fusion and dynamic inference. The approach introduces a learnable context-aware weighting module to effectively integrate multi-scale semantic information and incorporates a lightweight dynamic network to allocate computational resources on demand. Experimental results demonstrate that the model significantly outperforms state-of-the-art methods across multiple benchmark datasets, achieving higher accuracy and robustness while maintaining inference efficiency. This study offers a new perspective on efficient and adaptive visual understanding, exhibiting strong theoretical value and practical potential.

This study addresses the deployment of reinforcement learning controllers in urban arterial signal networks to enhance traffic throughput. It systematically evaluates centralized, fully decentralized, and parameter-sharing decentralized multi-agent reinforcement learning strategies against the classical MaxPressure method in terms of capacity region and average travel time. The work proposes a novel, generalizable parameter-sharing decentralized architecture that enables agents to spontaneously generate coordinated "green wave" effects without explicit communication or coordination. Experimental results demonstrate that the proposed approach not only outperforms baseline methods significantly but also maintains superior performance when transferred to larger, previously unseen road networks, exhibiting both high efficiency and strong generalization capability.