Decentralized Multi-Agent Urban Traffic Management via Spatio-Temporal Mobility Profile Planning

Local Rules, Global Efficiency: Emergent Coordination on Large-Scale Road Networks

Urban transportation systems face a fundamental efficiency paradox: individually optimal routing decisions can create system-wide congestion that increases travel time for all participants. Traffic engineers address this through rigid infrastructure solutions—for example, roundabouts that enforce circular flow, or turn restrictions that prevent conflicting maneuvers. However, these manually designed interventions lack flexibility: optimized for peak demand, they unnecessarily constrain traffic during off-peak hours. Autonomous vehicles (AVs) offer a transformative opportunity for dynamic coordination that adapts to traffic conditions. However, achieving such adaptive coordination at city scale with thousands of decentralized vehicles remains an open challenge. We propose Guided Local Coordination (GLC), a decentralized framework combining global shortest-path guidance with priority-based local conflict resolution. GLC precomputes optimal distance labels for the entire network, then uses priority inheritance and backtracking to resolve real-time collisions without replanning. We demonstrate that GLC scales to city-scale networks with 4K nodes and 10K agents, achieving high performance. As a key qualitative result, GLC produces emergent roundabout-like behavior: under high density, left-turn traffic spontaneously forms circulating flows while straight-through traffic proceeds unimpeded—adaptive coordination arising purely from local rules without explicit programming.

Multi-Agent Coordination Framework for Autonomous Delivery Robots in Dense Urban Environments

The proliferation of last-mile autonomous delivery fleets requires robust, scalable, and communication-efficient multi-agent coordination frameworks to safely navigate dense urban environments. Traditional multi-agent pathfinding approaches frequently scale poorly under high agent density or suffer severe performance degradation during sudden communication dropouts. To resolve these operational challenges, this paper presents a novel distributed hybrid coordination framework that integrates macroscopic consensus-based task allocation with localized, dynamic conflict resolution strategies. By implementing a decentralized token-passing auction model alongside asynchronous dynamic window path updates, the system guarantees conflict-free trajectories without relying on a persistent, centralized server. Extensive software co-simulations and physical field trials demonstrate that the proposed framework achieves a 22.4% reduction in path conflict frequency and a 16.8% improvement in fleet resource utilization compared to baseline prioritized planning models. These results prove that the system is highly resilient and viable for high-density, real-world autonomous logistics infrastructures.

RACER: Real-Time Adaptive Congestion-Aware Emergency Routing in Urban Vehicular Networks

This article proposes a new approach to routing, termed RACER (Real-time Adaptive Congestion-aware Emergency Routing), which dynamically responds to changing traffic conditions without requiring additional traffic-signal-control infrastructure, relying instead on congestion information obtained through standard vehicle-to-infrastructure telemetry.

An Agent-Based Crowd Simulation Framework Integrating Urban Walkway Networks and Individual Pedestrian Behaviors for Congestion Analysis

: High-density crowd accidents in complex urban walkway networks represent significant public safety risks characterized by non-linear dynamics and unpredictable emergence. While existing crowd models have primarily focused on evacuation within controlled facilities, they often lack the capacity to explain how behavioral heterogeneity interacts with real-world walkway networks to generate localized congestion. This paper proposes an Agent-Based Modeling (ABM) framework specifically designed to analyze the impact of diverse pedestrian behavior compositions on network-wide congestion dynamics. The proposed framework adopts a centralized time-stepped execution structure that integrates complex movement factors, including multi-destination trip chains and density-responsive navigation. To demonstrate its practical utility, the framework is applied to a realistic urban scenario in the Itaewon district, Seoul, utilizing empirical datasets on transportation demand and network topology. The study investigates how varying the distribution of behavior profiles—such as shortest-path, route-choice, and crowd-avoid strategies—reshapes emergent crowding patterns. The results highlight the framework’s strength as a robust scenario-based analysis tool for identifying systemic risks and providing actionable insights for crowd management in dense urban environments.

UrbanMix: LLM-Guided Simulation of Mixed Autonomy Traffic with Heterogeneous Behavioral Profiles

DT-MR-FALCON: Digital Twin–Assisted Mixed Reality Emergency Corridor Optimization for Intelligent Ambulance Navigation

Urban traffic congestion critically impairs emergency medical services (EMS) response times, often preventing ambulances from reaching patients within the life-saving “golden hour.” Existing traffic management systems are predominantly reactive and infrastructure-focused, lacking integrated support for real-time emergency vehicle navigation. Although reinforcement learning-based signal control and Digital Twin modeling have each demonstrated promise independently, their separate deployment fails to deliver coordinated, predictive, and driveraware emergency routing. This paper presents DT-MR-FALCON, a unified framework for Emergency Corridor Optimization (ECO) that simultaneously addresses predictive traffic modeling, distributed signal coordination, and driver-centric navigation. ECO is formally defined as a dynamic, congestion-sensitive path optimization problem on urban road networks. The proposed solution integrates: (i) a Digital Twin for short-horizon traffic state forecasting, (ii) a Federated Multi-Agent Reinforcement Learning (FMARL) framework for scalable, privacy-preserving signal coordination, and (iii) a Mixed Reality (MR) interface for real-time visualization of dynamically generated emergency corridors. The framework establishes a closed-loop system coupling prediction, optimization, and human-centered decision-making, supported by theoretical guarantees on corridor optimality and delay reduction under bounded prediction error. Large-scale SUMO simulations on real-world urban networks demonstrate that DT-MR-FALCON reduces average intersection delay by 35.0%, queue length by 37.5%, and ambulance travel time by 46.2% relative to fixed-time control, achieving a 95% corridor-clearance success rate.