Urban traffic congestion poses increasing challenges for intelligent mobility systems, particularly as robotic platforms require reliable infrastructure-aware information for decision-making. This paper presents a cross-domain digital twin framework that integrates empirically grounded traffic demand modeling, multi-agent simulation, adaptive signal control, and synchronized 3D visualization within a unified architecture. Signalized intersections are modeled as interacting vehicle and traffic light agents operating over a graph-based network. Infrastructure-level control is formulated as a Markov Decision Process (MDP), enabling adaptive and interpretable signal phase allocation based on observed queue states. Simulation outputs-including congestion dynamics, delay metrics, and signal evolution-are exported and synchronized with a Unity-based environment, forming a spatially and temporally consistent digital twin. Comparative evaluation against a fixed-time baseline demonstrates reductions in waiting time and queue length, together with increased throughput under realistic demand conditions. The framework further establishes a structured interface through which robotic agents, such as neighborhood electric vehicles (NEVs), can consume predictive infrastructure information to support navigation decisions. The results demonstrate that integrating adaptive multi-agent control with digital twin synchronization provides measurable traffic-level improvements while enabling scalable cross-domain coordination between infrastructure intelligence and robotic autonomy.
