A Parallel Intelligence Framework Integrating Digital Twins for Smart City Cyber-Physical-Social Systems: Architecture, Methodology, and Case Studies

Abstract

Introduction: The accelerating urbanization of the twenty-first century has transformed cities into hyper-connected cyber-physical-social systems (CPSS) whose complexity exceeds the descriptive capacity of classical modeling paradigms. Two complementary research traditions—parallel intelligence (PI) grounded in the ACP approach and digital twins (DT) based on high-fidelity virtual replication—have emerged in response but have evolved largely in isolation, with PI emphasizing social behavior and policy learning while DT emphasizes physical fidelity and real-time synchronization.

Objectives: This paper proposes PI-DT-CPSS, a unified framework that integrates parallel intelligence and digital twins to govern complex urban systems, organizing a smart city into five interoperating layers—physical, digital twin, parallel intelligence, social, and decision support—and formalizing their interaction through a closed feedback loop.

Methods: We detail the mathematical formulation of multi-fidelity twin models, a data fusion procedure combining Kalman filtering with deep residual correction, and a parallel execution strategy that combines model-predictive control with deep reinforcement learning.

Results: The framework is validated through three case studies: adaptive traffic signal control, renewable-dominant microgrid management, and urban emergency evacuation. Across these domains, simulation experiments show reductions of 19–34% in key performance indicators relative to state-of-the-art baselines, while preserving human-in-the-loop accountability.

Conclusions: We conclude by discussing theoretical contributions, deployment implications, and open challenges in privacy, fairness, and computational scalability