Abstract: In the context of constructing digital twin cities and road traffic systems, this paper analyzes the concept and challenges of digital twins, exploring the bidirectional mapping essence behind it. A digital twin theory and system based on multivariate composite mapping is derived. Considering the professional characteristics of road traffic and the diversified needs for data and engine integration, a road traffic digital twin system called "endpoint-edge-cloud-IoT-twinbase-engine" is proposed. This system emphasizes building bidirectional mapping and interaction between the physical entities of "human-vehicle-road-environment" and comprehensive digital models through an independent middle layer supporting data and engine adaptation. Case studies confirm that this system is feasible and effective.