<p>This study addresses the modeling and structured integration of multi-scale uncertainties in the Intelligent Vehicle Cyber-Physical System (IVCPS) for traffic optimization. Multi-spatial and multi-temporal uncertainties are analyzed across four levels (node, unit, system, region) and quantified using soft set theory. A macro–micro coupled optimization framework is developed. In the cyber space, regional-level congestion uncertainty is embedded into a multi-objective optimization model (along with fuel consumption and travel time) to generate recommended speeds at the road segment level. In the physical space, node-, unit-, and system-level uncertainties are incorporated into an improved Intelligent Driver Model (IDM) through parameter adaptation, enabling dynamic compensation of microscopic traffic behavior. Simulation results under both normal and accident scenarios show that the proposed approach significantly improves traffic efficiency while reducing congestion and fuel consumption.</p>