Abstract To properly evaluate the cold-start emissions of vehicles, four light-duty vehicles were selected to complete a Real Driving Emissions (RDE) test. The analysis and evaluation of the RDE test results reveal that the cold-start emissions account for a significant proportion of the urban trip emissions, and that the moving average window (MAW) method is not suitable for statistically processing the RDE test data that include the cold-start emission data. Additionally, three on-road emission tests including a cold-start test and two hot-start tests at different vehicle thermal states were designed to evaluate the effect of the initial vehicle thermal state on the vehicle start emissions. The results demonstrate that the thermal state of the TWC converter thermal state has a more direct and important effect on the cold-start emissions in the RDE test. Via a comparative analysis of the instantaneous characteristics of pollutant emission rates, fuel–air equivalent ratios, catalytic converter temperatures, vehicle trip dynamic parameters, etc., it is inferred that driving behavior has a more significant and sensitive impact on the cold-start emissions than on the non-cold-start emissions in the RDE test. Based on the RDE tests performed at different ambient temperatures, it is also indicated that the impact of the ambient temperature on the cold-start emissions is only a secondary factor relative to driving behavior. As a result, the trip dynamic conditions, specifically those for the cold-start period, should be set more reasonably and effectively in future RDE regulations to ensure the consistency of the cold-start emission test results.