Predicting Operating Speeds of Passenger Cars on Dual–Carriageway Road Tangents

This research develops multiple linear regression models for predicting operating speeds (V85) of passenger cars on dual–carriageway road tangents in Croatia, separately for the right (driving) and left (overtaking) lanes. Thirty-nine locations were analysed, with 26 locations used for model development and 13 for validation. Operating speeds were measured with a drone under free–flow conditions, ensuring consistency and accuracy of observations. Following a correlation analysis, ANOVA testing and multicollinearity diagnostics, stepwise regression was applied to identify statistically significant predictors from 14 infrastructural, traffic and environmental variables. The final models include factors such as total tunnel length, speed limit, lane width, longitudinal slope, average summer daily traffic (ASDT) and traffic flow density, with results differing between lanes. The right (driving) lane model achieved an explanatory power of R² = 0.82 (RMSE = 4.85), while the left (overtaking) lane model achieved R² = 0.71 (RMSE = 6.36). Validation on test locations confirmed the models’ predictive capability, with an average absolute deviation of 5.16% for the right lane and 4.75% for the left lane. The results provide practical approaches for evaluating the consistency of road designs, managing vehicle speeds and assessing safety, while laying the groundwork for future improvements to lane–specific prediction models.