Engineering‑guided multi‑objective optimization for sustainable asphalt pavement maintenance

Asphalt pavement maintenance involves balancing conflicting objectives among pavement performance, economic costs, and environmental impacts. Existing studies often neglect the dynamic influence of pavement condition on vehicle operational emissions and rely on optimization algorithms with limited engineering adaptability. This study develops a decision framework integrating performance prediction and multi-objective optimization. A Gaussian Process Regression (GPR) model is constructed to quantify pavement roughness deterioration. Subsequently, an improved Non-dominated Sorting Genetic Algorithm II (NSGA-II) is proposed, incorporating an engineering-prior-driven initialization strategy to align the search process with real-world distribution characteristics. Results indicate that the prediction model achieves high reliability, with a mean absolute error of 0.048 m/km and a prediction interval coverage probability of 96.3%. Statistical evaluation based on 30 independent runs demonstrates that the improved algorithm achieves median hypervolume and inverted generational distance values of 1.078 and 0.047, respectively. These results significantly outperform the standard algorithm (p < 0.001), confirming the robustness of the proposed enhancement. Sensitivity analysis shows that the optimal strategy region remains stable despite unit cost fluctuations. The final solution identified via entropy-weighted Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS) corresponds to the marginal benefit inflection point, providing a scientifically grounded and practically adaptable pathway for sustainable pavement management in the low-carbon era.