Nonparametric method for optimizing the start time of short-term road maintenance operations using real-time traffic map data

Short-term road maintenance activity is an important factor leading to nonrecurring congestion. To reduce such congestion, a nonparametric method is proposed to optimize the maintenance operation start time. A real-time traffic map is adopted as the traffic data source. This approach is of low cost, provides timely data updates, and includes wide road network coverage. Corresponding data acquisition and data conversion methods are proposed, and traffic data (e.g., speed and volume) are acquired under maintenance and normal conditions. A public travel open platform is used to obtain maintenance activity data. Based on these data, a nonparametric model, a support vector machine for regression (SVR), is established to estimate the traffic delay caused by maintenance operations. The optimal start time that leads to the minimum total cost can be obtained by assessing all possible operation start times and calculating the associated costs, such as maintenance, accident, and delay costs. The method is applied to the Shanghai expressway network to optimize the start time of maintenance operations. A total of 159 short-term maintenance activities and more than 300,000 corresponding real-time traffic records are considered, and continuously updating is performed. The SVR model is established and used to predict the traffic delay of actual maintenance activities with different operation start times. The results show that the proposed method is effective for start time optimization in complex scenarios. Compared with the traditional parametric method, the proposed method yields an increase in the delay estimation accuracy based on ground truth data.