Institute of Transportation, MOTC

       Developed countries are dedicated to terminal efficiency, and truck turnaround time is regarded as an important performance metrics. To reduce congestion at domestic container terminals, this study, based on international literature and field investigation, realizes why domestic terminals failed with truck appointment system was lack of motivation to encourage truck drivers to cooperate. Thus, it develops an approach with combination of machine learning and operational research for optimizing a truck appointment system considering empty-truck-trip problem. First, considering truck-company operations, similar export and import containers are matched in tuples with clustering analysis to reduce the number of empty trips and, then, tuples are regarded as parameters in optimization model. Based on non-stationary queuing theory, it estimates time-varying truck network of terminal managing truck arrivals and assigns truck companies to suitable time to minimize the numbers of trucks at peak hours.
       Numerical experiments related to domestic container terminals are conducted to illustrate the validity of the model. Results show that our optimization model can effectively reduce total truck waiting time up to 26% and find that our clustering analysis can achieve a time reduction of 28%. Besides, there is a bottleneck of terminal operations in yard caused by external trucks; and, it finds that total waiting time would be minimized when threshold of internal-truck queue length is 4, and that port-operator cost has a greater impact on overall costs. Furthermore, it shows that the number of visits to different yards would have a negative impact on truck congestion and that total cost would be minimized when double-move proportion is around 9-13%. It can be applied to related research units and container terminals.