Special Issue On Combining Optimization And Machine Learning Applications In Vehicle Routing Network Design And Crew Scheduling PDF Download

This study developed a mathematical model for optimizing vehicle routing and scheduling, which can be used to collect travel time information, and also to perform winter road maintenance operations (e.g., salting, plowing). The objective of this research was to minimize the total vehicle travel time to complete a given set of service tasks, subject to resource constraints (e.g., truck capacity, fleet size) and operational constraints (e.g., service time windows, service time limit). The nature of the problem is to design vehicle routes and schedules to perform the required service on predetermined road segments, which can be interpreted as an arc routing problem (ARP). By using a network transformation technique, an ARP can be transformed into a well-studied node routing problem (NRP). A set-partitioning (SP) approach was introduced to formulate the problem into an integer programming problem (I PP). To solve this problem, firstly, a number of feasible routes were generated, subject to resources and operational constraints. A genetic algorithm based heuristic was developed to improve the efficiency of generating feasible routes. Secondly, the corresponding travel time of each route was computed. Finally, the feasible routes were entered into the linear programming solver (CPL EX) to obtain final optimized results. The impact of travel time variability on vehicle routing and scheduling for transportation planning was also considered in this study. Usually in the concern of vehicle and pedestrian's safety, federal, state governments and local agencies are more leaning towards using a conservative approach with constant travel time for the planning of winter roadway maintenance than an aggressive approach, which means that they would rather have a redundancy of plow trucks than a shortage. The proposed model and solution algorithm were validated with an empirical case study of 41 snow sections in the northwest area of New Jersey. Comprehensive analysis based on a deterministic travel time setting and a time-dependent travel time setting were both performed. The results show that a model that includes time dependent travel time produces better results than travel time being underestimated and being overestimated in transportation planning. In addition, a scenario-based analysis suggests that the current NJDOT operation based on given snow sector design, service routes and fleet size can be improved by the proposed model that considers time dependent travel time and the geometry of the road network to optimize vehicle routing and scheduling. In general, the benefit of better routing and scheduling design for snow plowing could be reflected in smaller minimum required fleet size and shorter total vehicle travel time. The depot location and number of service routes also have an impact on the final optimized results. This suggests that managers should consider the depot location, vehicle fleet sizing and the routing design problem simultaneously at the planning stage to minimize the total cost for snow plowing operations.