Search

Sequential auto carrier loading is a postprocess of the auto carrier loading problem (ACLP) and can be formulated as an integer programming problem for each auto dealer along the auto carrier route except the dealer that is the last stop. The associated objective function for each individual subproblem is to minimize the number of reloads and, for those vehicles that are reloaded, to minimize the total arcs in the path from the reassigned slots to the exit slot. The sequential ACLP (SACLP) process also relaxes the assumption made in the literature that if at any auto dealer a vehicle is unloaded and reloaded again, it is reloaded back to its original slot. To solve the SACLP, an iterative heuristic algorithm that takes into consideration the effect of empty slots is proposed. At each auto dealer, after its demand has been unloaded, any unloaded vehicles destined for the upcoming dealers are reloaded back to better unassigned slots on the auto carrier and are subject to various loading constraints. Computational experiments on five instances indicated that use of the SACLP improved the solutions obtained from the ACLP by itself.

Travel time information is generally nonlinear and nonstationary in a dynamic environment, and therefore no consistent tendency can be easily observed. This research developed a novel approach that combined the empirical decomposition method for speed data analysis and gray theory for travel time prediction to predict the arrival time at each stop along a bus route. In addition, sensitivity analysis was performed for different numbers of stops. With an average prediction error of less than 3.5%, the experiments showed that the proposed prediction approach, which employed both historical and real-time speed data collected from the geographic positioning system, outperformed Chou's approach, which used only historical speed data. The proposed prediction method could be readily incorporated into a cell phone–based information retrieval system that indicated bus position en route as well as its arrival times at all stops.

The reverse supply chain network design problem is formulated as a bilevel model. In the upper level, the total cost incurred in the reverse supply chain network is minimized subject to the budget constraint, among others, whereas in the lower level the corresponding flow pattern in the network is characterized by the variable demand equilibrium conditions; this is essentially a variation of the Wardrop second principle. A sensitivity analysis–based algorithm, with special treatment of the material conversion factor at each recycler processor agent, or both, is proposed for solutions. A numerical example is then provided for demonstration.

One link-based algorithm (Frank-Wolfe) and three route-route based algorithms are compared for use in the dynamic user-optimal route choice problem. Computational experience with five test networks indicates that the route-based algorithms are in general superior to the link-based algorithm in terms of execution time but slightly inferior in terms of memory requirement.

Link capacity side constraints are explicitly incorporated into the variational inequality-based dynamic user-optimal route choice problem. The corresponding dynamic capacitated user-optimal conditions for each origin-destination pair and time interval state that the generalized route travel times (composed of route travel times and queuing delays) experienced by travelers are equal and minimal. An augmented Lagrangian dual method that embeds the gradient projection algorithm is then proposed and validated with a numerical example. The computational performance of the gradient projection algorithm is also compared with the disaggregate simplicial decomposition algorithm in a preliminary attempt.