| Summary: | One of the common problems a yard crane dispatcher in the container terminal faces every day is to complete the given jobs on time as it is predicted. As the quay cranes’ operations depend greatly on the arriving and departing of vehicles, the vehicle must make sure not to make any delays in their schedule. It is easy for the dispatcher to predict the vehicle arrival time based on the given job schedule. However some time it is difficult for the driver of a vehicle to arrive to the specific loading and unloading point at yard block on the predicted time as movement of vehicle is a dynamic process and incident may occur in the midst of travelling from point to point. Thus it makes it hard to predict the actual vehicle arrival time.
Two algorithms, MMT-RBA and MMT-localSeach, are presented to find the Yard Crane job sequence for serving a fleet of vehicle with actual arrival time. MMT-RBA is provable to find the optimal job sequence by using recursive backtracking with A* algorithm to re-compute remaining jobs that fall after the deviated job , whereas MMT-localSearch selects a few jobs around the deviated job and do a local search to find best sequence to minimize the maximum tardiness. Simulation of the deviation from predicted vehicle arrival time is computed by sampling a random value from a probability distribution with the predicted arrival time as the mean. The result of the experiment shows that recomputation by RBA performs better than no-computation and local search. However, based on statistical means, it shows that recomputation by RBA, the vehicle waiting time are statistically lower even though there is no statistical difference in the other performance indicators like tardiness, makespan and YC travelling distance.
|
|---|