Mitigating End Effects in the Dynamic Vehicle Allocation Model

A class of network models developed for optimizing the use of a fleet of vehicles can be viewed as time-staged optimization models. Such models which have linear objective functions suffer particularly from end effects, the distortions introduced by employing a finite planning horizon. This paper brings together the end effects literature and the vehicle allocation literature to find a finite linear programming problem that gives an approximation to the first stage optimal solution of the infinite model. In addition, we develop another method of approximation and show that both methods can be formulated as generalized networks. Numerical results are presented for each method for comparison.