Greedy Local Improvement of SPEA2 Algorithm to Solve the Multiobjective Capacitated Transshipment Problem

We consider a multi-location inventory system where inventory choices at each location are centrally coordinated through the use of lateral Transshipments. This cooperation between different locations of the same echelon level often leads to cost reduction and service level improvement. However, when some locations face embarrassing storage capacity limits, inventory sharing through transshipment may cause undesirable lead time. In this paper, we propose a more realistic multiobjective transshipment model which optimizes three conflicting objectives: (1) minimizing the aggregate cost, (2) maximizing the fill rate and (3) minimizing the transshipment lead time, in the presence of different storage capacity constraints. We improve the performance of the well-known evolutionary multiobjective algorithm SPEA2 by adequately applying a multiobjective quasi-gradient local search to some candidate solutions that have lower density estimation. The resulting hybrid evolutionary algorithm outperforms NSGA-II and the original SPEA2 in both spread and convergence. It is also shown that lateral transshipments constitute an efficient inventory repairing mechanism in a wide range of system configurations.