Genetic Algorithms for the Multiple Container Packing Problem

This paper presents two variants of Genetic Algorithms (GAs) for solving the Multiple Container Packing Problem (MCPP), which is a combinatorial optimization problem comprising similarities to the Knapsack Problem and the Bin Packing Problem. Two different representation schemes are suggested, namely direct encoding and order based encoding. While order based encoded solutions are always feasible, a repair algorithm is used in case of direct encoding to ensure feasibility. Additionally, local improvement operators have been applied to both GA variants. The proposed algorithms were empirically compared by using various sets of differently sized test data. Order based encoding performed better for problems with fewer items, whereas direct encoding exhibited advantages when dealing with larger problems. The local improvement operators lead in many cases not only to better final results but also to shorter running times because of higher convergence rates.