Solving multidimensional knapsack problems with generalized upper bound constraints using critical event tabu search

A critical event tabu search method which navigates both sides of the feasibility boundary has been shown effective for solving the multidimensional knapsack problem. In this paper, we apply the method to the multidimensional knapsack problem with generalized upper bound constraints. This paper also demonstrates the merits of using surrogate constraint information vs. a Lagrangian relaxation scheme as choice rules for the problem class. A constraint normalization method is presented to strengthen the surrogate constraint information and improve the computational results. The advantages of intensifying the search at critical solutions are also demonstrated.