Improved Heuristic Flower Pollination Algorithm for Solving Multi-Dimensional Knapsack Problems

The optimization algorithm plays an important role in solving the complex problems, and many complex problems can be modeled as a combinatorial optimization problem. The multi-dimensional knapsack problem is a kind of typical combinatorial optimization problem. The pollination algorithm is a kind of natural heuristic algorithm proposed in recent years, which has the characteristics of few parameter tunings and high convergence rate. In this paper, we propose an improved hereditary pollination algorithm (IHFPA), which is suitable for solving multidimensional knapsack problem. The hybrid heuristic repair strategy and simulated annealing mechanism is introduced to solve the knapsack problem, which has two advantages: retaining the characteristics of the original algorithm's high convergence rate, and increasing the diversity of the population with the result of improving the global search ability. At the end, IHFPA is simulated in comparison with quantum genetic algorithm (QGA), cuckoo algorithm (BCS) and particle swarm optimization algorithm (BPSO) for eight multi-backpack test problems of ELIB database. The experimental results show that the robustness and accuracy of the proposed algorithm are generally superior to those of the other three algorithms, which suggests IHFPA can effectively solve the combinatorial optimization problem of multiple backpack problems.