Multi-ant colony optimization algorithm based on game strategy and hierarchical temporal memory model

To solve the problems of slow convergence and insufficient accuracy of traditional ant colony algorithms in solving large-scale problems, this paper proposes a multi-ant colony optimization algorithm based on game strategy and hierarchical temporal memory model (GHMACO). Firstly, the heterogeneous multi-ant colony model is constructed, and each colony collaborates to improve the performance of the algorithm. Secondly, in order to enhance the communication among the heterogeneous colonies, a non-cooperative game strategy is introduced. The heterogeneous ant colonies are divided into the propagating colony and the absorbing colonies, where the propagating colony propagates the optimal payoffs of the game, and the absorbing colonies choose optimal strategies for absorption to balance the convergence and diversity of the algorithm. Further, the hierarchical temporal memory model is adopted to perform hierarchical optimization strategies based on path memories which includes: local exploration strategy, pheromone redistribution strategy and path replacement strategy, thus improving the accuracy of the algorithm and helping the colonies to jump out of the local optimum. Experiments on the traveling salesman problem show that the improved algorithm is effective in improving the convergence and accuracy of the traditional ant colony algorithms, especially in large-scale problems.

Automated summary

This paper proposes a multi-ant colony optimization algorithm based on game strategy and hierarchical temporal memory model (GHMACO) to solve the problems of slow convergence and insufficient accuracy in traditional ant colony algorithms when solving large-scale problems. The algorithm improves the convergence and accuracy of traditional ant colony algorithms, especially in large-scale problems, by constructing a heterogeneous multi-ant colony model, introducing a non-cooperative game strategy, and adopting hierarchical optimization strategies based on path memories using the hierarchical temporal memory model.