Research on improved ant colony optimization for traveling salesman problem

Area of Science:

• Artificial Intelligence
• Computer Science
• Operations Research

Background:

• The Traveling Salesman Problem (TSP) is a significant combinatorial optimization challenge.
• Existing Ant Colony Optimization (ACO) methods for TSP suffer from slow convergence and local optima.
• There is a need for improved algorithms to address TSP limitations.

Purpose of the Study:

• To propose an enhanced Ant Colony Optimization algorithm, GCNIACO, integrating Graph Convolutional Networks (GCNs).
• To improve the efficiency and effectiveness of solving the TSP.
• To enhance the algorithm's ability to find optimal solutions and escape local optima.

Main Methods:

• Integration of Graph Convolutional Networks (GCNs) to generate superior initial solutions.
• Conversion of GCN-generated solutions into pheromone trails for ACO.
• Adaptive dynamic adjustment of pheromone volatility.
• Incorporation of the 3-opt algorithm to improve local search capabilities.

Main Results:

• GCNIACO demonstrated enhanced pheromone guidance in the initial stages of the ACO process.
• The adaptive adjustments and 3-opt algorithm improved the algorithm's ability to escape local optima.
• Simulations on TSP datasets and an engineering example showed superior performance compared to classical algorithms.

Conclusions:

• The proposed GCNIACO algorithm offers a significant improvement for solving the Traveling Salesman Problem.
• The integration of GCNs and adaptive strategies effectively addresses the limitations of traditional ACO.
• GCNIACO provides a more robust and efficient approach to achieving optimal solutions for TSP.