A reinforcement learning-enhanced discrete zebra optimization algorithm for solving the traveling salesman problem.

Area of Science:

• Computational Intelligence
• Operations Research
• Algorithm Design

Background:

• The Traveling Salesman Problem (TSP) is a computationally complex combinatorial optimization challenge.
• Existing algorithms often struggle to balance exploration and exploitation for optimal TSP solutions.

Purpose of the Study:

• To introduce a novel hybrid algorithm, the Reinforced Zebra Optimization Algorithm (RZOA), for solving the TSP.
• To enhance the decision-making process in optimization algorithms using Deep Reinforcement Learning.

Main Methods:

• Developed a discrete Zebra Optimization Algorithm (DZOA) with discrete operators for route updates.
• Integrated Deep Reinforcement Learning (DRL) using a Deep Q-Network (DQN) for adaptive operator selection.
• Implemented an experience replay memory and a carefully designed state-action-reward structure for balanced search behavior.

Main Results:

• RZOA demonstrated significant superiority over comparative algorithms on 42 TSPLIB benchmark datasets.
• Achieved near-optimal solutions with average Percentage Deviation of the Best solution (PDB) below 1% and average Percentage Deviation of the Average solution (PDA) below 0.5% for small/medium instances.
• Showcased robust performance and consistency, outperforming competitors in most cases with statistical validation.

Conclusions:

• The proposed RZOA offers an intelligent and efficient approach to complex combinatorial optimization problems like the TSP.
• The hybrid integration of ZOA and DQN provides adaptive, self-learning capabilities for enhanced optimization.
• RZOA presents a promising advancement in solving large-scale optimization challenges with high accuracy and stability.