An Application of Particle Swarm Optimization for Solving Traveling Salesman Problem

For my graduation project, I implemented Particle Swarm Optimization (PSO) to solve the Traveling Salesman Problem (TSP) with 51 cities. The implementation features a flexible architecture where both the number of cities and their spatial coordinates can be modified through parameter adjustments. The algorithm employs velocity-position updates with social and cognitive components, utilizing Euclidean distance calculations for fitness evaluation. Key implementation aspects include: - Particle representation using permutation encoding for city sequences - Inertia weight adjustment for convergence control - Local and global best position tracking mechanisms - Swap sequence operators for maintaining valid tours The solution incorporates comparative analysis capabilities against Genetic Algorithm (GA) implementations, featuring tournament selection, ordered crossover, and swap mutation operators. This comparative framework enables comprehensive performance evaluation of PSO versus GA in terms of solution quality, convergence speed, and computational efficiency for TSP instances. Benchmarking metrics include optimal route distance calculation, computational time analysis, and convergence behavior tracking across multiple algorithm iterations. The modular design allows seamless testing with various city configurations and algorithm parameter settings.