A GUI Simulation for the Travelling Salesman Problem (TSP) and solving/approximation algorithms in comparison
The following algorithms are currently implemented:
- Exact:
- Brute-force (you guessed it)
- Dynamic-programming
- Heuristics for opening:
- Nearest neighbour
- Best nearest neighbour
- Minimum Spanning Tree (MST) Transform
- Optimization:
- k-Opt (currently only 2-opt)
For testing and benchmarking purposes TSPSim is connected to a fork of Heidelberg's infamous TSP Library, allowing you to import any TSP File and make local changes to it. The Library is hosted at tsplib.felixcornelius.de
You have 3 options to create a new TSP instance:
Generation pseudo-random vertices of a given number, adding vertices by mouse input or entering a weighted adjacency matrix (experimental. very experimental.)
A good practice strategy then would be to create a first tour with the Nearest neighbour method (or Best nearest neighbour, which is just independent of a starting point) and compare it with the transformed MST tour (click the stats checkbox for detailed tour info). The better one (which is NN in about 90% of all cases) can then be optimized by identifying and reconnecting crossing edges of vertex 2-pairs (2-Opt method)
- Import/Export instance to TSPLib compatible .tsp File
- Export instance as PNG Graphic (with optional alpha channel)
- Color/weight customization of vertices and edges
- Define a background image for any TSP instance.
(with basic image manipulation such as zoom, translation and opacity). A few maps are also included
- English localization
- Fix 2-Opt issues
- 3-Opt implementation
- Commandline version for speed