Traveling Salesman Problems

Directory Structure

TSP/
  app/               contains algorithms and features
  data/              contains data instances
  distanceMatrices/  contains distance matrix files
  documentation/     contains documentation for the algorithms
  evaluation/        contains evaluation of algorithms
  logs/              contains log files
  optimalResults/    contains optimal results
  output/            contains exported solution files
  parameters/        contains parameter files
  statistics/        contains statistical analysis of the data instances

Data Description

Data http://comopt.ifi.uni-heidelberg.de/software/TSPLIB95/tsp/

Large data http://www.math.uwaterloo.ca/tsp/world/countries.html

Data description http://comopt.ifi.uni-heidelberg.de/software/TSPLIB95/tsp95.pdf

Data FAQ http://comopt.ifi.uni-heidelberg.de/software/TSPLIB95/TSPFAQ.html

One can download the datasets .tsp, .opt.tsp from here by:

For Windows:
- Install winrar and set path:
```
set path="path\to\WinRAR\";%path%
```
- Run getData-Win.bat:
```
> getData-Win.bat
```

For Linux/MacOS:

Install curl:
```
> sudo apt install curl
```
Run getData-Unix.sh:
```
> bash getData-Unix.sh
```

Instances Description

The file dataInstances.json contains general statistics on some basic information about the problems (or instances):

NAME: name of the problem
DIMENSION: the number of cities in the problem
BEST_KNOWN_SOLUTION: the best value of a solution has been found from all sources (currentely based on TSPLIB_BestKnownSolution.json file in statistic/instances)
EXIST_OPTIMAL_TOUR: true if the optimal tour for the problem has been found and put in optimalResults folder, else false
GIVEN_COORDINATES: true if the problem provides coordinates, else false

Remark on datasets

The problems whose dimensions are greater than or equal to 10000 are not considered.
The two problems lin318.tsp and linhp318.tsp have the same name and data.
Therefore, the linhp318.tsp problem is not considered.
The problem dsj1000.tsp has CEIL_2D as its EDGE_WEIGHT_TYPE. Meanwhile, the TSPLIB_BestKnownSolution.json contains two records of the dsj1000.tsp problems: dsj1000EUC_2D and dsj1000CEIL_2D with EUC_2D and CEIL_2D as EDGE_WEIGHT_TYPE respectively.
Therefore, the record dsj1000EUC_2D is removed and the record dsj1000CEIL_2D is renamed to dsj1000.

References

Edmund K. Burke, Graham Kendall. "Search Methodologies", 2014, 2nd Edition, Springer

Tabu Search Method for Solving the Traveling salesman Problem - Isra Natheer Alkallak, Ruqaya Zedan Sha’ban

David L. Applegate, Robert E. Bixby, Vaˇsek Chvátal, and William J. Cook. "The Traveling Salesman Problem: A Computational Study", Princeton University Press

Installation

Note: The directory in Linux is different from that in Windows. Therefore, if errors with '/' and '\' encountered, the path need to be modified.

Example:

On Linux: https://github.com/godfeed27
On Windows: https:\\github.com\godfeed27

Clone repository

First of all, you need to download the repository. You can either run the script below on the command-line or terminal:

git clone https://github.com/godfeed27/TSP.git

or download zip file of repository and unzip.

If you have the problem related to personal access token, try following the steps below:

Log in to your GitHub account
Go to Settings, then choose Developer Settings
Choose Personal access tokens
Generate and copy your tokens
Clone again and paste tokens to password space

Set PYTHONPATH

Add TSP/ directory pathname to PYTHONPATH.

For Windows:

Run command line as administrator and execute the following command:
```
 setx PYTHONPATH "%PYTHONPATH%;\path\to\TSP" /M
```
Note:
- After run this command, you need restart terminal to update PYTHONPATH.
- Windows can't set PYTHONPATH for the current terminal session, so this command sets PYTHONPATH permanently.
For Linux/MacOS:
```
export PYTHONPATH=$PYTHONPATH:/path/to/TSP
```
Note: This command only set PYTHONPATH for the current terminal session.

Change directory

Change the path that points to your TSP folder.

cd path/to/TSP

Create python virtual environment:

Create environment by following command:
```
python3 -m venv env
```

Activate environment:

env/Scripts/activate # For windows
source env/bin/activate # For Linux or MaxOS

Install libraries in by:
```
pip install -r requirements.txt
```

Deactivate environment:

env/Scripts/deactivate # For windows
source env/bin/deactivate # For Linux or MaxOS

Create parameter file

In folder parameters/, one has to create parameter files based on the corresponding sample parameter files:

TSP_Folder: path to the TSP directory.
PROBLEM_NAMES: problem names seperated by commas.
ALGORITHM_NAMES: algorithm abbreviations seperated by commas.
TIME_LIMIT_OPTIONS: time limits that depend on the dimensions of the problems.
For example,
```
TIME_LIMIT_OPTIONS:100~60|1000~600|10000~600|100000~1200
```
is equivalent to:

Number of Cities Time Limit

1 <= n <= 100 60 s

100 < n <= 1,000 600 s

1,000 < n <= 10,000 600 s

10,000 < n <= 100,000 1200 s

Note: The last bound's dimension must be greater than the greatest dimension in the problems.
DATA_FOLDER: path to the folder that contains data instances.
Highly recommend setting data as default.
DATA_INSTANCES_FOLDER: path to the folder that contains the statistics of the data instances.
Highly recommend setting statistics/instances/ as default.
DISTANCE_MATRICES_FOLDER: path to the folder that contains the distance matrix files.
Highly recommend setting distanceMatrices as default.
EVALUATE_IMAGE_FOLDER: path to the folder that contains the evaluation images.
LEADER_BOARD_FOLDER: path to the folder that contains the algorithms leader board.
OPTIMAL_RESULTS_FOLDER: path to the folder that contains the optimal result files.
Highly recommend setting optimalResults as default.
OUTPUT_FOLDER: path to the output folder.
Highly recommend setting output as default.
LOGS_FOLDER: path to the report folder that contains log files.
STATISTICS_IMAGE_FOLDER: path to the folder that contains the statistical images.
EXPORT_SOLUTION: decides whether the solution is exported or not (yes or no).
DATE_RUN: date when script was run (format: yyyymmdd).
VERBOSE: decides how much detailed the output should be.
If VERBOSE = 0, prints nothing.
If VERBOSE = 1, prints header and final result only.
If VERBOSE = 2, prints the whole algorithm's process.
Algorithm parameters: parameters' values for the algorithms.

Number of Cities	Time Limit
1 <= n <= 100	60 s
100 < n <= 1,000	600 s
1,000 < n <= 10,000	600 s
10,000 < n <= 100,000	1200 s

Note: Default parameters' values are given in the sample parameters files in the parameters/ folder.

Features

1. Create distance matrix files

One can generate and save distance matrix files locally by running the following:

python3 app/feature/saveDistanceMatrix.py path/to/parameter.txt

The distance matrix files (e.g. a280.matrix.tsp) will then be generated and saved in DISTANCE_MATRICES_FOLDER (in the parameter file) folder.

2. Create json file that stores data information

One can get data instances' information summary in a json file dataInstances.json by running the following:

python3 app/feature/getDataInstances.py path/to/parameter.txt

The dataInstances.json will then be saved in dataInstances (in the parameter file) folder.

3. Draw statistical analysis

One can get statistical analysis of the problems on dimensions, coordinates and optimal tours by running the following:

python3 app/feature/statisticsPlot.py path/to/parameter.txt

The statistical analysis will then be saved in STATISTICS_IMAGE_FOLDER (in the parameter file) folder.

4. Validate optimal results

One can check if the optimal results are reasonable by running the following:

python3 app/feature/checkOptimalResults.py path/to/parameter.txt

The result will then be printed to the output.

5. Solve TSP

One can solve the TSP problems using the algorithms above by running the following:

python3 app/feature/problemSolve.py path/to/parameters.txt

The results will then be saved in OUTPUT_FOLDER (in the parameter file) folder.

Note:

The distance matrix files should be generated locally beforehand (feature 1) for faster performance.
Rerunning the same parameter file will not guarantee the same solution

6. Evaluate results

One can evaluate the results after solving by running the following:

python3 app/feature/evaluateResults.py path/to/parameters.txt

The algorithms' results comparison and the leader board for the algorithms will then be saved in STATISTICS_IMAGE_FOLDER and LEADER_BOARD_FOLDER (in the parameter file) folders , respectively.

7. Check distinct optimal tours

In the TSPLIB95 data description, there is a remark:

There may exist several shortest tours for a problem instance.

One can check if a solution tour with the optimal total distance is different from the optimal tour by running the following:

python3 app/feature/checkExistDistinceOptTours.py path/to/parameters.txt

The result will then be printed to the output.# TSP

godfeed27 / tsp Goto Github PK

tsp's Introduction