Traveling Salesman Problem (TSP)

Introduction

This is a Free and Open Source Software for Geospatial (FOSS4G) inspired tutorial. It focuses on Network Analysis using a Geographic Resources Analysis Support System (GRASS).

About the Tutorial

This tutorials' main aim is to be clear and understandable. I made every effort to use real data even though this presents many problems. However, it is a reality that data is dirty and software is faliable. This tutorial will be very time consuming to complete, however, you will learn alot along the way and this knowledge will easily transfer to future projects.

In this tutorial you will

• Install GRASS
• Download and prepare data
• Start a GRASS database
• Build a network and perform analysis on it

Note If the command line is not working and you don’t wish to take the time to troubleshoot this problem I have been given a great work around from my teacher Scott Mitchell. If you do not have GRASS folders set up refer to step named Getting on our way.

• 1. Launch the OSGeo4W
• 2. Type grass64 –gui and then hit enter on the keyboard. (notice there is a space between 64 and the dash)
• 3. You will see the old GRASS version load, so simply click File from the menu bar and click Quit GIS Manager.
• 4. Now in the command box type g.gui gui=wxpython&

Installation

Install OSGEO Package (30-60mins)

Quick Start

First, you will need to install free geospatial software. I am using GRASS 6.4.2. This can be found through the secure and trusted site of Open Source Geospatial Foundation OSGeo4w is Your Open Source Compass for Windows This package provides a set of open source geospatial software for a Win32 environment. Simply follow the directions from this site: Click Here to go to Download Page The great thing about this pkg is that is contains several versions of GRASS. So if something doesnt work in one version it might work in another.

Launch GRASS(1-1.5 hrs)

Launching GRASS Tutorial This tutorial will give you a simple walk through of GRASS allowing for an increased understanding of the software, though it is not necessary to complete it in full.

The Tutorial

Getting Started: GRASS Database (just minutes)

Grass Database

1. Prior to starting a GRASS session, create a New Folder named Grassdata and a second named nonGrassData, on your Windows Desktop. Then start GRASS and once the screen opens, browse to the Grassdata folder. Ensure all file and folder names contain NO spaces, that they are short, and do not begin with numbers.

2. Push the button for Georeferenced files, name the newLocation Ottawa and then browse to the nonGrassData folder. Select the Roadways.shp file. Now click define.

3. Click on both Ottawa and then PERMANENT in each of the grey boxes and then push the ENTER GRASS button. Now we are ready to start working in GRASS.

Next:Import data files and prepare them for processing and analysis

Importing .shp

1. You now have three screens open and running under GRASS. A command line black box, a GRASS GIS Layer Manager, and a Map Display pane. In the Layer Manager window click File | Import Vector Data |Common Import Formats.

2. A GUI window will appear to lead you forward. Choose Roadways.shp from the drop down menu and push Run. If you watch the Layer Manager window it will tell you when all is complete. At the bottom of the window you will have to hit the Map Layers tab as you have been watching the command console tab.

3. Now the Roadways.shp file is loaded to the Layer Manager window simply toggle it on to view it in the Map Display window.

Next do the same for the wards.shp file.

Importing .csv

This is not as straight forward as the .shp files. You must first open the Addresses.csv file in OpenOffice Calc for editing.

When OpenOffice asks which format the file is in pick the one that has ASCII. Then toggle on: 'separted by commas' and untoggle anything else like periods or spaces ect. In the preview box at the bottom it should now display a more normal spaced looking file.

Calc

Edit all the column headings to simple, short, alphabetical names with no spaces. Delete unnessary columns. Leave the first one, pin number,addressNumber,road names and the x and y coordinate columns. You have six columns now.

The x,y columns need to be formated. You can use the 'Find and Replace' tool. First replace all commas with periods. Then right justify the x and y columns. These are necessary steps.

Save the file as a .csv. OpenOffice might say that there is a problem, just ignore it.

To bring it into GRASS:

Click File|Import vector data|ASCII points/GRASS ASCII vector import [v.in.ascii]

Under the Required Tab:

Browse to the file and choose it. Then back in the GUI, name the output file.

Under the Imput Format Tab:

Change the Field Sparator to a comma ','

Under the Points Tab:

Number of header lines to skip: 1

Number of column used for x: 5

Number of column used for y : 6

Number column used for category: 1

Hit Run

Processing in Grass

Type this in the command line box, it will adjust the region parameters.

g.region -p vect=Roadways

Next extract the location you specifically wish to focus on. In this case it will be the Somerset ward.

v.extract input=Wards output=Somerset type=area where=cat=16

Next extract the customers that are within that ward from the larger customer data set.

v.select ainput=Addresses atype=point binput=Somerset type=area output=addSomerset 
operator=within

Next extract the Roadways

v.overlay ainput=Roadways atype=line binput=Somerset output=SomerRoads operator=and

Network Preperation

Type this in the command line box, it will adjust the region parameters.

g.region -p vect=Somerset

Next clean the data

v.clean input=SomerRoads output=SomerRoadCl type=line tool=break,rmdupl

Extract a Start point

When you right click on the addSomerset layer in the Layer Manager window, select to veiw the Attribute table. Scroll to find the exact address you want to use and note its 'cat#', which will be the one in the first coloum for the row you choose, you will need to input that number later in the GUI. We will use the GUI for the next 2 extrations.

Click on menue bar: Vector|Feature selection|Query with attributes

Browse to the addSomerset file and choose it

Name the output map as: Start

In the next tab: Selection

types to be extracted: point

Where conditions of SQL type: cat=40698 OR the start point you choose earlier

click Run but do not close the window!!!

Next extract all other points minus the one we just extracted.

All you need to do is simply toggle back to the first tab to change the output name to Customers.

And then in the Selection tab: toggle on the Revese Selection option and it will do the rest.

Hit run and close when done.

Attribute Editing

We know that the homes are not located directly on top of the roads so we must calculate and attach them to the network for processing and analysis. First we will create 3 new columns in the attribute tables and then populate them.

v.db.addcol map=Customers columns='x double precision,y double precision,dist int'

v.db.addcol map=Start columns='x double precision,y double precision,dist int'

v.to.db map=Start type=point option=coor columns=x,y

v.to.db map=Customers type=point option=coor columns=x,y

v.distance from=Start to=SomerRoadCl  to_type=line upload=dist column=dist

v.distance from=Customers to=SomerRoadCl  to_type=line upload=dist column=dist

Connect Customers to the Network

Once we have the distance from all points to the roads,connect all the addresses to the road-network according to a threshold (distance)

v.net input=SomerRoadCl points=Start output=network_v2 operation=connect thresh=10

v.net input=network_v2 points=Customers output=net operation=connect thresh=50

Click to see the connections. Customers are pink, Roads are grey, Connections are black

Report

we can view information about the network

v.category net op=report

Network Analysis Examples: Shortest path, no cost

We have a start point, now it is time to choose a destination. Once again choose the 'cat#' of an address point, as you did earlier, using the attribute table from Customers.

v.extract input=Customers output=End type=point where=cat=44017

Open the Attribute tables for both the Start point layer and the End point layer.

Open Notepad to make a .txt file of these layer's x,y coordinates.

Format is:

You will get the Start and End layer's x and y coordinates from their respective attribute tables. This is a finicy process, be sure not to put any extra spaces. Also do not hit the retun key after you type the last number. If you do simply you the backspace key and delete everything up til the last number entered. Save the file as od.txt

Shortest Path No-Cost

v.net.path input=net output=route_nocost type=line file=od.txt

If you get an error then use the GUI found under Vector|Network Analysis|Shortest path. Or retype the code and in front of od.txt put its full address, like: file=/home/desktop/od.txt

Network Analysis Examples: Shortest path, with costs

First add two new coloums to the network file.

 v.db.addcol map=net columns='saftey int,cost int'

Next create at .txt file in Notepad

UPDATE Routes SET saftey=1 WHERE RD_SUFFIX='ST';
UPDATE Routes SET saftey=1 WHERE RD_SUFFIX='AVE';
UPDATE Routes SET saftey=1 WHERE RD_SUFFIX='CRT';
UPDATE Routes SET saftey=1 WHERE RD_SUFFIX='DRWY';
UPDATE Routes SET saftey=1 WHERE RD_SUFFIX='LANE';
UPDATE Routes SET saftey=100 WHERE RD_SUFFIX='PKWY';
UPDATE Routes SET saftey=1 WHERE RD_SUFFIX='PL';
UPDATE Routes SET saftey=1 WHERE RD_SUFFIX='PRIV';
UPDATE Routes SET saftey=100 WHERE RD_NAME LIKE 'HWY';
UPDATE Routes SET saftey=100 WHERE RD_NAME LIKE 'HIGHWAY';
UPDATE Routes SET saftey=100 WHERE RD_NAME = 'TURN LANE';
UPDATE Routes SET saftey=100 WHERE RD_NAME LIKE 'DRIVEWAY';

Now you can

1 Save the .txt as saftey.txt and then rename the extention saftey.sql and then use the command line to execute it, if you know the address.

db.execute input=/home/desktop/saftey.sql

2. OR you can open the GUI under the menu Database|Query|SQL statement

Simply copy and paste the UPDATE arguments from above in the space provided, and hit run

This is what I did but I had to do all steps in GRASS6.5 as it didnt work in 6.4.2

Now that the saftey coloumn has been updated we can preform a cost related analysis. You will notice the non-residential like streets have a high cost and the residentials have a low cost. The calculation will search for the shortest path with the highest saftey rating.

Now you can calulate with costs by either:

v.net.path input=net output=route_to_libraries_withcost file=/home/desktop/od.txt afcolumn=saftey dmax=10000

if you know the address of the od.txt file :

Safest Route

OR by using the GUI at Vector|Network Analysis|Shortest path

Network Analysis Examples: Traveling Salesman Problem

If you have a set of points that you must travel to, extract them into a file of their own. Clean a new roads file for use Connect the two layers as we did earlier

Run a report on the network to discover the min and max of the second layer

v.category network name op=report

Run the Traveling Salesman Analysis

put the min and the max from the report for layer 2. ccats=8-21 is just an example of how you should type it

v.net.salesman input=network output=salesman_nocost ccats=8-21

Data

• Go to The city of Ottawa, Open Data Site
• Download the .csv file Address Points (main) Or click this link. Note: When you unzip this file change its name to customers.csv.
• While at the same site download the Roadways.shp file and the wards2010.shp file too.

Conclusion

About This Tutorial

This tutorial was created for GEOM4008 which is part of the Geomatics program at Carleton University, located in Ottawa, Ontario, Canada.

frame

External Links

• Travelling salesman problem explained

Here is a list of links that offer additional support and information;

• Quantum GIS Web site
• Quantum GIS Wiki

• Grass GIS Web site
• Grass GIS Wiki
• Spatial Reference

• GDAL
• OSGeo
• Carleton University Web site
• Carleton University Wiki

• HTML Codes

References

Data

Tutorial http://meta.wikimedia.org/wiki/Meta:Sandbox