The Graphical User Interface for
GSST, Extra Nodes and Extra Searcher Algorithms
Ath. Kehagias,
G. Hollinger,A. Gelastopoulos
and K.Tobakidis
GSST, Extra Nodes and Extra Searcher
are algorithms for graph search; more specifically GSST clears a graph with a node-located
intruder by an internal, monotone, connected (IMC)
search. The algorithm is described in the paper "Searching the
Nodes of Graph: Theory and Algorithms", Ath.
Kehagias, G. Hollinger and A. Gelastopoulos,
2009 . Extra Nodes and Extra Searcher clear a graph with an edge-located
intruder by an internal, monotone, connected search. These edge-clearing
algorithms need to make use of GSST (or any other node-clearing IMC
algorithm) and are described in the paper "Pursuit-Evasion
Problems using Graph Theory methods: Software development and experiments ",
Ath. Kehagias, K. Tobakidis, 2010 (in Greek only).
§
The paper "Searching the
Nodes of Graph: Theory and Algorithms" can be downloaded at http://users.auth.gr/~kehagiat/KehagiasPublications.htm
(5-14), Section: Selected Technical Reports.
§ The gsearchGUI.zip package can be downloaded at
http://users.auth.gr/~kehagiat/KehagiasSoftware.htm. The package contains the implementations of the above algorithms as described
in the Components Section along with their graphical
front end which runs on Windows computers.
§
The original implementation of GSST with it’s source code for Windows and Linux operating systems and graph
files for testing can be found at http://robotics.usc.edu/~geoff/software.html Geoffrey Hollinger’s software
page.
Installation of
gsearchGUI is extremely simple:
download gsstGUI.zip and unzip it into a
directory of your choice; then run gsearchGUI.exe. GsearchGUI is distributed for use as is,
without any explicit or implicit warranty. Feel free to use it and disseminate
it.
GsearchGUI v1.0 was implemented in MS
Visual Basic 6.0 by Ath. Kehagias
and K. Tobakidis.
The program corresponds closely to the command line components, as
described in the Components Section. Launching the program brings up the
window of Figure 1. On the left side we see several input boxes. All of these
correspond to options of the gsearch.exe program (for an explanation
of the options see the
next section of this
document or the above mentioned paper) with one exception: in addition to the
edge list file (graphs/Edge01.txt in Figure 1) there is also a file
containing the x- and y-coordinates of the nodes (graphs/Node01.txt in Figure 1 -- this is required for
visualization). At bottom, there are the options of the functions of the Eclear.exe
program (as described in the Components Section and the second paper),
using which we can run an edge-clearing search.
Figure
1. A screenshot of the gsearch
GUI.
On the right side there are
several command buttons:
1. The “Plot Graph” button creates an image of
the given (always named as graph.jpg) and invokes the AutoPics image viewer to
project it.
2. The “Search” button performs
the graph search (it actually invokes the gsearch and if needed the Eclear executables and stores their results
in the output directory). Time to complete the search depends
on the size of the graph and the number of spanning trees used.
3. The “Plot Node Search” and “Plot
Edge Search” buttons plot the graph
search for the node or the edge game. Actually, first the search strategy files
are used to generate a sequence of image files, one image corresponding to each
step of the search (these files are stored in the directory pix) and then the image viewer is
invoked to view the files. The user can step through the images using the arrow
keys, or use the start / stop
buttons
to run a slide show of the search.
4. The “Help”
button invokes the current file.
5.
The “Exit” button terminates
the program.
When starting the program, as we see in Figure 1, some buttons are by default disabled. Those buttons
cannot be used as there is no search strategy saved yet. Also the “Extra Searcher” and “Extra Nodes”
radio button will only be enabled when the “Edge
Clearing” checkbox is checked. Let’s see an example of the
program’s usage, following the next steps:
Figure 2. Choosing the
graph files.
1. Choose the edge and node files.The GUI starts with the default selection of Edge01.txt and Node01.txt ; if you want a different graph,
type the corresponding file names in the input boxes or click the browsing
buttons (Labelled with “…”). Some examples of these files are located in the graphs directory.
Figure 3. Browsing the
example files in the graphs directory.
2. Create and view the image of the
graph by pressing the “Plot Graph”
button.
Figure 4. Autopics
projecting the graph’s image file.
3. Choose the arguments of gsearch.
When there is a specific range, we can choose from a combobox
list.
Figure 5. Choosing how to
break edge ties.
4. If we wish to perform an edge-clearing
search we check the checkbox bellow and choose the desired algorithm. This step
can be omitted if we wish to perform a node search (initially).
Figure 6. Edge clearing options.
5. Perform the search by pressing the “Search” button. While the necessary
commands are running, a small tool called Gsearch
killer will pop up. In case the search is
taking too long, we can interrupt it by pressing the “KILL
IT” button. Afterwards we can change our options to minimize the
search time and rerun the search.
Figure 7. Gsearch killer tool.
When the search is over, a message box will
inform us for the minimum number of searchers we could achieve.
Figure 8. Message box after the search runs.
6. We can view the saved search strategy for
the node (edge) game by pressing the “Plot Node
(Edge) Search” buttons, which are now enabled. Every node of the
graph are colored with light green if one searcher is present, with dark green if
more than one searchers are present and with white if
a searcher has visited it in the past steps of the search. The edges of the
graph are colored with white if they have been node or edge cleared (depending
on the game we chose to visualize). Every element of the graph, node or edge,
than hasn’t yet ever been clear is colored with light brown. The root of
the search is highlighted by a brown circle around it. We can follow the search either manually or by using the “Start” and “Stop”
buttons.
Figure 9. Edge game search visualization.
The edge {4-6} would be considered as clean for the node game.
7. If we omitted the fourth step, meaning we
performed a node search, we can now apply the “Teleport
Clear” or “Extra Searcher”
functions on the saved node search. The program informs us when the operation
is over.
Figure 10. End of teleport clearing or
applying the Extra Searcher algorithm on a saved search.
Press “OK”
and plot the search once again. Observe the possible differences from
the initial search.
8. We can change our options and run as many
searches we wish; then exit the program.
The executable components that gsearchGUI uses
are four:
1. The command
line program gsearch.exe. The program was implemented by G. Hollinger in
ANSI C using the gcc 4.1 compiler. For the sake of
completeness we also list the help file of gsearch.exe.
USAGE: gsearch -m [graph] -n [no.trees] -s [startnode] -g [gen.tree] -y [wr-search] -w [wr-tree] -v [visualize]
-t [edge traversal], -i [improve tree] -l [low number], -r [redundancy check]
EXAMPLE: gsearch -m graphs/Edge14.txt -g exhaustive -s 1 -n 500000 -i 0 -t bh -y one -w one
The various options
available are as follows.
-m [graph]: string, name of file with edge list of the
graph.
-n [no.trees]: int, how many sp.trees to
generate (DEFAULT is 1).
-s
[start node]: int, which node to start (DEFAULT is 1,
random choice is 0)
-g [gen.tree]:
string, method of generating spanning trees
(acceptable
values: readtree,
exhaustive, uniform, dfsrand; DEFAULT is uniform)
-y [wr-search]: string, how
many best searches to write
(acceptable
values: one, all, none; DEFAULT is one)
-w [wr-tree]: string,
how many best sp.trees to write
(acceptable
values: one, all, none; DEFAULT is one)
-t
[traversal]: string, how to break edge label ties
(acceptable
values: bh,
random, bhrand, bhweight, bhdom; DEFAULT is bh).
-i [improve tree]: boolean, use
tree improvement technique or not.
(acceptable values
are 0 / 1; DEFAULT is 0: do not use)
-r
[redundancy check]: boolean, check for
redundant trees
(acceptable values
are 0 / 1; DEFAULT 1: check for redundancy)
-l [low
number]: break if a tree is found with this number of searchers
(DEFAULT
is 0: do not break)
-v
[visualize]: boolean, use visualizer (only
supported on linux)
(DEFAULT
is 0: do not use)
-h
[help]
The -m option indicates the
file which contains the graph description. This must be an ascii (plain text) file containing a list of the
edges of the graph, one edge per line, indicated as a pair of nodes. The nodes
must be continuously numbered from 1 to N and these numbers are
used as labels. N is assumed to be the largest number
appearing in the edge list (the graph is assumed to be undirected and
connected). Examples of edge lists can be found in the graphs directory. (Many
examples of edge files are stored in the subdirectory graphs).
The remaining options of gsearch correspond to the
description of the algorithm as given in the paper cited above. The -n option corresponds to
the M parameter of GSST
(number of spanning trees), -s is the root of the search, -g describes the uniform
and DFS methods for generating spanning trees (there
are also options for exhaustive enumeration of all spanning trees and for
reading a specific spanning tree from file) and the -t option corresponds to
the traversal methods BH, UR, BR, BW and BD. The -i, -r and -l options are self
explanatory. The -y option writes one or more minimal
strategies in the file(s) output/strat*.txt; each row of the file
corresponds to one step of the strategy and shows the nodes in which the
searchers are currently located. The -w option writes the
rooted tree(s) corresponding to optimal strategies, in the file(s) output/tree*.txt; each row of the file
shows one parent and her child. The -v visualization option
only works on Linux computers (for a visualization of the search on Windows
computers use the GUI).
2. The command line program Eclear.exe.The implication
of the Extra Searcher and Extra Nodes algorithms was created by K.
Tobakidis in C using the Dev C++ v.4.9.9.2 compiler.
The help of the program is listed bellow.
Usage:
Eclear -i Edgefile :(Extra
Nodes)
prints the fake-noded Edgefile
Eclear -o Edgefile Searchfile(with fake nodes) :(Extra Nodes)
prints the edge
clearing strategy in the original graph
Eclear -x Edgefile Searchfile :(Extra Searcher)
applies the
Extra Searcher on a node clearing strategy
Eclear -t Edgefile Searchfile :(Node Search)
prints the node
clearing strategy, teleport-cleared
The program can also perform a clearing of the
teleporting moves on any IMC search strategy, with the –t option.
3. The Gsearch
killer program. A program to interrupt the above two
command line programs when we wish to. The program is shown at Figure 7
(Quick start Section) It is used as VB6 does not
support multithreading so the Operating System’s multithreading does the job.
4. The Autopics picture viewer. The image viewer is a freeware stand
alone program made available by the company Mydesktophelp (at http://www.mydesktophelp.com/development.php).
Also the program uses the freeware
graphics library called GD available at http://www.libgd.org/Main_Page.
1. Do not use the exhaustive
spanning tree enumeration option on large graphs because the program may run
out of memory and crash.
2. When first running
the program, if an error message regarding the msstdfmt.dll file is shown, run the msstdfmXP.bat file, located in the main
folder. This problem is mainly occurred on Windows XP OS. The msstdfmXP.bat command must be run with Administrative rights.
3. On Windows
4. If the program still doe not run
properly on Windows Vista or 7, try using the compatibility mode.