Using Path-Finding

Algorithms of Graph
Theory for Route-
Searching in Geographical
Information Systems
By
Radhika Kumaran
09MW13
I ME Software Engg
 Abstract
► This paper deals with graph theory application in large-
scale geographical data searching and visualization.

► A comparison of two path-finding algorithms of graph

theory, i.e. blind-search and A* algorithm, that results
into the selection of the second one for
implementation within a geographical information
system.

► The main goal is to process the terrain map

representation in the form of a graph during route-
searching between start and target positions, and
during the visualization and animation of the results of
this search.
 Introduction to Graphs
► Graphs can be imagined as a set of points (vertices)
interconnected by lines (edges).
► The edges can have specified direction
(orientation).
► Paths are sequences of vertices and edges between
them.
► Path-finding between two graph points is used in
many areas, like GPS navigation, CAD systems
(auto-routing in design of circuit boards), and
applications of artificial intelligence, computer
games, virtual reality, military forces or robotics.
► The main aim is to pass through environment in
secure form and to avoid of obstacles.
 Introduction to GIS
► A geographical information system (GIS) is an
information system (IS) designed to work with data
that represent geographical or spatial coordinates.
► These data can be further analyzed and statistically
evaluated.
► Another way to specify a GIS is that it is a specialized
IS defined as a collection of computer hardware and
software and geographical data designed for
effective gathering, retaining, editing, processing,
analysis and visualization of all kinds of geographical
information.
► Generally, it is a spatial specialized IS.
► Being an IS, GIS has to provide information
 Definitions
Let G = (V, E) be a coherent graph and let Distance d(u, v)
between vertices u and v of graph G be the shortest path.
Vertices distance d(u, v) of graph G has the following
properties:
• d(u,v) 0;u = v,
• d(u,v) = d(v,u),
• z V : d(u,v) d(u,z) + d(z,v).
And
1. For every vertex u belongs to V the number e(u,G) = max
d(u,v), is called eccentricity of vertex u.
2. Number P(G) = max e(v,G), v belongs to V is average of
graph G.
3. Number r(G) = min e(v,G), v belongs to V is radius of
graph G.
 Concepts in Maps, Routes and
Algorithms
► The basic working structure of algorithms is the map, which
is represented as a set M = N x N (two dimensional
quadratic grid).
► Square is a member of set M.
► Every square represents position in map and has its own
coordinates [x,y] according to beginning Pm.
► Squares [x1,y1] and [x2,y2] are neighbors, if (|x1–x2|=1
and |y1–
► y2|<=1) or (|x1–x2|<=1 and |y1–y2|=1).
► The beginning in map Pm is a defined square.
► Function to assign regress or wieght to squares wm:
M ->R+ U infinity
► The route length is the sum of all regresses of the squares.
 Algorithms for Route
Searching
► There are several algorithms for route-searching in
static maps:
• blind-search
• divide & conquer
• breadth–first search
• bidirectional breadth–first search
• depth-first search
• iterative depth-first search
• Dijkstra’s algorithm
• best-first search
• Dijkstra’s algorithm with heuristics (A*)
 Blind Search Algorithm
► Search based on “Progress till Collision”.
► The evaluation process is as follows:
abs[(ex tx ) + (ey ty )], (8)
where ex,y are the [x,y] coordinates of the actuallyevaluated
square and tx,y are coordinates of target squareon the map.
► This algorithm works with aim on one square and usually does not
take to consider the whole evaluating function w, but only two
states + or 1 and doesnot cosider regress.
► The evaluation process is as follows:
abs[(ex -tx ) + (ey - ty )], where e(x,y) are co ordinates of the
evaluated square and t(x,y) co ordinates of the target square.
 Advantages of Blind Search
► speed,
► low memory requirements and
► simple implementation.
 A* Algorithm
► The A* algorithm is a natural generalization of Dijkstra’s
algorithm and the scanning is based only on heuristics.
► In A* the element classification process is done by a
binary heap.
► All square near the evaluated square are also evaluated
in eight directions.
► Function g(x) is evaluated from the starting square. If
the evaluated square is in the diagonal, then the
number 14 is assigned to this square, and if not, then
number 10 is assigned to this square,i.e g(x)=14 or 12.
► As heuristics, the Manhattan method is used
h(x)=10* [abs(ex - tx ) + abs(ey - ty )]
f(x)= g(x) + h(x)
 Contd…
► Two values are stored for every already processed vertex:
1.d[v] will store the length of the shortest path to the vertex,
2. p[v] will be the vertex before v on the path.
► The algorithm also uses two sets: OPEN and CLOSED.
1.OPEN is the set of vertices to process,
2.CLOSED is the set of vertices already processed.
 A* Algorithm Working
► At the beginning, the starting vertex is inserted into OPEN.
► The iteration cycle starts with the choose of the vertex n with
best value of f(n).
► This vertex will be inserted into set CLOSED and all from n
reachable vertices v are selected, and if they are not in set
OPEN, they will be inserted with p[v]=n and the corresponding
value of d[v]=d[n]+l(n,v), where l(n,v)=wg(v) is the cost function
for the selected edge from n.
► If v was already a member of OPEN, we need to test if the new
path is shorter than the old by comparing d[v]<d[n]+l(n,v).
► If the comparison fails, we need to update the values p[v] and
d[v].
► The presented cycle continues until the target is found or the set
OPEN is empty.
► After the algorithm had finished, we need to reconstruct the path
using values of p[v].
 Improving A* Algorithm’s
Efficiency
► The A* algorithm can be made more efficient by
using techniques like optimizing the number of
vertices in the set OPEN and CLOSED.
► This can decrease memory usage and fasten the A*
algorithm.
algorithm
► The optimum values for the OPEN and CLOSED set
are experimentally found to be 1000 and 3000
respectively.
 Case Study-1
► BlindSearch
Algorithm gives this
route

► A*algorithm gives
this
 Case Study-2
► BlindSearch Algorithm
gets into an infinite
loop and never finds
the path.

► A*Algorithm efficiently
finds the path.
 Case Study-3
► Route found by
Blind Search
Algorithm

► Route found by A*
Algorithm
 Comparison of Squares
Searched by both the
Algorithms
► Thenumber of
squares searched
by blind search is
more than that of A*
as evident from the
graph.
 3D – Case Study
► Upper left corner has also
two buttons: 1 –
openmap, 2 – options.
Options are like map
resolution etc.
► right corner are
navigation buttons (3 –
minimize, 4 –close).
► The number 5 indicates
the search button.On
selecting this last option,
a search dialog opens .
 Contd..
► 6 is the field for the search
query
► 7 is the results displaying
area
► 8 is for search Execution
► 9 is for animation button
► 10 is the Path display button
► 11 is object display button.
► 12 is close dialog button
 Contd..
► Using the A* search
algorithm technique
on entering a area to
be searched the path
is displayed.
► Lower resolution might
miss out narrow paths.
► Buttons 13 and 16 are
for classical
navigation.
 Results of the Case Study
► Fromthe case study on the algorithms some
conclusions can be made like:
1. The blind-search algorithm is fast but is not accurate
enough and can be used in maps with low number
ofbobstacles

2. Despite time results (the speed of this algorithm is not

the fastest), this algorithm is reliable on 100% and has
provided better results.

3.The optimization techniques of set size restrictions

were useful and brought higher speed into execution of
path-finding.
 Conclusion
► As the map is the core of any GIS, a possible
way of use of path-finding on graphs within
such a system is the mapping of these maps
onto searchable graphs

► Collision detection is the main aim of the

algorithms, but, as this article implies, the
information system area should follow these
trends as well by implementing its algorithms
into the visualization and searching modules.

► Future GIS would include 3D interfaces as well.

Thank You