A Campus KioskA Campus KioskKendal Hershberger

& David Moses

Union University Scholarship Symposium

April 30, 2007

PurposePurpose

► Make an application to give directions to people new to campus who may not know their way around

► Make the application easily accessible

► Scalability

Choosing the LanguageChoosing the Language

► OpenGL using C Advantage - already familiar with the

language and graphics library Disadvantage - difficulty of drawing an

entire campus

►DirectX Disadvantage – Microsoft-specific systems

Choosing the LanguageChoosing the Language

► Flash and ActionScript Advantages

► Very accessible and easy to update► Opportunity to learn a new language► ActionScript has many similarites to Java and

other high-level languages► Easy drawing utilities

Disadvantages► Unfamiliar

ActionScriptActionScript

► scripting language behind Flash applications

►ActionScript 2.0 new object-oriented model

►Class-based, but classes defined in external .as file

Graph SearchingGraph Searching

►Graphs consist of nodes and vectors (or edges).

►A node can be connected to another node via an edge.

►For our case, the edges are paths that traverse Union's campus, and each node represents a geographical point where they connect or terminate.

Graph SearchingGraph Searching

►Example:

http://www.csie.ntu.edu.tw/~wcchen/algorithm/graph/Graph4.htmhttp://www.csie.ntu.edu.tw/~wcchen/algorithm/graph/Graph4.htm

Graph SearchingGraph Searching

►Depth Depth First Iterative

Deepening Best First

►Breadth Breadth First Uniform Cost Dijkstra's

Algorithm

►Problem: Find the best path from one node (one point on campus) to another node (another point on campus)

►Techniques:

Graph SearchingGraph Searching

►Depth First Go to a node, mark it as visited, push it on

a stack Go to a neighbour node, and do the same

thing If you’re at a dead end, pop the stack until

you find an unvisited neighbour Repeat until you’ve traversed the entire

tree (optimal path) or until you have found the destination node (non-optimal)

Graph SearchingGraph Searching

►Depth First

http://www.csie.ntu.edu.tw/~wcchen/algorithm/graph/Graph4.htmhttp://www.csie.ntu.edu.tw/~wcchen/algorithm/graph/Graph4.htm

Graph SearchingGraph Searching

►Iterative Deepening Same as Depth First, but depth of search

is controlled Better for very large graphs, still slow.

►Best First Greedy algorithm (not uninformed) that

simply selects the closest node to the destination

VERY fast, but can generate horrible paths under certain conditions

Graph SearchingGraph Searching

►Breadth First Go to a node, put all of its neighbour in a

queue Visit each neighbour in the order they

were placed, and put all of its neighbours in a queue

Continue until all nodes have been visited (optimal path) or until destination node is found (non-optimal path)

Graph SearchingGraph Searching

►Breadth First

http://www.csie.ntu.edu.tw/~wcchen/algorithm/graph/Graph4.htmhttp://www.csie.ntu.edu.tw/~wcchen/algorithm/graph/Graph4.htm

Graph SearchingGraph Searching

►Uniform Cost Same as Breadth First Search, but

prioritizes nodes that have shortest path from origin rather than simply the next node in the queue

More realistic paths, but still takes long

►Dijkstra's Algorithm Same as Uniform Cost, but actively keeps

track of shortest path to node Uses relaxation of the path cost Much faster than all others

Graph SearchingGraph Searching► dijkstra = function (origin:Node, destination:Node) {dijkstra = function (origin:Node, destination:Node) {

► var S:Array = new Array();var S:Array = new Array(); // list of visited nodes// list of visited nodes► var Q:Array = Node.getNodes(); var Q:Array = Node.getNodes(); // list of unvisited nodes// list of unvisited nodes► origin.dv = 0;origin.dv = 0; // initialize distance at origin to // initialize distance at origin to ►

► do {do {► var u:Node = extractMin(Q);var u:Node = extractMin(Q); // remove node with lowest // remove node with lowest

distancedistance► S.push(u);S.push(u); // mark it as visited// mark it as visited

► var neighbors:Array = u.getNeighbors();var neighbors:Array = u.getNeighbors(); // relax u's neighbors// relax u's neighbors► for (i=0; i<neighbors.length; i++) {for (i=0; i<neighbors.length; i++) {► var v:Node = neighbors[i];var v:Node = neighbors[i];► if (u.dv + u.getDistance(v) < v.dv) {if (u.dv + u.getDistance(v) < v.dv) {► v.dv = u.dv + u.getDistance(v);v.dv = u.dv + u.getDistance(v);► v.prevNode = u;v.prevNode = u;► }}► }}► } while (u != destination)} while (u != destination)► };};

Graph SearchingGraph Searching

►Dijkstra’s Algorithm

http://cgm.cs.mcgill.ca/~msuder/courses/203/lectures/greedy_algorithms/http://cgm.cs.mcgill.ca/~msuder/courses/203/lectures/greedy_algorithms/

Creating the Campus MapCreating the Campus Map

►Deciding on using an image or drawing the map ourselves

► Using Flash to draw graphics vector vs. raster graphics

► Quality► Scalability

Creating the Campus MapCreating the Campus Map

Creating PathsCreating Paths

►Create a datatype to store the nodes and edges in ActionScript

►Make a list of points to place the nodes►Display it in a Flash scene

Creating PathsCreating Paths

►Kirupa’s Graph ADT►Node and Edge

classes var a:Node; a = new Node("a", 200, 50); a.addEdge(b, "a->b");

►Uses Depth First to find path

►Dynamically draws the path from origin to destination

Finding the Best PathFinding the Best Path

►Our implementation of the Graph ADT

►Tested using points around the Belltower area

►Depth First search is fairly accurate and fast

Finding the Best PathFinding the Best Path

►Completed entry of points for the rest of campus

►Our naïve Depth First search algorithm reveals its limitations

Finding the Best PathFinding the Best Path

►Switched from Kirupa’s Depth First Search and implemented our own Dijkstra’s algorithm in ActionScript

►Still needed to provide interface and integrate map with graph

Building an InterfaceBuilding an Interface

►Added map layer to Flash scene

►Started adding buttons and designing interface components

Building an InterfaceBuilding an Interface

► Choosing locations in a List Box

DebuggingDebugging

►Remaining Bugs Refresh problem

►Fixed by deleting the calls to draw.

Speed degradation►Caused by

unintended doubling of the graph size.

Faulty edges► Incorrect data entry…

over 500 individual edges!

Future PossibilitiesFuture Possibilities

►Interiors of buildings, including classrooms and faculty offices Zoom feature Database integration

►Touch screen kiosks located around campus

►Print feature

Demonstation / QuestionsDemonstation / Questions