1. By: Ziji Song Advisor: Prof. J. Spinelli
EER Senior Project Wireless Sensor Placement in Arbitrary 3D Environment
By: Ziji Song
Advisor: Prof. J. Spinelli

2. Origin and Application of the Problem
Union Outing Club students were trapped in a cave during caving due to water flooding
Application:
Wireless network setup

3. Goal of the Project
For any given 3D environment, setup a wireless sensor network in the environment based on its shape and the communication range of the sensor.
Use different algorithms to place motes and compare the different results

4. Design Specifications
All empty spaces in the environment should be reached by at least one sensor mote
The least number of sensor motes should be used

5. Description of the 3D Environment
The 3D Environment  “Cave”
Cave is divided into nodes which are separated by x,y and z paralleled planes (i.e a 3*3*3 cave will look like a Rubic’s cube)
Empty spaces nodes “Free Space Nodes” (refer to the center point of the cube area)
Non-empty space nodes  “Rock”

6. Neighbor Density
Definition: For a given free space node in the cave, denote Q as the set of free space nodes which direct lines between p and Q maps only free space nodes. The neighbor density of p equals ∑1/dist2(p,q) for all q in Q.

7. 2D Example of Calculating ND

8. Another 2D Example with Rock

9. Important Definitions
Covered Node: If a free space node is reachable by at least one sensor mote, then it is a covered node.
Occupied Node: If a sensor mote is placed on a free space node, then the node is occupied.

10. Two Algorithms to Place Motes
The First Algorithm
Adding motes to an empty cave
Add motes to the uncovered node with highest ND
Repeat step 2 until all FS nodes are covered

11. Two Algorithms to Place Motes
The Second Algorithm
Subtracting motes from a fully occupied cave
Place all FS nodes in an array and sort the array ASC with respect to ND
For each node in the array, remove the sensor
If not all FS nodes are covered, put the sensor back on that particular FS node

12. Comparison of the Two Methods
Run time
Simulation results

23. Conclusion No known algorithm to find the optimal solution
Run Time might go up exponentially
Both algorithms give reasonable solutions

24. Special Thanks Prof. Chang Prof. Fernandes
Prof. Almstead Prof. Hedrick