Lessons Learned Building TeamTrak: An UrbanOutdoor Mobile Testbed

1
Lessons Learned Building TeamTrak An
Urban/Outdoor Mobile Testbed
Jeffrey Hemmes, Douglas Thain, Christian
Poellabauer, Chris Moretti, Phil Snowberger, and
Brendan McNutt University of Notre Dame
2
Overview

• System Components
• Experience with Sensor Measurements
• Experience with Field Tests
• Lessons Learned and Future Work

3
System Components

• Hardware
• 32 Lenovo X41 Thinkpad tablet computers
• 8 HP iPAQ hx2795b PDA
• Garmin GPS-18 or HP iPAQ Navigation
• PNI V2Xe digital 2-axis compass
• Watchport/V2 digital camera
• SparkFun SerAccel v5 digital accelerometer
• Network
• Wireless Ethernet (802.11b) in ad hoc mode
• Standard Windows IP configuration
• Routing Protocol
• Simple distance-vector routing à la RIP or DSDV
• Sensor data included in routing packets

4
System Components
5
System Components
6
Experience With Sensors

• GPS position subject to drift errors

7
Experience With Sensors

• Effect of compass roll on heading

8
Experience With Field Tests

• Conducted field tests with undergraduate class
• Purpose
• Gain practical experience with the system
• Uncover implementation and design issues
• Suggest areas for future research
• Objectives
• Find a designated node hidden somewhere on campus
• Communicate its location to others via routing
  protocol
• Construct longest unbroken chain of links possible

9
Experience With Field Tests
10
Lessons Learned and Future Work

• GPS needs assistance when deployed in urban
  environments at human scale
• Obtaining a fix in many locations practically
  impossible
• Partial obstructions of the horizon can delay
  obtaining a GPS fix by as much as 5 to 10 minutes
• Extreme cases of position drift can render GPS
  information useless for many applications
• Need new ways of estimating location based on the
  reported position and location quality of other
  nodes

11
Lessons Learned and Future Work

• Sensors can compensate for poor network
  connectivity
• Ad hoc network connections haphazard at best
• Displaying last known position of other nodes
  proved quite helpful for locating a particular
  node
• Even very transient connections with few packets
  exchanged allowed users to identify other nodes
  positions fairly quickly
• How long do we maintain stale data?

12
Lessons Learned and Future Work

• Human operators play an important role
• Users generally tend to prefer maintaining
  connectivity, and will adjust their position
  accordingly
• Such behavior can be exploited to augment network
  protocols in mobile applications
• Much work in this area remains to be done

13
For More Information

• General
• http//www.nd.edu/teamtrak
• Downloads
• http//www.nd.edu/teamtrak/start
• Contact
• Jeffrey Hemmes
• jhemmes_at_cse.nd.edu
• www.nd.edu/jhemmes