Asset Tracking with RollCallTM: Design, Deployment and Analysis

1
Asset Tracking with Roll-CallTM Design,
Deployment and Analysis
Yanyong Zhang, Ben Firner, Rich Howard, Wade
Trappe WINLAB, Rutgers University
2
Asset Tracking Applications

• We envision that an efficient asset tracking
  application should
• Continuously track the position of each item
• Detect the missing of an item in real time
• E.g., jewelry tracking

3
Limitations with Existing RFID
Passive RFID Tags Inexpensive Long Lifetime -
Poor Range
Active RFID Tags Good Range - More Expensive -
Short Lifetime - Larger
Im still here!
Obstacle
Obstacle
Anyone there?
Passive Tag Reader
Active Tag Receiver
4
Our Solution Roll-CallTM

• Tag
• Tag periodically announces its ID.
• Receiver disabled.
• Life time 1 year.
• Basestation
• Receives and timestamps the packets, then
  transfer them to the processing station

• Multiple receives used to provide overlapping
  coverage.
• Central Processing Unit
• Tracks the tags using data from multiple
  receivers.

5
Challenges Packet Losses

• Packet losses can lead to false alarms.
• There are three types of losses
• Ambient losses due to radio environment

Solution better placement of basestations
6
Challenges Packet Losses

• Packet losses can lead to false alarms.
• There are three types of losses
• Ambient losses due to radio environment
• Collision losses
• Random transmissions lead to collisions
• Our model shows that in a 100-tag system with
  each tag transmitting a 100 bit packet in 100ms
  epochs, the collision probability is 5

7
Challenges Packet Losses

• Packet losses can lead to false alarms.
• There are three types of losses
• Ambient losses due to radio environment
• Collision losses
• Dead-time losses due to unavailability of the
  basestation
• Processing time
• Transfer time

Solution Reduce the processing at the
basestation, and buffering the packets for a
batch transfer
8
Throughput Limit Test

• Experiment setup a tag transmits 20,000 packets
  at different packet intervals
• Combining buffering and use of multiple
  basestations achieves the best performance

9
Raw Packet Loss Ratio
Packet loss ratio
Number of basestations
10
Presence Detection

• Raw packet loss ratio information is not fed into
  applications instead, we need to detect whether
  a tag is missing
• Declare miss ratio (DMR)
• how resilient a detection scheme is against
  packet losses
• n packet losses, out of which we detect m tag
  misses, then the DMR is m/n
• Report Latency (RL)
• The interval between the last when a tag was
  heard and the time it was declared missing given
  that it is actually lost
• Detection Heuristics

11
Conclusions and Future Directions

• The initial Roll-CallTM design is promising
• Packet losses should be further reduced
• More enhancements are on the way
• Extracting information from collided packets
• Employing basestations with multiple processors
• More sophisticated detection algorithms are
  needed
• Will extensively study the performance in
  different environments

12
Questions Answers