Sensor Data Collection through a DelayTolerant MANET of Small Unmanned Aircraft

1
Sensor Data Collection through a Delay-Tolerant
MANET of Small Unmanned Aircraft
Daniel Henkel, Timothy X Brown Interdisciplinary
Telecommunications Electrical and Computer
Engineering University of Colorado Presented at
the Second International Workshop on Multi-hop
Ad Hoc Networks from Theory to Reality (REALMAN)
Florence May 26, 2006
2
Overview

• Widely dispersed sensors require a mechanism to
  collect the sensor event data.
• Mobile nodes can move among the sensors to
  collect the data.
• We describe a sensor data collection
  implementation that uses unmanned aerial vehicle
  (UAV) nodes combined with a delay tolerant
  networking (DTN).
• We also describe some of the issues in service
  discovery and mobility control.

3
Sensor Data Collection Problem

• We have built a large-scale ad hoc network test
  bed
• Traditional ad hoc networking is not optimized
  for sparse intermittently connected sensor
  networks
• Data must be multicast out to multiple Sensor
  Monitoring Stations (SMS). SMS must also be able
  to send unicast commands

More Details
Skip Details
4
AUGNet
AUGNet
Ad hoc UAV Ground Network
Group 1
Group 2
UAVs provide contemporaneous end-to-end
connectivity
5
Collection Tasks
SMS-3
Sensor-1
Sensor-2
Sensor-3

• Data Delivery
• Sensor event delivery (multicast)
• SMS to sensor control (uni-cast)
• Controlled Mobility

6
Data Collection Problems
SMS-3

• Multicasting of event messages to all SMS
• Intermittent connectivity on any link
• Service discovery of GW, SMS

7
Data Delivery

• We use a combination of
• Event multicasting
• Reliable, staged forwarding
• Gateway and SMS service discovery

More Details
Skip Details
8
Event Multicasting
SMS
Gateway
ISR
Terminus
SMS
Gateway
SMS

• Staged delivery (custody transfer)
• Multiplication of messages in each stage

• NAPT

9
Reliable Packet Forwarding
TX stage
RX stage
put insend buffer
Event packet
TCP
seconds to minutes
ACK packet
delete fromsend buffer

• Modified UDP, no TCP, no RUDP
• Sequence numbers, RX seq hash table
• ACK Timeout-based reliability

10
Service Discovery

• Multiple GW, SMS in out of coverage
• Heartbeat and GW advertisements

11
Implementation

• Software Click Modular Router
• Hardware Soekris Single Board Computers
• Performance is good when connected and can
  deliver packets despite multiple extended
  connection failures.
• Have integrated gateway with cellular backhaul

More Details
Skip Details
12
Click Modular Router
13
Implementation/Performance

• RTT 40ms, 15hrs sustained operation

• Soekris SBC, embedded Gentoo Linux
• Atheros miniPCI, Madwifi-ng driver

14
Backhaul
close-up of soekris board, cdma phone, base
station

• CDMA phone with USB data cable

15
Controlled Mobility

• Intermittent networking shown given
  eventual connectivity
• How do we guarantee eventual connectivity?
• Solution Controlled Mobility
• Have investigated several strategies.

More Details
Skip Details
16
Ferrying

• Store, Carry, Forward network
• Data physically carried to destination by special
  ferry nodes
• Prior work Zhang/Ammar Path planning for
  ferries TSP theory

controlled movement
Send Buffer
Ferry Buffer
Receiver
17
Ferrying Models

• Conveyor-belt Model
• Chain-Relay Model

18
Conveyor vs. Chain-Relay

• Metrics throughput, packet delay
• Which model is better?
• Fast ferries
• Conveyor Belt Model
• High data rate, long-range radios
• Chain-Relay Model
• (Henkel D., Brown T., On Controlled Node Mobility
  in Delay Tolerant Networks of Unmanned Aerial
  Vehicles, ISART 06)

19
Conclusion

• We have built a DTN for sensor data collection.
• We have theoretical results on ferry scheduling.
• Future
• Real DTN persistent storage in db
• Plane trajectory design depending on observed
  traffic
• Real-time sensor traffic notification
• henk,timxb_at_colorado.edu augnet.colorado
  .edu recuv.colorado.edu