1. Delay-Tolerant Communication using Aerial Mobile Robotic Helper Nodes
recuv.colorado.edu
Delay-Tolerant Communication using Aerial Mobile Robotic Helper Nodes
Daniel Henkel
April 4, 2008

2. Overview DTN Test Bed Direct, Relay, Ferry Models Relay Optimization
Choosing Optimal Mode
Sensor Data Collection
We have developed a test bed, which is a challenged network – motivation for exploring DTN
Looking at theoretical limits of communication in models of D, R, and F
Modeling such that upper bounds can be seen
Find optimum operation parameters for RELAY networks along a communication path; T-R separation distance, environment (PL-exponent)
In Ferrying Mode we optimize the path a ferry takes through the network
Question: Which mode yields highest t-put and lowest delay?
Practical application is sensor data collection as hinted at in the paper.

3. University of Colorado Location

4. Unmanned Aerial Vehicles (UAVs)
Small (10kg) Low-Cost (<$10k) UAVs
60-100km/h, 1hr endurance
5HP gas engine
Built in-house

5. Ad hoc UAV-Ground Networks
Scenario 1: increase ground node connectivity.
NOC
Scenario 2: increase UAV mission range.

6. Applications Military Scientific Civil / Commercial
Intelligence, Surveillance, Reconnaissance (ISR)
Border Patrol
Scientific
Atmospheric Research (NIST, NCAR)
Tornado/Hurricane & Arctic Research
Civil / Commercial
Disaster Communication & Intelligence (Fire)
Sensor Data Collection
NIST - National Institute for Standards and Technology (like PTB)
NOAA –
NCAR – National Center for Atmospheric Research

7. AUGNet Ad Hoc UAV Ground Network recuv.colorado.edu 241cm
Group 1
16cm
Group 2
241cm
Our research started with a real-life test bed. Wanted to know how MANETs behave in real world. Simulations have severe limitations and don’t accurately reflect what’s going on.
- Explain the test bed elements
Ad Hoc UAV Ground Network
recuv.colorado.edu

8. AUGNET as Delay Tolerant / Challenged Network
Plane banking Simultaneous end-to-end paths might not exist
Antenna configuration Links might be very lossy
Unmanned planes Nodes might move at high speeds
Links might have extremely long delays
Links might be intermittently up or down
From these networks come new challenges for protocols!

9. Using node mobility control to enhance network performance
Research Goal
Using node mobility control to enhance network performance
UAV2
Ferrying
Relay
Direct
UAV1
UAV3
GS1
- Now: question becomes more theoretical. What combination of these three modes optimizes performance?
- explain the modes
- explain controlled mobility of our planes -
GS2

10. Assumptions Controllable helper nodes Known communication demandsz y
Controllable helper nodes
Known communication demands
Single link perspective
Theoretical rather than implementation x λ
- The trade-offs of these modes will be explained with theoretical models. Not implemented yet.
- Assumptions might be unrealistic, but give insights into capacity of modes and trade-offs S R

11. Direct Communication Shannon capacity law Signal strength
Thermal noise (normalized)
- This might look simplistic, but Gaussian Channel is often used and Rappaport has verified (SINR) models.
- Prediction models of SINR are common in cellular systems; planning is done based on these models. Can be determined for given, fixed environment, e.g., a city.
- We are only using P(d) for ease of computation. Could switch model to be more accurate.
- R(SINR) is any monotonically decreasing function. Again, Rappaport has concrete models, but we stick to Shannon formula. -
Data rate

12. Direct transmission (zero relays)
Relay Network d dk S R
Direct transmission (zero relays)
End-to-end data rate: RR
Packet delay: τ = L/RR
Relay transmission

13. “Single Tx” Relay Model
a.k.a., the noise-limited case
t=0 S dk R

14. “Parallel Tx” Relay Model
a.k.a., the interference limited case
> Optimal distance between transmissions?
t=0
t=0
t=0 ρ S R

15. Conveyor Belt Ferrying ModelA F B

16. Optimizing “Single Tx”
Where is the trade-off?
dk vs. # of transmissions
Optimal number of relays:
Initially: rate increase higher than ‘relaying cost’ (put graph R over #relays here)
Then: additional relay decreases R
with

17. Optimizing “Parallel Tx”
Where is the trade-off?
interference vs. # parallel transports
Use Matlab! R k ρ
link reuse factor

18. “Triple Play”
2km
4km
8km
16km
eps= PN=10-15 W

19. Distance—Rate Phase Plot

20. Delay—Rate Phase Plot

21. Delay—Rate Animation

22. Sensor Data Collection
Challenges:
No end-to-end connection
Intermittent connectivity
Sensors and SMS unknown
SMS-3
Gateway-2
CDMA
Sensor-1
SMS-1
Sensor-2
SMS-2
Gateway-1
Sensor-3
Have sensors all on left side
Highlight ferrying to the right side SMS locations
Sparsely distributed sensors
Limited radio range, power
Multiple monitoring stations

23. Hardware Implementation
RTT 40ms, 15hrs sustained operation
Soekris SBC, embedded Gentoo Linux
Atheros miniPCI, Madwifi-ng driver

24. Functional Evaluation
FS2
83 on & off 84 85
FS1 82 81

25. Functional Evaluation II

26. Next Steps Ferry route planning with Reinforcement Learning
Multi UAV operations/hybrid with MAVs
UAV Swarming
Phased array antenna
WiMAX trial

27. Research and Engineering Center for Unmanned Vehicles (RECUV)
Daniel Henkel,
recuv.colorado.edu