4th Scandinavian Workshop on Wireless Ad-hoc Networks, May 4-5, Stockholm, 2004

Ad hoc UAV-Ground Network (AUGNet) Test Bed

Timothy X Brown, Brian Argrow, Cory Dixon, Sheetalkumar Doshi, Phillip Nies
University of Colorado, Boulder, 80309-0530
timxb@colorado.edu

1 Introduction:
Communication networks between and through aerial vehicles are a mainstay of current
battlefield communications. Present systems use specialized high-cost radios in
designated military radio bands. Current aerial vehicles are also high-cost manned or
unmanned vehicles. Small low-cost Commercial Off-The-Shelf (COTS) radio equipment
combined with powerful computer processing can be mounted on small (10Kg)
unmanned aerial vehicles (UAVs) to form ad hoc wireless networks that will be a key
technology in multi-UAV, tactical-UAV, and swarming-UAV operations for military,
scientific, and commercial applications.
Ad hoc networks have been a rich area of research in recent years. But, the bulk of this
research has been in theory and simulation with some limited attempts at real test beds
[LNN02,MBJ99,SBB03]. There is a significant need for careful studies of real ad hoc
network behavior. We expect this need is even stronger as we introduce non-standard
nodes such as nodes mounted on UAVs.
In this paper we consider ad hoc networks consisting of ad hoc nodes on the ground and
ad hoc nodes mounted in small UAVs, which we denote Ad hoc UAV-Ground Networks
(AUGNets). We envision two broad AUGNet scenarios as shown in Figure 1. In the first
scenario, an ad hoc network of ground nodes is disconnected because of distance and/or
terrain. The UAV, with a better view of the nodes, maintains connectivity as an ad hoc
relay. In the second scenario, a small UAV because of power and payload constraints has
limited communication range which in turn may limit operational range. Ad hoc relaying
between multiple UAVs extends communication range. In this paper we describe our
efforts to construct such an ad hoc network. It builds on our earlier work in 802.11b ad
hoc network protocols and small UAV construction. We are in the process of
constructing a test bed to evaluate the performance of such networks.

Figure 1: Scenario 1 (left): ad hoc networking with the UAV increases ground node connectivity.
Scenario 2 (right): ad hoc networking between UAVs to increase operational range.
2 Approach:
An AUGNet poses several challenges. Ad hoc nodes can be in a variety of
configurations. Some potential configurations include a fixed site mounted on a high
pole, a mobile node mounted on a vehicle, a personnel-carried node, or an aerial node
mounted in a small UAV. The design should be modular enough so that it can be applied
to all of these configurations. The UAV nodes present special considerations. The UAV
can usually communicate with most of the ground nodes.1 Ad hoc routing algorithms will
tend to route the traffic through the UAV thus limiting communication to the UAV
bottleneck bandwidth. The UAV may also have additional flight navigation and control
radios that can interfere with the ad hoc communication.
Our approach consists of four efforts: (1) ad hoc network software, (2) communication
hardware, (3) UAV, and (4) test bed monitoring architecture.
2.1 Mesh Network Radio and UAV:
The ad hoc network software combined with the communication hardware we denote the
mesh network radio (MNR). The MNR is shown in Figure 2. The MNR hardware
consists of a Soekris single board computer, Orinoco 802.11b card, a Fidelity-Comtech
bidirectional amplifier with up to 1W output, and a GPS. A key to our approach is that all
nodes whether mounted in a UAV, at a fixed ground site, or mounted on a vehicle node,
use the same core MNR. The radios only differ in their packaging. This greatly simplifies
our software and hardware development.

21cm

16cm

Figure 2: Mesh Network Radio (MNR) (left). MNR mounted in environmental enclosure for vehicle
or fixed ground mounting (center). MNR mounted in UAV (foreground right)
The MNR runs the dynamic source routing protocol (DSR) [JoM96] communicating with
other nodes via 802.11b. We chose DSR because its routing is on-demand so idle nodes
do not consume resources, and because it uses source routing which facilitates our test
bed monitoring. We implemented DSR ourselves using the Click modular router

1
The small UAV is capable of high altitude flight. But, AMA RC rules limit the UAV altitude to 500ft
(150m) above ground level for our testing. Even so, a UAV communicating with a ground node will be
more likely to have line-of-site transmission and less interference with the ground than other ground nodes.
[KMC00,DBB02]. With our own implementation we are free to modify the protocol to
include embedded monitoring (see next section) and to address the special UAV
considerations. To avoid traffic converging on the UAV node, we use a simple
hierarchical routing strategy. All ground nodes are level 1 nodes and all UAV nodes are
level 2. All route requests use one or two attempts. In the first attempt, only level 1 nodes
participate in the route request. If the initial attempt fails the node uses a second route
request using all nodes. Combined with source routing, the ad hoc nodes can control their
use of the UAV bandwidth. Note that our use of the UAV differs with the use for ad hoc
networking in [GPL00] where the UAV is large, flies at high-altitude (20km), and has a
separate radio for UAV-ground communication.
The UAV was constructed by the University of Colorado’s Aerospace Department for
this task using carbon fiber composite construction techniques. The plane has a 2.4m
wingspan and take-off weight is 15kg, 5kg of which is payload. It is designed to be
flexible for different flight configurations, have multi-hour endurance for uninterrupted
flight testing, and, have a large payload to carry the MNR and other instrument packages.
The UAV body is shown in Figure 2.
We tested the UAV 72MHz control radio and UAV 900MHz telemetry radio in the
presence of the 2.4GHz 802.11b radio. Measurements with a spectrum analyzer and a
Berkeley Variatronics Yellowjacket WLAN analyzer showed no discernable interference
power between the bands. Even with antennas placed within a few centimeters no
excessive packet loss was observed for either system. The 72MHz, 900MHz and 2.4GHz
antennas are mounted on different parts of the UAV to provide some isolation.
2.2 Test Bed Monitoring
The test bed must solve several challenges in order to be effective. The test bed data
should be available in real time. The UAV radio nodes are small size, with limited power
and payload. The ad hoc networking is complex with behavior distributed across the ad
hoc nodes. The test bed should scale to 10’s of monitored nodes. The monitoring should
have minimal impact on the normal operation of the network.
These constraints limited some approaches. The real time collection requirement
precludes simply storing monitoring data on each node to be collected after the
experiment. The limited UAV payload forces the use of the ad hoc network itself for
collecting monitoring data rather than a separate radio for monitoring feed back. The
distributed behavior suggests that data has to be centrally collected and correlated
between nodes. The scaling and interference constraints mean that the monitoring should
use minimal computing, storage, and bandwidth resources.
The monitoring approach is shown in Figure 3. The ad hoc nodes consist of the MNR
running a monitoring process that collects data on packets sent, packets received, and
packets discarded due to congestion induced overflow. This data is tagged with the GPS
time and location and sent out at 10 second intervals. This interval is chosen long enough
so that the impact of monitoring packets on other network traffic is minimal.
 University of Table Mountain
Colorado

Monitor Ad hoc radio

Server Fixed Fixed
Site 1 UAV Site 2
Fixed Nodes Fixed
Node Node
Test Bed Mobile
Internet Internet
Gateway Nodes

Remote
Monitor Wired Network

Figure 3: Monitoring Architecture

Monitoring packets are sent like other data packets to any one of the fixed nodes in the
network using an anycast address recognized by any of the fixed nodes. The fixed nodes
in turn send them using a wired Ethernet interface to the test bed gateway and on to the
monitoring server.
The monitor server has several purposes. The first is to archive the monitoring data. The
second is to produce derived statistics. For instance, the number of packets lost on a link
is the difference between the number of packets sent on the link and the number
successfully received on the link. The third purpose is to provide a playback function
where different aspects of an experiment can be replayed by remote users for analysis
purposes. The playback feature uses a Java interface so that remote users can analyze the
data from any web browser. A screen shot of the interface is shown in Figure 4.

Figure 4: Screenshot from the remote monitoring software. Situation map is on the top left showing
MRN locations (squares) at Table Mountain site. Status messages are displayed on the bottom left.
Performance graphs on highlighted MNR are shown on right.
The test bed is operated at the Table Mountain National Radio Quiet Zone [TM04], a
federal test bed for radio equipment near Boulder, Colorado. Table Mountain is a 4x2.5
km2 unobstructed fenced facility large enough for ad hoc network and UAV operations.

3 Status:
The ad hoc routing software has been implemented and tested on the MNR. The first
UAV has been completed and flight tested. A second UAV will be completed in April.
The monitoring test bed can collect data and a web interface provides access to the data.
Early AUGNet testing has begun with initial performance tests of throughput, range, etc..
Variations on the basic scenarios in Figure 1 will be made over the spring.

Acknowledgements
This work is supported through a contract with the L3-Comcept Corporation. The
AUGNet project is supported by Tom Bateman, Jack Elston, Harvey Gates, Sushant
Jadhav, Vivek Jayaraj, Jay Jones, and Roshan-George Thekkekunnel. Special thanks to
John Ewan, Wayde Allen, and Ken Allen of ITS for supporting our access to Table
Mountain.

References:
[DBB02] Doshi, S. Bhandare, S., Brown, T. X , “An On-demand minimum energy routing protocol for a
wireless ad hoc network,” Mobile Computing and Communications Review, vol. 6, no. 2, July 2002.

[GPL00] Gu, D.L., Pei, G., Ly, H., Gerla, M., Zhang, B., Hong, X., UAV Aided “Intelligent Routing for
Ad-Hoc Wireless Network in Single-Area Theater,” Wireless Communications and Networking
Conference, 2000. WCNC. 2000 IEEE ,
Volume: 3 , 23-28 Sept. 2000 Page(s): 1220 -1225 vol.3

[JoM96] Johnson, D., Maltz, D., “Dynamic Source Routing in Ad HocWireless Networks,” Mobile
Computing, Chapter 5, pp. 153-181, Kluwer Academic Publishers, 1996.

[LNN02] H. Lundgren, D. Nielsen, J. Nordstrom, and E. Tschudin, “A large-scale testbed for reproducible
ad hoc protocol evaluations,” IEEE WCNC, 2002.

[KMC00] E. Kohler, R. Morris, B. Chen, J. Jannotti, and M. F. Kaashoek, “The click modular router,”
ACM Transactions on Computer Systems, vol. 18, no. 3, pp. 263–297, August 2000.
http://www.pdos.lcs.mit.edu/click/

[MBJ99] Maltz, D., Broch, J., Johnson, D., “Experiences Designing and Building a Multi-Hop Wireless Ad
Hoc Network Testbed,” CMU School of Computer Science Technical Report CMU-CS-99-116, March
1999.

[SBB03] Sanghani, S. Brown, T. X, Bhandare, S., Doshi, S. ``EWANT: The Emulated Wireless Ad Hoc
Network Testbed,'' IEEE Wireless Communications and Networking Conference (WCNC), 16-20 March,
2003.

[TM04] Table Mountain national radio quiet zone, www.its.bldrdoc.gov/home/programs/tm_quiet_zone