A Testbed for Experiments with Sensor/Actuator Networks
ABSTRACT We describe a table-top testbed for experiments in mobile sensor networks. The testbed includes static nodes as well as robotic mobile nodes. The static beacon nodes are used for localization and multihop network setup. The static snooper nodes serve the purpose of debugging and visualizing the experiment at a later time. The snoopers passively listen to packets. Ground truth is provided by an overhead vision system, and the data logged by the snoopers are replayed to evaluate/debug the network using NAM.
- SourceAvailable from: Karthik Dantu
Conference Paper: Robomote: enabling mobility in sensor networks[Show abstract] [Hide abstract]
ABSTRACT: Severe energy limitations, and a paucity of computation pose a set of difficult design challenges for sensor networks. Recent progress in two seemingly disparate research areas namely, distributed robotics and low power embedded systems has led to the creation of mobile (or robotic) sensor networks. Autonomous node mobility brings with it its own challenges, but also alleviates some of the traditional problems associated with static sensor networks. We illustrate this by presenting the design of the robomote, a robot platform that functions as a single mobile node in a mobile sensor network. We briefly describe two case studies where the robomote has been used for table top experiments with a mobile sensor network.Information Processing in Sensor Networks, 2005. IPSN 2005. Fourth International Symposium on; 05/2005
- [Show abstract] [Hide abstract]
ABSTRACT: We introduce an algorithm which detects and traces a specified level set of a scalar field (a contour) on a plane. A network of static sensor nodes with limited communication and processing are deployed in a planar environment along with a mobile node which can both sense and move. As the mobile node moves through the environment, it computes the local spatial gradient of the field by communicating with its immediate neighbors in the static sensor network. The algorithm causes the mobile node to perform gradient descent on the scalar field till it arrives at a location on the desired contour. From this point onwards, the algorithm drives the mobile node to trace the desired contour without departing from it. Experiments in simulation indicate that the required contour is found with reasonable accuracy (between 80-90%) for networks with node degree of greater than or equal to six. Our results also indicate that the paths generated by our algorithm are near-optimal in terms of the distance traversed by the mobile node. Our preliminary experimental results with a physical robot show that our algorithm is feasible.Robotics and Automation, 2007 IEEE International Conference on; 05/2007
A Testbed for Experiments with Sensor/Actuator Networks
Mohammed Rahimi Rohit Mediratta
Robotic Embedded Systems Laboratory
Computer Science Department
University of Southern California
Los Angeles, CA 90089-0781, USA
Karthik Dantu Gaurav S. Sukhatme
We describe a table-top testbed for experiments in mo-
bile sensor networks. The testbed includes static nodes
as well as robotic mobile nodes. The static beacon nodes
are used for localization and multihop network setup. The
static snooper nodes serve the purpose of debugging and
visualizing the experiment at a later time. The snoopers
passively listen to packets. Ground truth is provided by
an overhead vision system, and the data logged by the
snoopers are replayed to evaluate/debug the network us-
Keywords: sensor network, sensor/actuator network,
experiments with sensor/actuator networks on the bench-
top. The testbed allows experimenters to populate the
benchtop with small (upto 25 nodes) multi-hop wireless
networks(typicallyupto 3 hops)of extremeembeddedde-
vices . In addition, we have roboticized some of the
sensor nodes such that they can move about on the testbed
and autonomously re-position themselves . An over-
head camera system  provides ground truth for local-
ization, and a set of snooper nodes capture network traffic
for offline analysis. Since the setup is indoors we can eas-
ily controlthe illuminationintensity (for experimentswith
the light sensors on the nodes), the locations of the nodes,
camera calibration, and the like. The testbed is logically
partitioned into a structured grid of 32 squares, with each
squarebeing1 squarefootin area. We describethe testbed
setup in the following four sections.
Beacons: the beacons are evenly distributed in the
environment and know their x,y coordinates. These
beacon nodes are used for localization as well as for
multihop network setup.
Snoopers: these nodes serve the purpose of debug-
ging and visualizing the experiment at a later time.
The snoopers passively listen to packets.
Overhead vision: ground truth is provided by the vi-
sion system using mezzanine, an overhead camera
system looking down onto the testbed.
Post Processing: the data logged by the snoopers are
replayed to evaluate/debug the network using NAM.
The testbed (Figure 1) is a 8’x 4’ tabletop bound by
a wooden fence 3/4” high. The tabletop is colored white
to aid the ground truth system in identifying the robots
based on color coding. The tabletop stands on top of three
columns 40” high.
The beacons are built around the UC Berkeley Rene
mote . The beacons are evenly spread over the testbed
and serve to make up the static part of the network. Each
beaconhas oneAtmega 163Lmicrocontrollerbased mote,
shielding to minimize interference and multi-path effects.
To ensure similar power levels for all beacons in the net-
work we provide power using a 3V 500mA DC adapter.
The beacons are programmed using Tiny OS ; a
component-based small operating system. Each beacon
executes a routing component as well as a localization
The programming board 2 was fabricated specially to
allow us to package it inside a case to minimize RF in-
terference. The programming board is similar to the pro-
gramming board from Berkeley, and differs only in the
physical layout. As shown in 2 the programming board
contains a DB25 male connector for programming any
mote on the board. It also contains a DB9 female con-
nector to read data over the serial port which helps in de-
The programming board was manufactured by Cross-
bow Technologies ??.
Figure 1: The testbed.
Gold Star Technologies
Super Shield Conductive Coating 841-340g
JJB Series Permanent 1/4 Wave
3vDC/500mA Regulated AC-to-DC Adapter 273-1756
RocketPort Multiport Serial PCI Octa card 95850-5
Figure 2: Manufacturer contacts
Figure 3: The programming board.
The ATMEL Atmega 163 mote is a 4 Mhz low power
microcontroller based mote. It contains a 16KB flash in-
struction memory,512 bytes of SRAM, a 256 KB of EEP-
ROM as secondary storage. There is also a RFM TR1000
radio for communication. More details and hardware de-
scription can be found at
The beacons are placed inside special cases 3 to min-
imize Radio Frequency (RF) interference. These cases
have interfaces for programming the motes and for de-
bugging over the serial connection. The casing also has
a power jack, to enable us to maintain constant power lev-
els using a DC power adapter. The case is essentially the
outer cover of a DB25 to DB 9 converter available off the
shelf from any electronics shop. Our programming board
was designed with the casing already available to us. We
hadtodrillholesinthecasingto allowusto providepower
to the programmingboard as well as to allow space for the
antenna to stick out.
To minimize RF interference and multi-path fading ef-
fects we coated our casing with conductive Nickel coat-
ing ??. This conducting coat absorbs small traces of RF
thus greatly reducing multipath fading effects. One to two
mil coating provides 40db-50db shielding across a fre-
quency range of 5 to 1800 Mhz. Since the coating is elec-
trically conducting,we painted the casing on the inside, to
prevent any possibility of a short circuit.
Figure 4: The case.
We purchased monopole antennae from Antenna Fac-
antennae. The antenna works for the 902-928Mhz fre-
To maintain a homogeneous setup of calibrated static
nodes, it is important to ensure that all nodes have the
same power levels available to them at all times. It can
be easily shown  that the communicationcapabilities of
any mote is greatly hampered by reducing power levels.
We power all our beacons with 3V 500mA regulated DC
power adaptors purchased from Radioshack ??.
Our testbed has snooper nodes which are placed at
specific places to sniff the packets being exchanged on
the testbed. Snoopers are used for debugging, localiza-
tion service, (the vision system is used for localization
ground truthing) as well as replaying the network packet
exchanges at a later time using our NAM post processing
tool. EachsnooperexecutestheGENERIC BASEcompo-
nent  and has a serial cable connected to one of the port
on an 8 port serial card on a central machine (a PC next to
the testbed). The serial card is polled on this machine to
generate a dump (log) of almost all packets exchanged on
the testbed. We ensure that packets, which may be read by
multiple snoopers simultaneously, do not get logged mul-
tiple times. To undertake duplicate suppression we main-
tain a log of upto the last ten packets that were logged.
Whenever a new packet is being logged into the log file, it
is first checkedagainst the last ten packets to ascertainthat
the packet was not already logged by some other snooper.
The snooping system is thus composed of
Atmel 8535 motes
Comtrol Rocketport 8 port serial card.
Serial Extension cables.
The programming board contains a DB25 male con-
nector for programming any mote on the board. It also
contains a UART chip and a DB9 female connector to
read data over the serial port which helps in debugging.
The programming board was manufactured by Crossbow
3.2ATMEL 8535 MOTES
The ATMEL 8535 mote is a 4 mhz low power micro-
controller based mote. It contains a 8KB flash instruction
memory, 512 bytes of SRAM, a 512 bytes of EEPROM as
secondary storage. There is also a RFM TR1000 radio for
communication for upto 10kbps data rate. More details
and hardware description can be found at
3.3COMTROL ROCKETPORT 8 PORT PCI SERIAL
The card uses a 36Mhz ASIC (Application Specific In-
tegrated Circuit) that enables the RocketPort ?? to trans-
mit and receive data at rates up to 921 Kbps full duplex
across all ports simultaneously. It provides us with 8 ports
to which the snoopers communicate, to send the data they
into the same log file.
3.4SERIAL EXTENSION CABLES
Theserialextensioncables areusedto connecttheDB9
serial interface on the programming boards to the serial
DB9 interface on the Y cable of the Rocketport PCI serial
The vision system is used as ground truth for the lo-
calization component executing on the Robomote. The
vision system returns accurate x,y coordinates as well as
orientation of the robots. We used Mezzanine1  for our
vision system. The system is freely available at player-
Our camera is placed directly over the testbed. A wide
angled camera provides us with a warped picture of the
testbed, howeverMezzanine is equippedwith a dewarping
utility to remove any dewarping effects.
We used a regularoff the shelf webcam with a wide an-
gled lens. The camera was mounted on the roof, 6’ above
the testbed, looking down on the testbed. It was attached
totheroofusingVelcro. Thissetupis pronetoaging(cam-
era angle looking down will change over time) but Mez-
zanine provides an easy GUI based calibration tool which
override any loss of calibration due to movement of either
the testbed or the camera. The camera provides an output
in YUV space in the NTSC formatontothe compositesig-
nal pin of the frame grabber. The camera is powered by a
5V 1000mA DC adaptor.
We use a 640x480 interlaced NTSC PCI framegrabber.
The camera’s output serves as a composite input for the
framegrabber. The framegrabber captures 30 frames per
second. However we use only 1 in 3 frames for an accept-
able user interface display. The frame grabber is queried
by Mezannine which processes the captured frames.
5 Post Processing
The timestamped log file obtained by the snoopers is
used for network replay, debugging and for analysis. The
log file is converted from the raw packet format to a for-
mat acceptable to the Network Animator . The post
processor assumes that prior knowledge of the layout of
the testbed is knownand available in an options file. Since
we had the liberty to manipulate the packet payload, we
transmitted each packet with a pre-defined data structure,
such that source, destination and present coordinate infor-
mation was contained in each packet. This information
was obtained from the log file and replayed by converting
it into NAM format. The post processor is responsible for
displaying to the user the packet exchange that took place
when an experiment was being conducted.
The post processor has an options file which is used
for initializing the Network Animator input file, for speci-
fying the size of the testbed and for specifying the scal-
ing factor from testbed space to NAM space (they use
different units).Every node’s (beacon and robomote)
TOS LOCAL ADDRESS is required in the options file or
else that node’spackets will be ignoredbythe post proces-
sor. The post processor itself has the option of specifying
whether the user would like to follow the path that each
robot followed during the experiment or whether just the
current position at any time is required.
A very simple RF-localization scheme has been imple-
mented on the testbed using the beacon nodes placed two
feet apart. The transmit power levels of the beacon nodes
have been fine tuned to perform optimal triangulation .
A query packet is sent from either a mobile robot node or
a stationary node. The beacons reply to this query (if they
hear it) reporting their coordinates. The querying node lo-
calizes itself to the centroid of the location values of the
replies it receives. The software on the beacon nodes has
We are in the process of incorporating signal-strength
based weighting of the beacons received to provide a bet-
ter location estimate .
Figure 5: The view from the overhead camera
This work is supported in part by ONR grant N00014-
00-1-0638 under the DURIP program and by NSF grants
 David E. Culler, Jason Hill, Philip Buonadonna,
Approach to Embedded Software for Tiny Devices
 Sibley, G.T., Rahimi, M.H. and Sukhatme, G.S.
Robomote:A Tiny Mobile Robot Platform for Large-
Scale Sensor Networks 2002 IEEE International
Conference on Robotics and Automation, pp. 1143-
 Andrew Howard Mezzanine User Manual; Version
 Jason Hill, Robert Szewczyk, Alec Woo, Seth Hol-
lar, David E. Culler and Kristofer S. J. Pister System
Architecture Directions for Networked Sensors, pp.
93-104, ASPLOS 2000
trin In Proceedings of the 21st International Con-
ference on Distributed Computing Systems (ICDCS-
21), Phoenix, Arizona, USA. April 16-19 2001.
Whitehouse Towards a Location-Based Context-
Aware Sensor Infrastructure CS Division, EECS
Department, University of California at Berkeley
Klemmer, Sarah Waterson andKamin
 TheNetworkSimulator at USC/ISI
Figure 6: A screenshot of the mezzanine tool.
Figure 7: A screenshot of the post processing using NAM