DOI: 10.5923/j.jwnc.20130303.01
1
Wireless and M obile (WAM ) Systems Lab, Department of Electrical and Computer Engineering, Gainesville, FL 32611, USA
2
Center for High Performance Reconfigurable Computing (CHREC), Electrical and Computer Engineering, University of Florida
3
Space Systems Group (SSG) Department of M echanical and Aerospace Engineering University of Florida Gainesville, FL
Abstract Many planned Earth observation satellite missions use distributed satellite systems in constellation or cluster
orbits instead of one, large satellite with many instru ments. This network of satellites and ground stations allows data to be
efficiently downlin ked. Ho wever, to create a netwo rking protocol for constellations of small Earth observing satellites,
several limitations need to be addressed, including distributed topology management, slow downlin k data-rates, and single
point-to-point communication. Since distributed satellite constellations exacerbate the severity of these limitations, a
thorough analysis of a constellation's network performance is required to ensure that task objectives are achievable. In this
paper, we co mpared low Earth orbit (LEO) Earth observing satellite constellations using delay tolerant networking
protocols and the user datagram p rotocol (UDP). The constellations' inter-satellite lin ks and downlin ks were evaluated
using network metrics such as access window time, drop-rat io, and throughput. We simu lated these network metrics using
the Network Simu lator 3 (ns-3). Prev ious works have proposed satellite constellations for Earth observation, however,
constellations of s mall satellites have not been analy zed for network metrics. Results show that a sun-synchronous
constellation with a repeating ground track outperforms a flo wer constellation with respect to increased access time,
reduced drop-ratio, and higher throughput. The simulat ions also determined the optimu m med ia access control slot time
and packet transmission intervals for long distance satellite links. Additionally, we were able to co mpare the throughput
performance of UDP to abundle protocol by adding nodes with delay tolerant networking protocol imp lementations.
Keywords Co mmunicat ion networks, Delaytolerant networking (DTN), Network topology, Satellite co mmun ication,
Satellite constellations, Network simu lations, User datagram protocol (UDP), Veh icular networks, W ireless mesh networks
satellite missions could also benefit greatly fro m DTN method between nodes. The architecture of BP works as an
protocols because these satellites typically coordinate as overlay network and uses Endpoint Identifiers (EID) as a
constellations[2]. naming scheme to identify the final destination nodes.
In order to compare candidate constellation types for For DTN nodes, the bundle layer is immed iately below the
effective satellite mission design, the effects that the application layer. This stack position allows BP to
constellations typeshave on the inter-satellite link network masklayers fro m the application, as depictedin Figure 1.
performance must be evaluated. Simulat ion is common ly
used to evaluate network perfo rmance, and satellite packages
are available for OM NeT++, OPNET, QualNet, and the
Network Simulator (ns)[4]. However, most satellite
simu lation packages are for large satellites with mo re
powerful transmitters than the transmitters available on small
satellites. So me research focused on evaluating network
performance of specific protocols for satellite constellations
using network simu lators[5-6]. However, these studies have
focused on evaluating protocol performance o r optimizing a
single satellite constellation rather than comparing candidate
constellation types using network performance as a criterion.
The rest of the paper is organized as follows. Section 2
provides background on Delay Tolerant Networking
including User Datagram and Bundle Protocols and DTN
simu lation issues. In Section 3, we describe the
constellations modeled in satellite tool kit, and the
Figure 1. A diagram of the network stack for DTN and T CP/IP
parameters used for the network simu lation. Then, in Section
4, we detail the dropout ratio, throughput, and data rate The motivation fo r using DTN co mes fro m the fact that
results from simu lation. Last, in Section 5, we draw mu ltip le satellites in constellations can serve as relays,
conclusions regarding optimu m slot time values for BP/UDP allo wing more opportunities to downlink with ground
and discuss future work. stations. With varying bandwidth, a DTN can leverage the
maximu m data rate availab le in point-to-point connections,
instead of the weakest end-to-end rate[14]. Challenges in
2. Background on Delay Tolerant implementing these protocols on small satellites include
Networking (DTN) processing speed, power constraints, and topology
management.
2.1. User Datagram and Bundle Protocol
DTN is a suite of communicat ion protocols that has been 2.2. Eval uating BP/ UDP for Small Satellites
used to address challenging environmental conditions that Typically, s mall satellites have operated over the AX.25
create delay and intermittent connectivity. Examp les of these lin k layer protocol at point-to-point links[2]. Related pro jects
challenging network environ ments include sensor networks use Digital Smart Technologies for A mateur Rad io (DSTAR)
for wildlife tracking(ZebraNet)[7], connectivity in remote on satellites such as OUFTI-1[15]. For small satellites, DTN
and undeveloped regions (DakNet)[8], vehicular network can run on top of AX.25, and a DTN convergence layer
infrastructures (SUVNet )[9], social pro ximity-based adapter has been implemented for the lin k layer protocol
networks (Peoplenet)[10], underwater acoustic networks AX.25[16]. Ho wever, DSTAR and AX.25 protocols are
(NURC WHOI M icro-Modem)[11], and networking space exclusively point-to-point and downlink on VHF and UHF
missions (IRIS)[2]. spectrums. These spectrums do not contain the bandwidth
In 2002, the co mmon goals of MANETs and sensor needed for data rates above 38.4 kilobits per second[15].
networks led the Internet Engineering Task Fo rce (IETF) to Thus, an evaluation of running DTN protocols over UDP
form the DTN Research Group (DTNRG)[12]. The DTNRG for LEO constellations and cluster topologies is needed. By
works on protocol design, technical specifications, and lin king Linu x containers (LXCs)[17] through virtual local
develops code for the reference imp lementation, called area networks (LANs), we can determine the data rates and
DTN2. In 2007, the IETF published the request for comment access times for LEO satellites to form intersatellite links
(RFC) 5050, which defined a standard protocol for DTN and downlink with ground stations.
known as the Bundle Protocol (BP)[13]. Bundles defined in With our method, gateways and routers are authenticated
BP contain data that helps a network application move as DTN agents. Forwarding nodes also authenticate the
chunks of data larger than packets, known as bundles. BP sender data such that traffic can be stopped at the earliest
operates with non-IP (internet protocol) and IP transport point in the route. To deal with the high frequency of
layers, known as convergence layers, in a store and forward disruptions, bundles can be stored in forward ing nodes for a
Journal of Wireless Networking and Communications 2013, 3(3): 19-28 21
long period of t ime(e.g., minutes or days, as needed). In this for distributed processing and higher data relay to a ground
way, retransmission does not have to occur end-to-end. Store station.
options are configured at the bundle layer but can be set by
the application. 3.2. Network Management i n the Satellite Toolkit (STK)
To evaluate the network perfo rmance of arbitrary satellite
2.3. Past DTN Network Simul ations constellations, we used Network Simu lator 3 (ns-3), an
Network simu lators, such as DTNSim2 and the established network simulator[21]. To include DTN
Opportunistic Network Environ ment (ONE), have been used implementations on ns-3 nodes, our experiment used
in previous studies, however, DTNSim2 is no longer standardized, open source, physical and link layer simu lation
supported and ONE development was discontinued in modules, and leftthe higher layers as standard DTN
2011[18]. Often, these simulation p latforms require high implementations on virtualnodes.
computing resources. We mitigated the effects of limited LEO satellite access to
Additionally, for DTN simulators to be most effective for ground stations by employing the data mu le methodology
space inter-networking, the simulators must be highly proposed by[22]. The data mule methodology consists of
realistic. Many DTN simulators do not currently implement usinga set of source satellites to collect data, and a set of sink
realistic channel models or the networking stack. There is satellites to transport the collected data to a ground station.
also an issue of cross-simulator co mparab ility. Researchers In order to maintain relevancy between the two candidate
often create their own simu lators to test algorithms, making constellations, we designed each candidate constellation to
it difficu lt to compare new algorith ms with existing contain ninesource satellites and six sink satellites.The sink
algorith ms unless the new protocol is implemented on a satellites are larger, more powerfu l satellites with mo re
variety of simulators. Many DTN imp lementations have access time to the ground stations than the source satellites,
been written for various platforms. For our experiments, we but less access time to the target areas than the source
used DTN2 for a BP reference implementation. satellites.
Inclination 97.3 70
Right Ascension of
0 0
the Ascending Node
Satellites 1-9: 0, 40, Satellites 1-6: 0,
True Anomaly 80, 120, 160, 200, 60, 120, 180, 240,
240, 280, 320 300
(3)
un-slotted ALOHA-net as the link layer protocol, we used modules that use the nodes positions.
the mobile node model with each satellite represented as a
node.
The mobile node model is robust and supports a wide
range of protocols. Ho wever, since this model is most
appropriate for terrestrial wireless networks, we mod ified the
mobile node model to simulate our constellations. Creating a
mobile node simu lation typically consists of plotting the
nodes movements as tool command language (Tcl) scripts,
called scenarios. However, since three-dimensional (3D)
positioning is not supported, we could not simp ly write a
script to translate the 3D satellite movements into a scenario.
In order to integrate 3D movements, we mod ified the
positioning system. Unlike the other simu lation models, the
mobile node model is not easily modifiable and inhib its
direct rep lacement of the positioning system. To overco me
Figure 6. The ns-3 wireless transmission simulation model
this restriction, we wrote new modules to rep lace the existing
modules interfaced with the positioning system. The new The second primary use fo r the positioning system is to
modules used an external database of satellite positions to calculate the rad io propagation delay. When the
provide node position information. WirelessChannel module first receives packets from a
To use the STK constellation data in our new modules, we transmitting nodes PHY, the WirelessChannel module
used STK to expo rt our SSRGT and flower constellations as calculates the radio propagation delay for each receiving
comma separated value (CSV) files. The CSV files node by dividing the distance between the nodes in meters by
contained a position for each satellite at every minute. We the speed of light. To transmit the packet with delay, the
wrote a Python script to translate the satellites positions into WirelessChannel module requests the scheduler module to
a structured database and we wrote a C++ library to load, schedule a receive event.
cache, and linearly interpolate the satellites positions at Since the propagation methods in the WirelessChannel
times between the recorded minute positions. Using our C++ module are not inherent, we added a new module, Wireless
lib rary, we replaced any module that used node positions ChannelSTK, wh ich duplicated the WirelessChannel module.
with a version that used STK-exported data. We modified the WirelessChannelSTK modu le to use the
A nodes position has two primary uses in the mob ile node external satellite position databases for calculating the
model. First, the nodes position is used to calculate the distance between the nodes. Using the FreeSpaceSTK
receive power o f the t ransmissions that a node receives. Each module, the W irelessChannelSTK module, and STK-export
node contains a wireless PHY module that activates when ed constellation information, our modifications allow
any node in range transmits a packet. The wireless PHY wireless traffic to resemble t raffic between orb iting
module contains a propagation module that calculates the satellites.
packets receive power.
The propagation module uses the transmitting and 3.6. Bundle Protocol (BP) Experi ment Setup
receiving nodes positions to calculate the transmissions To simulate BP, and co mpare with UDP, DTN2 impleme
propagation distance. For simp licity and due to the direct ntations ran on virtual mach ines. Virtual machines create a
line-o f-sight transmissions in LEO, we used the FreeSpace native environ ment for the DTN software and provide virtual
propagation module, wh ich is based on the Friis transmission hardware on wh ich operating systems can be installed on a
Equation (1). The FreeSpace module uses the distance host system and configured according to the DTN software
between the transmitting and receiving nodes to calculate requirements with minimal impact to the underlying host. Of
each packets receive power. the various virtual platforms available, Oracles VirtualBo x
In order to interface with our external satellite position [32] was used because, with available open source editions,
databases, we added a new module, FreeSpaceSTK. modification is feasible. However, VMWare Player or
Modules can define Tcl co mmands, which Tcl scripts can VMWare Server[33] could also be used.
call to perform module-specific act ions. We wrote a For this experiment, ns-3 modules simulated physical
command handler fo r the FreeSpaceSTK module for a new DTNs up to the data link layer as WiFi-type given topology
command that logically lin ks a node to a corresponding and mobility patterns, ranges, and data rates.For the higher
satellite position database. The FreeSpaceSTK receive layers, connections to virtual machines were bridged using
power calculation uses the satellites position databases IEEE 802.3 in the host system.
instead of the mobile node positions to calculate node The entire experiment ran on one host machine. The
distance. FreeSpaceSTK doesnot, however, calculate the components includedns-3 and the virtual mach ines with the
radio propagation delay. Figure 6 shows the structure of the DTN2 network software stack and DTN applications, subject
Journal of Wireless Networking and Communications 2013, 3(3): 19-28 25
to testing and integration, run on top. ns-3installations running on Ubuntu Linu x. To run our
The experimentmodeled co mmun ication channel properti experimental simu lations, we created two Tcl scripts: one
es, such as delay, transmission rate, error, and packet loss script defined the nodes using our flower constellation
distribution, with detailed scheduling. The simu lation positions and the other script defined the nodes using our
allo wed configurat ion for mobility patterns of wireless nodes, SSRGT constellation positions.
networking device properties at the physical and link layer, The Tcl scripts specified the protocol for each satellite
logging, and packet tracing. If required, new models could be nodes network layers. We used the simulation modules for
constructed and used in the simulat ions. each nodes MAC layer since 802.11b-1999 has acceptable
As a main network simu lation and server Linu x platform, long-range performance and a wide range of availab le
Fedora Linu x 64-b it distribution was selected as the host commercial-o ff-the-shelf hardware. For each nodes PHY
mach ines operating system. We used the Ubuntu Linu x layer, we used the standard wireless PHY modu le with the
distributionfor the virtual machines due to the availability of propagation configured to use our FreeSpaceSTK module.
the DTN2 reference imp lementation. The host hardware We configured the FreeSpaceSTK module in one of the
could be any standard personal co mputer or laptop Tcl scripts to use the satellite position database for the flower
compatible with the operating system with at least 8 GB of constellation and the other Tcl script to use the satellite
RAM and disk space to allow fo r many virtual machines. position database for the SSRGT constellation. To simulate
Different hardware platforms of interest, such as Broadcom the commun ication channel, we used our WirelessChannelS
or ARM,could be integrated by physically connecting these TK module. The nodes defined by the Tcl scripts behaved
platforms into the system through an access point. The ns-3 like satellites in a constellation and could support traffic
comes with a number of channel and device models, fro m most agents.
including WiFi and WiMax. The DTN2 systems and We wrote a Tcl scenario to generate sample traffic for the
application software that run on the virtual nodes come fro m simu lation. In the Tcl scenario, each non-sink node
the DTNRG. generated constant bit rate (CBR) traffic over a UDP agent to
The Wifi network models are built into the ns-3 source each of the six sin k nodes. To prevent the source nodes that
code shown in Figure 7. Wi-Fi is considered the common were out of range of any sink satellites fro m transmitting, we
type of link for the DTN tested. Bundles were assumed that satellites could detect the presence of sink
onemegabytein size. satellites within 2000 km. Source nodes only transmitted
data when they were within 2000 km of a sink node or the
ground station in Gainesville, Florida.
Figures 8 and 9 show that for both our flo wer and SSRGT SSRGT constellations between 10bps and 80bps.
constellations, the optimal slot time ranges fro m 500us to Figure 11 shows the packet drop-ratio for the SSRGT and
1500us, while the standard 802.11b-1999 slot time causes flower constellations as a function of source traffic density.
the MAC to drop nearly all packets. While the round-trip The flower constellation maintained a similar throughput as
propagation delay between distant nodes may be higher than the SSRGT constellation even though the SSRGT
1500us, the optimu m 802.11b-1999 slot t ime reflects a constellation dropped fewer packets than the flower
tradeoff between unnecessary slot time delay for nearby constellation. The higher packet drop-ratio for the flower
nodes and propagation delay for distant nodes. Since the constellation as compared to the SSRGT constellation,
optimal slot time range is similar for the sink satellite combined with the constellations similar throughputs,
connections to both the ground station and the source suggests that the flowerconstellation has more opportunities
satellites, sink satellites can use the same PHY for than the SSRGT constellation to transmit data fro m long
transmitting to sink nodes and the ground station. distances. Both constellations suffered very lo w throughputs,
For the source satellite to sink satellite connections, the which is expected due to the weak transmitters and long
SSRGT constellation dropped significantly fewer packets distances involved in the simulat ion.
than the flower constellation. With the slot time set at 640us,
the SSRGT constellation dropped fewer than 50% of the
packets, while the flower constellation dropped more than
75% of the packets. For both the flower and SSRGT
constellations, the high drop-ratio in Figures 8 and 9
demonstrate that any network protocol used by the satellites
must perform reliab ly with intermittent connections.
Figure 11. Comparison of packet drop ration versus source traffic density
spacing of 2,000 km, the average data rate is 10.44 Kb it/sec. REFERENCES
In comparison,the UDP/IP network of nodes at distances of
[1] A.Toorian,K.Diaz,andS.Lee,The cubesat approach to space
2,000 km had a data rate of 0.08 Kb it/sec. access, in the IEEE Aerospace Conference, pp. 114, 2008.
Table 5. Comparison of the average data rate between DT N/UDP and [2] P. M uri, and J. M cNair. A Survey of Communication
UDP
Sub-systems for Intersatellite Linked Systems and CubeSat
M issions. Journal of Communications[Online], 7.4 (2012):
Environment Average Date Rate
290-308. Web. 24 Jan. 2013
DT N/UDP 50 km spaced cluster 755.85 kbit/sec [3] M . Smith, S. Seager, C. Pong, J. Villasenor, G. Ricker, D.
DT N/UDP 2000 km spaced constellation 10.44 Kbit/sec M iller, M . Knapp, G. Farmer, and R. Jensen-Clem,
UDP 2000 km spaced constellation 80 bit/sec Exoplanetsat: detecting transiting exoplanets using a
low-cost cubesat platform, in Proceedings Vol. 7731. SPIE,
2010.
5. Conclusions and Future Work [4] L. Xiangqun, W. Lu, L. Lixiang, H. Xiaohui, X. Fanjiang, and
C. Jing, Omnet++ and maximum-based protocol simulator
In this paper, we developed a method for co mparing the for satellite network, in the IEEE Aerospace Conference, pp.
network performance for constellations designed using the 1 9, 2011.
satellite toolkit (STK) based on sink time, drop-rat io, and
[5] A. Houyou, R. Holzer, H. M eer, and M . Heindl,
throughput. Using this method, we designed and compared Performance of transport layer protocols in leo pico-satellite
the network performance of two novel s mall satellite constellations, Technical report, University of Passau, Tech.
constellations: a flower constellation and a sun-synchronous Rep., 2005.
repeating ground track (SSRGT) constellation. In order to
[6] H. Bedon, C. Negron, J. Llantoy, C. Nieto, and C. Asma,
compare the constellations network performance, we Preliminary internetworking s imulation of the qb50 cubesat
modified the Network Simulator 3s (ns-3s) mob ile node constellation, in Communications (LATINCOM ), IEEE
model to simulate co mplex satellite constellations. Results Latin-American Conference on, pp. 1 6, September 2010.
revealed that as the satellites opportunistically communicat [7] P. Juang, H. Oki, Y. Wang, M . M artonosi, L. Shiuan Peh, and
ed during a week in simulat ion time, the satellites in the D. Rubenstein, Energy-efficient computing for wildlife
SSRGT constellation dropped fewer packets than the tracking: design tradeoffs and early experiences with
satellites in the flower constellation. During a period of 500 ZebraNet SIGPLAN Not. 37, 10, pp. 96-107, October 2002.
ms to 1 second, the SSRGT satellites showed a higher [8] A. Pentland, R. Fletcher, A. Hasson, DakNet: rethinking
throughput. Further, modifying ns-3 with a delay tolerant connectivity in developing nations, Computer , vol.37, no.1,
bundle protocol (BP) increased data rates substantially. Our pp. 78-83, Jan. 2004.
simu lation files and related scripts described in the art icle are
[9] H. Huang, P. Luo, M . Li, D Li, X. Li, W. Shu, M . Wu,
available upon request. Performance Evaluation of SUVnet With Real-Time Traffic
Future work includes benchmarking delay tolerant Data, IEEE Transactions on Vehicular Technology, vol.56,
networking (DTN) protocols on embedded hardware viable no.6, pp.3381-3396, Nov. 2007.
for s mall satellites. Future tests can compare the
[10] M . Motani, V. Srinivasan, and S. Nuggehalli, PeopleNet:
performance and determine likely limit ing factors for engineering a wireless virtual social network In Proceedings
hardware p latforms.In addition,we will incorporate mu ltip le of the 11th annual international conference on M obile
ground stations[35] and compare the network performance computing and networking (M obiCom '05). ACM , New York,
and coverage of satellite constellations to satellite cluster NY, USA, pp. 243-257, 2005.
formations in which several satellite nodes fly in a single [11] L. Freitag, M . Grund, S. Singh, J. Partan, P. Koski, K. Ball,
orbit, with one satellite node in the cluster serving as a sink. "The WHOI micro-modem: an acoustic communications and
navigation system for multiple platforms," Proceedings of
MTS/IEEE OCEANS, pp.1086,1092 Vol. 2, 17-23
ACKNOWLEDGEMENTS September 2005.
[12] A. M cM ahon, S. Farrell, Delay and Disruption-Tolerant
This work was supported in part by the National Science Networking IEEE Journal of Internet Computing, vol.13,
Foundation (ECCS-0901706). Any opinions, ndings, and no.6, pp. 82-87, November 2009.
conclusions or reco mmendations expressed in this material
are those of the author(s) and do not necessarily reect the [13] K. Scott, S. Burleigh, Bundle Protocol Specification, IETF
RFC 5050 2007.www.rfc-editor.org/rfc/rfc5050.t xt
views of the National Science Foundation.We would also
like to thank NASA Office of Chief Technologists Space [14] M . Khabbaz, C. Assi, W. Fawaz, Disruption-Tolerant
Technology Research Fellowship (NSTRF) grant Networking: A Comprehensive Survey on Recent
Developments and Persisting Challenges, Communications
#NNX11AM73H for supporting this work.
Surveys & Tutorials IEEE , vol.14, no.2, pp.607-6402012.
28 Paul M uri et al.: User Datagram and Bundle Protocol for Distributed Small Satellite Topologies
[15] B. Klofas, J.Anderson, K. Leveque,A Survey of Cubesat [25] G. Tyc, J. Tulip, D. Schulten, M . Krischke, and M . Oxfort,
Communication Systems In the 5th Cubesat Developers The rapideye mission design, Acta Astronautica, vol. 56, no.
Workshop, San Luis Obispo, California, pp. 1-29, 2009. 1-2, pp. 213219, 2005.
[16] J. Ronan, K.Walsh, and D. Long, Evaluation of a dtn [26] D. M ortari, M . Wilkins, Flower constellation set theory. part
convergence layer for the ax.25 network protocol Proc. of i: Compatibility and phasing, Aerospace and Electronic
the 2nd International Workshop on M obile Opportunistic Systems, IEEE Transactions on, vol. 44, no. 3, pp. 953962,
Networking, New York, NY, ACM , pp. 72-78 2010. 2008.
[17] M . Helsley, LXC: Linux Container Tools IBM [27] P. M uri,J. M cNair, J. Antoon, A. Gordon-Ross, K. Cason, N.
developerWorks Technical Library. February Issue 2009. Fitz-Coy, Topology design and performance analysis for
http://www.ibm.com/developerworks/linux/library/l-lxc-cont networked earth observing small satellites in the IEEE
ainers/ M ilitary Communications Conference, M ILCOM ,pp.1940-1
945, 7-10 Nov. 2011
[18] J. Karvo and J. Ott, Time scales and delay-tolerant routing
protocols InProceedings of the third ACM workshop on [28] Analytical Graphics, Inc. (AGI), August 2011.
Challenged networks (CHANTS '08). ACM , New York, NY, http://www.agi.com/.
USA, 33-40 2008.
[29] P. M uri, O. Challa, J. M cNair, Enhancing small satellite
[19] J. Wertz, H. M eissinger, L. Newman, and G. Smit,M ission communication through effective antenna system design in
geometry; orbit and constellation design and management the IEEE M ilitary Communications Conference, M ILCOM ,
M icrocosm Press Dordrecht: Kluwer Academic Publishers, pp.347-352, Oct. 31 2010-Nov. 3 2010.
El Segundo, CA 2001.
[30] J. Jensen, Remote sensing of the environment: An earth
[20] C.P. Bridges, L. Sauter,P. Palmer, Formation deployment resource perspective p. 544, 2000.
and separation simulation of multi-satellite Scenarios using
SatLauncher, in the IEEE Aerospace Conference, pp.1-9, [31] W. DAmico, M . Lauss, Wireless Local Area Network
5-12 M arch 2011 Flight Demonstration for High Doppler Conditions, John
Hopkins APL Technical Digest, Vol 25, No 4, pp. 335-342,
[21] The Network Simulator - ns-3, August 2012.http://www.ns 2004.
nam.org/
[32] Oracle VM VirtualBox, April 2013. https://www.virtualbo
[22] R. Shah, S. Roy, S. Jain, and W. Brunette, Data mules: x.org/.
M odeling and analysis of a three-tier architecture for sparse
sensor networks, Ad Hoc Networks, vol. 1, no. 2-3, pp. [33] "VM ware Player, August 2011.http://www.vmware.com/.
215233, 2003.
[34] Vladimirova, T., Wu X., Bridges C.P., Development of a
[23] V. Salomonson, A. Park, An overview of the landsat-d Satellite Sensor Network for Future Space M issions, in IEEE
project with emphasis on the flight segment, in LARS Aerospace Conference, Big Sky, MT, 2008 pp.1-10
Symposia, p. 236 1979.
[35] B. Preindl, H. Page, V. Nikolaidis, Genso: The global
[24] P. Slater, Survey of multispectral imaging systems for earth educational network for satellite operations, in Proceedings
observations, Remote Sensing of Environment, vol. 17, no. 1, of the 59th International Astronautical Conference, 2008.
pp. 85102, 1985.