3400 International Drive, NW 
Washington, DC · 20008
Tel: 202-364-4700
Fax: 202-364-8554
E-mail: university@mil3.com
Web: www.mil3.com

OPNET is a registered
trademark of Opnet Technologies, Inc.
© 2001 Opnet Technologies, Inc

Professor: Dr. Zygmunt J. Haas
Department: School of Electrical Engineering, Cornell University

Modeling Mobile Ad-Hoc Networks (MANETs) Using the OPNET Modeler / Radio Simulation Environment (Marc Pearlman, School of Electrical Engineering, Cornell University)

1. Introduction to Mobile Ad-Hoc Networks (MANETs) 

Mobile Ad-hoc networks (MANETs) are self-organizing wireless networks composed of mobile nodes and requiring no fixed infrastructure. The limitations on power consumption imposed by portable wireless radios result in a node transmission range that is typically small relative to the span of the network. To provide communication throughout the entire network, each node is also designed to serve as a router. The result is a distributed mutli-hop network with a time-varying topology. Because ad-hoc networks do not rely on existing infrastructure and are self-organizing, they can be rapidly deployed to provide robust communication in a variety of hostile environments. This makes ad-hoc networks very appropriate for providing tactical communication for military, law enforcement and emergency response efforts. Ad-hoc networks can also play a role in civilian forums such as electronic classrooms, convention centers and construction sites. With such a broad scope of applications, it is not difficult to envision ad-hoc networks operating over a wide range of coverage areas, node densities and node velocities. 

2. Framework for MANET Simulation in OPNET 

The OPNET modeler / radio has been a valuable tool in our group's development and analysis of protocols for mobile ad-hoc networks. Using the network editor, we can instantly deploy a random distribution of network nodes. For initial testing purposes, our MANETs may contain only a handful of nodes. After verifying that our protocols operate correctly, they are generally applied to larger networks, on the order of 50 - 500 nodes. Because we generally assume that all ad-hoc network nodes are alike, all node attributes can be configured simultaneously, saving a great deal of time. Rather than predefine node trajectories in the network editor, we specify a customized mobility process, which dynamically updates each node's position during run-time. 

The ad-hoc network is characterized not only by the attributes of its constituent nodes, but also by the environment in which it operates. The environment has an effect both node mobility and signal propagation. These factors are taken into account by our mobility process model. For example, our nodes can be configured to conform to an urban "grid" road system or meander according to a versatile Gauss-Markov mobility model [1]. Additionally, our mobility process confines nodes to the physical / logical boundaries of their network. 

Channel behavior is controlled by the simulator's 14 stage radio channel pipeline stages. By default, received signal power is calculated based on a free-space model, which is not realistic enough for our purposes. Fortunately, OPNET provides a number of alternatives for channel modeling, all of which we have used during different stages of our ad-hoc network development. First, the simulator comes bundled with a more realistic (and configurable) cellular radio propagation model [A]. Second, we have used the op_mko tool to compile our own pipeline stages, allowing us introduce additional features (such as shadow fading [2]) into our propagation model. Finally, we have also tried the Terrain Modeling Module (TMM), which uses ray tracing to calculate the path loss over a specified terrain. Unfortunately, the relatively low antenna heights and short separation distances of our ad-hoc network nodes exceed the bounds of the TMM's "Longley-Rice" model. While not viable for our ad-hoc network research, we've been able to apply the TMM to our research in cellular systems. 

The details of each node's behavior are reflected in its node model. In particular, each MANET node contains an instance of the customized mobility process, along with a full protocol stack. We employ a variety of traffic models, (both OPNET standard/contributed models and custom-built models), depending on the particular network scenario. End-to-end connectivity is enabled by a customized implementation of the TCP/IP protocol suite, suitable for our ongoing research in mobile networking (including, but not limited to ad-hoc networking). Our IPv4 implementation provides support for loose and strict IP source routing. An implementation of Mobile IP [B] supports general user mobility, allowing users to travel between different networks. Finally, we've included protocols such as IGMP [C] and DVMRP [D] , to support our research of reliable mobile multicasting [3]. 
Each MANET node is equipped with a low power half-duplex radio transceiver (matching transmitter and receiver pair), typically operating on the order of 1[Mbps]. Typically, a single data channel is reused throughout the entire network. Transmission access to this channel is provided by a media access control protocol specially designed for the MANET environment (described in section 3). In situations where reliable data link level communications needs to be enforced, OPNET's Sliding Window Protocol (SWP) model is included. 

3. New Protocols for the MANET Environment

Within the framework described above, our group has developed novel protocols for MANET media access control, routing, and mobility management. This work was motivated by the poor performance of traditional strategies when applied to the hostile MANET environment. In the following sections, we each introduce these protocols and provide some insight into their OPNET implementations. 

3.1. Medium Access Control for Mobile Ad-Hoc Networks: Dual Busy Tone Multiple Access (DBTMA)

The attenuation characteristics of the wireless medium, combined with additional environmental obstructions, can result in different channel conditions at neighboring nodes. As a result, traditional carrier sense multiple access (CSMA) protocols, which sense the status of the channel at the transmitting node, does not necessarily provide an accurate view of the intended receiver's availability. Improper carrier sensing by the transmitter leads to two distinct problems, which contribute to a significant loss in throughput. In the "hidden terminal problem", the transmitter senses the channel as idle, when the intended receiver is actually busy. The subsequent packet transmission results in a collision at the receiver. In contrast, the "exposed terminal problem" occurs when the transmitter senses its channel as busy, when in fact the intended receiver is free. In this case, the transmitter unnecessarily backs off from the channel, wasting available resources. 

In order to provide effective wireless media access control, a feedback mechanism is needed to provide nodes with information about the status of the channel at their neighbors. Our group has addressed this problem with the Dual Busy Tone Multiple Access (DBTMA) [4] protocol. Prior to transmitting a data packet, a node secures access to the channel through an RTS/CTS handshake (performed on a separate control channel). After completing the RTS/CTS handshake, the transmitter sends the data packet, while simultaneously activating a transmit busy tone. The intended receiver, in turn, activates a separate receive busy tone as soon as this data transmission is detected. The dual busy tones are used to block attempts by neighboring nodes to access a channel already in use. In particular, the transmit busy tone prevents neighbors of the transmitter from accepting incoming RTS requests. Likewise, the receive busy tone prevents the receiver's neighbors from initiating the RTS/CTS handshake. This effectively prevents the "hidden terminal problem" associated with wireless channel access. In addition, DBTMA inherently avoids the "exposed terminal problem", by permitting neighboring nodes to transmit data simultaneously to different (and available) receivers. 

DBTMA consists of two control mechanisms: the RTS/CTS handshake and the busy tone activation / detection. In analyzing the performance of DBTMA, it was useful to study the impact of the two mechanisms separately. As a result, two corresponding OPNET models were first constructed. The implementation of the RTS/CTS exchange, based on existing MACA[E] and MACAW[F] protocols, is a straightforward event-driven packet exchange. In contrast, the busy tones require a little ingenuity, since analog signaling is not directly supported by the OPNET simulator. We represent busy tones with packets, transmitted on separate transmit tone and receive tone channels. A busy tone is activated by sending a very long packet on the appropriate busy tone channel. To terminate the tone, the transmitter is sent an "abort" command on the appropriate busy tone channel (via the op_ima_obj_command() function). Busy tone detection is performed by examining the "signal lock" attribute on the receiver's busy tone channel. A special received power pipeline stage is needed to prevent receivers from locking on to a "busy tone" that is too weak. 

3.2. Routing for Mobile Ad-Hoc Networks: Zone Routing Protocol (ZRP) 

In "traditional" wired networks, routers proactively track the topology of the entire network. Changes in link status or node reachability are effectively flooded through the entire system. Such an approach can have disastrous consequences in MANETs, where network topology may change frequently and bandwidth is at a premium. Large amounts of route update traffic can be generated, potentially overloading the system and shutting down communication. Furthermore, because routes are updated so often, much of the acquired information is never used, making the route updates a true waste of channel capacity. In contrast to proactive routing protocols, reactive routing protocols acquire routes only on-demand. While some delay is incurred in route acquisition, the amount of overhead traffic is generally much less than proactive routing protocols, because only useful routing information is collected. However, reactive routing protocols are still inefficient because the search mechanism relies on the flooding of route query packets throughout the entire network.

Bearing these inefficiencies in mind, we propose the hybrid proactive / reactive Zone Routing Protocol (ZRP) [5]. Each node proactively maintains routes within a local region of the network (which we refer to as the routing zone). Knowledge of this routing zone topology is leveraged by the ZRP to improve the efficiency of a reactive route query/reply mechanism. The ZRP can be configured for a particular network through adjustment of a single parameter, the routing zone radius. 

The ZRP consists of two components, the proactive IntrAzone Routing Protocol (IARP), and the reactive IntErzone Routing Protocol (IERP). The ZRP philosophy allows for a great degree of flexibility in the implementation of either component. Using OPNET, we have developed link state and distance vector versions of the IARP. Their basic operation is similar to traditional protocols like OSPF [G] and RIP [H], except that the propagated route information is restricted to the scope of a routing zone (controlled by means of a hop count field). Likewise, link state and distance vector models of the IERP have also been constructed, based on the recently proposed reactive Dynamic Source Routing (DSR) [I] and Ad-Hoc On Demand Distance Vector (AODV) [J] protocols. 

Initial evaluation of the ZRP focused on the role of the routing zone radius on protocol performance. The routing zone radius was made a simulation attribute, which was varied to determine the optimal ZRP configuration for any given type of network. In addition, by setting the routing zone radius to its extreme values, we were able to compare the ZRP's performance with a variety of existing purely proactive (i.e. OSPF, RIP) or purely reactive (i.e. DSR, AODV) routing protocols. Current ZRP development is focussed on the development of algorithms to estimate the locally optimal routing zone radius for each node, and an accompanying protocol to perform the necessary updates in local routing information. 

3.3. Mobility Management for Mobile Ad-Hoc Networks: Uniform Quorum Systems (UQS)

In the absence of an established hierarchy or informational infrastructure, the Zone Routing Protocol, (described above), is an effective approach for discovering routes in MANETs. Alternatively, logical structure can be imposed on the ad-hoc network to simplify the complex route discovery process. A common approach is to organize geographically close nodes into a subnet, with an elected gateway to provide access beyond the subnet. Rather than learn the topology of the entire network, nodes need only track the local topology of their subnet. Additionally, gateways need to maintain routes for the routing backbone (formed by the gateways of other subnets). 

The presence of a routing backbone can lead to network congestion, as all traffic between nodes needs to pass through gateways. To address this problem, location databases (rather than gateways) can be distributed through the network to form a virtual backbone. These databases serve only as containers for location storage and retrieval. Routing is carried out in the flat network structure, involving every node in the network. That is, the routes do not necessarily go through the databases. However, the location provided by the databases can provide vital information to the routing protocol, so that route searching is more localized. 

Because MANETs are susceptible to partitioning, it is advantageous to have location information replicated in multiple databases. This idea has led to the formulation of our group's Uniform Quorum Systems (UQS) [6]. In this system, databases are logically organized into quorums, such that any two quorums intersect in a constant number of databases. Storage and retrieval of location information is performed by writing to or reading from all databases in any quorum. 

The UQS's quorum construction algorithm has been thoroughly analyzed, and the affect of quorum size/intersection on the cost of information access (and information unavailability) is well understood. Work is now underway to incorporate the UQS into our OPNET ad-hoc networking framework. Individual protocols handle quorum construction, virtual backbone maintenance, location identification and location storage/retrieval. Once fully implemented, the UQS will be evaluated against other mobility management models. 

4. Summary 

The OPNET Modeler / Radio provides a rich environment for developing protocols for mobile ad-hoc networks (MANETs). The product ships with a large set of standard communication protocols, along with number of models for simulating the wireless channel and wireless communication systems. Some of these models have been applied directly to our MANET framework. In other cases, the unique nature of the MANET environment has forced us to either modify existing implementations of standard protocols, or develop totally new protocols. Because the source code for OPNET process and channel models is open and well documented, the creation of new models has been a fairly straightforward process. 

Without a doubt, OPNET's modular philosophy has been the key to turning our independently developed processes into a unified MANET framework. Individual components can be updated and interchanged, allowing us to select mobility, application and channel models that are appropriate for any given MANET scenario. Moreover, our new protocols can be replaced with models of traditional protocols, allowing us to easily perform comparative simulation analyses. Finally, we have been able to share our models with OPNET users from other research institutions, providing us with valuable feedback and independent verification of our protocols' performance. We hope that in the future, there will be some degree of interoperability between OPNET and other popular event driven network simulators (such as NS-2, PARSEC, Entrapid, etc.), allowing us to easily share our work with all members of the MANET development community. 

5. References

5.1. Wireless Networks Laboratory -- Cornell University

[1] B. Liang and Z.J. Haas, "Predictive Distance-Based Mobility Management for PCS Networks," IEEE INFOCOM'99, New York City, NY, March 1999. 

[2] M.R. Pearlman, "A Fully Correlated Model for Shadow Fading", in preparation. 

[3] Z.J. Haas, R.H. Gau, and E. McCarthy, "Reliable Mobile Multicast Protocols", internal technical report, 1999.

[4] J. Deng and Z.J. Haas, "Dual Busy Tone Multiple Access (DBTMA): A New Medium Access Control for Packet Radio Networks," IEEE ICUPC'98, Florence, Italy, October, 1998. 

[5] Z.J. Haas and M. R. Pearlman, "The Performance of Query Control Schemes for the Zone Routing Protocol," ACM SIGCOMM '98, Vancouver, BC, Canada, Sept. 1998. 

[6] Z.J. Haas and B. Liang, "Ad Hoc Mobility Management with Uniform Quorum Systems," IEEE Transactions on Networking, April 1999. 

5.2. Additional References

[A] W. C. Y Lee, Mobile Cellular Telecommunications: Analog and Digital Systems, 2nd edition, McGraw Hill, 1995. 

[B] C. Perkins, "IP Mobility Support," RFC 2002, October 1996.

[C] S. Deering, "Host Extension for IP Multicasting," RFC1112, August 1989.

[D] S. Deering, C. Partridge, and D. Waitzman, "Distance Vector Multicast Routing Protocol," RFC 1075, November 1998.

[E] P. Karn, "MACA - A New Channel Access Method for Packet Radio," in ARRL/CRRL Amateur Radio 9th Computer Networking Conference, pp. 134-140, ARRL, 1990. 

[F] V. Bharghavan, A. Demers, S. Shenker, L. Zhang, "MACAW: A Media Access Protocol for Wireless LANs," ACM SIGCOMM'94, September 1994. 

[G] J. Moy, "OSPF Version 2," RFC 2328, April 1998. 

[H] G. Malkin, "RIP Version 2," RFC 2453, November 1998. 

[I] D.B. Johnson and D.A. Maltz, "Dynamic Source Routing in Ad-Hoc Wireless Networking," in Mobile Computing, T. Imielinski and H. Korth, editors, Kluwer Academic Publishing, 1996.

[J] C.E. Perkins and E.M. Royer, "Ad-Hoc On Demand Distance Vector Routing," IEEE WMCSA'99, New Orleans, LA, February 1999. 


Last modified by Zygmunt Haas on February 13, 2001