JOINT DEVELOPMENT OF MAC AND PHYSICAL LAYER ALGORITHMS FOR AD HOC NETWORKS WITH MIMO LINKS
Mary Ann Ingram and Ragupathy Sivakumar
School of Electrical and Computer Engineering
Georgia Institute of Technology
We wish to study ad hoc wireless networks of nodes, such that each node has multiple and independently controlled antenna elements. A transmit and receive pair of such nodes constitutes a multiple-input-multiple-output (MIMO) link, which is well known to achieve extraordinary spectral efficiency through simultaneous transmission of multiple data streams. A network of MIMO links also has unprecedented flexibility. For example, co-PI Ingram has shown that for two closed-loop MIMO (CL-MIMO) links operating simultaneously in the same channel, any of a continuum of possibilities for relative qualities of service in the two links can be “dialed up” through judicious choice of a constraint in a distributed optimization algorithm at the physical layer [1,2]. In contrast to CL-MIMO, open-loop MIMO (OL-MIMO) links do not exploit channel information at the transmitter. While these links have lower performance than CL-MIMO links, they do not have the overhead of signaling channel information to the transmitters. However, CL-MIMO and OL-MIMO have not been compared under conditions of severe co-channel interference, where CL-MIMO’s spatial filtering at the transmitter may significantly improve network throughput. Therefore, one of the objectives of the proposed work is to determine if the overhead of CL-MIMO is justified in an ad hoc network.
Just what the “dial” settings should be when new nodes want to transmit is a medium access control (MAC) protocol question. Due to the unique characteristics of wireless networks including the hidden and exposed terminal problems, collision avoidance techniques through control packet exchanges are typically employed by MAC protocols. The CSMA/CA (Carrier Sense Multiple Access with Collision Avoidance) protocol uses a request-to-send (RTS)- clear-to-send (CTS) handshake prior to the data transmission. CSMA/CA (or its variants) is typically employed in ad-hoc networks since such networks lack the services of an established backbone infrastructure. However, CSMA/CA is designed for a network with omni-directional links and can be shown to be highly ineffective in networks with smarter physical layers. For example, co-PI Sivakumar has developed MAC protocols for ad-hoc networks with switched beam links, and has developed a switched-beam aware variant of CSMA/CA called OP-RTS/D-CTS(G) that outperforms the baseline protocol significantly. Figure 1 compares the per-flow throughput performance of the baseline CSMA/CA protocol, the OP-RTS/D-CTS(G) protocols, and a simple extension of CSMA/CA (D-RTS/D-CTS) with directional transmissions, for a network of 25 nodes. Each node has four antenna elements, and 5 end-to-end CBR/UDP flows with a data rate of 4Kbps are employed. As is evident, physical layer aware MAC schemes can leverage any flexibility offered by underlying technologies and can thus improve network performance.
We propose to develop a MAC protocol for ad-hoc networks with multiple input multiple output (MIMO) links. While our past work in developing MAC protocols for networks with switched beam links has provided us with insights that will help us in the effort, the problem to be addressed has challenges and issues that are unique to networks with MIMO links. In the rest of the section, we present some of these challenges and therein outline the key tasks involved in the proposed work.
In a network with omni-directional links, a transmission inhibits any other transmission in the neighborhood of both the transmitter and the receiver (assuming CSMA/CA). In a network with switched beam links, a transmission inhibits any other transmission in an area that is a function of the direction of transmission, the transmission range, and the beam-width. In contrast, for networks with MIMO links, transmissions in the same vicinity can simultaneously occur as long as the transmit-receive pairs of the interfering links have adequate resources at the receivers to suppress the interference, and for CL-MIMO, adequate resources at the transmitters to avoid making interference.
Assuming linear processing, a receiver has adequate resources if the number of its antennas, or equivalently, its number of adaptive degrees of freedom (DOFs), is at least the number of received desired and interfering data streams. Fewer DOFs may be sufficient if some of the interference streams are weak. Thus, adapting CSMA/CA to this environment would entail enriching the basic RTS-CTS handshake with more information that conveys the number of streams to be used for the upcoming transmission, and the available number of DOFs at the transmit and receive nodes. For a contending link, a transmission can be initiated as long as the number of streams to be used does not overwhelm the receivers of existing transmissions. We call such a mode of operation as the Master-Slave mode. We will investigate how such DOFs-related information can be effectively used to optimize network performance. Another problem in the scenario portrayed above is that a certain number of available DOFs indicated by a node can potentially be contended for by all of the node’s neighbors. We will investigate how the resource partition can be performed effectively.
Co-PI Ingram has shown that two interfering CL-MIMO links operating simultaneously (albeit at a capacity lower than their respective maximums) can have a larger aggregate capacity than when they operate in a TDM (time-division multiplexing) fashion [Fall VTC ‘01]. Consider a scenario in which node A is transmitting to node B using 4 streams, and the maximum number of DOFs at a node is 4. On an interfering link between nodes C and D, it might be beneficial – from the network standpoint – for node C to start transmitting on the CD link, and for both links to adapt to an operating point where less than 4 streams are transmitted [PIMRC, winters]. Note that this is different from the earlier described mode of operation where interfering links will not operate in a manner that would overwhelm receivers of existing transmissions. We refer to such a mode of operation as the Peer-Peer mode. We will investigate how average case performance measures provided by the PHY layer algorithms can be effectively used to arrive at MAC layer decisions that would facilitate such adaptive operation of MIMO links.
In both the above modes of operation, it is possible that certain nodes lie at a certain distance from an ongoing transmission such that they cannot decode the control messages transmitted, but can cause interference at the receiver if they were to initiate a transmission. In such cases, the quantizing of control messages to reflect the available DOFs will serve no purpose. We will study the impact of distance of the interfering link on the number of DOFs required at the receiver to suppress the interference. We will also consider the use of a guard DOF that would protect receivers from interference caused by such links.
Another problem in the operation of the MIMO links is the required feedback channel between the receiver and the transmitter to exchange channel state information. Such feedback is critical to the effective functioning of the adaptive PHY layer algorithms outlined earlier in the proposal. We will investigate how such a feedback channel can be established to work alongside the MAC scheme, and how it can be optimally realized.
Brief Vitae
Dr. Ragupathy Sivakumar
Dr. Sivakumar received his M.S. and Ph.D. degrees in Computer Science from the University of Illinois at Urbana-Champaign in 1998 and 2000 respectively. He joined Georgia Tech as an assistant professor in the School of Electrical and Computer Engineering in 2000. His research interests are in the areas of next generation wireless systems, multi-hop ad-hoc networks, transport layer protocols, network quality of service, and programmable networks. He has over 30 publications in refereed conferences and journals. He has received several honors and awards for outstanding research and teaching including the David J. Kuck outstanding Ph.D. thesis award for his work on “Dynamic Networks: Architectures and Algorithms”. Sivakumar is the guest editor for the ACM MONET (Mobile Networks) Journal’s Special Issue on Quality of Service in Heterogeneous Wireless Networks, local arrangements chair for the ACM International Conference on Mobile Computing and Networking (Mobicom 2002) conference, vice program chair for the IEEE International Conference on Communication (ICC 2003) conference, and an editor for the Elsevier-Science Computer Networks Journal. He has served as a reviewer for numerous IEEE and ACM conferences and journals. Sivakumar is a member of the IEEE and the ACM.
Relevant Publications
Spine Based Routing in Ad-hoc Networks, Cluster Computing Journal 1998
CEDAR: Core Extraction Distributed Ad-hoc Routing, IEEE JSAC Special Issue on Ad-hoc Networks 1999
Enhancing Ad-hoc Routing using a Virtual Infrastructure, IEEE INFOCOM 2001
Improving Fairness and Throughput in Multi-hop Wireless Networks, IEEE ICN 2001
Performance Comparison of Cellular and Multi-hop Wireless Networks: A Quantitative Study, ACM SIGMETRICS 2001
IEEE 802.11 over Multi-hop Wireless Networks: Problems and Perspectives, IEEE VTC 2002
Submitted
Medium Access Control for Ad-hoc Networks with Switched Beam Antennas, Submitted to Globecom 2002
IEEE 802.11 over Multi-hop Wireless Networks: Problems, Perspectives, and Solutions, submitted to IEEE JSAC Special Issue on Home LANS 2002
Sivakumar has several publications in the area of wireless ad-hoc networks, including some in the specific area of medium access control. 1, 2, and 3 are papers on routing protocols for ad-hoc networks. 4, 5, 6, and 8 are papers on medium access control performance in ad-hoc networks with omni-directional antennas. 7 is a paper on a CSMA/CA variant for ad-hoc networks with switched beam antennas.
Dr. Mary Ann Ingram
Dr. Ingram received her Ph.D. degree in Electrical Engineering from the Georgia Institute of Technology in 1989, and soon after joined the faculty there as an Assistant Professor. Her early work was in optical communications, and later in airborne radar. She became an Associate Professor in 1997, and since then, she has focused on the application of array antennas, or “smart antennas,” to wireless communications. Specific activities include system performance analysis, channel sounding and modeling, and prototype development.
Some Relevant Publications:
Demirkol, M.F. and Ingram, M.A., “Power-controlled capacity for interfering MIMO links,” Proceedings of the 54th IEEE Fall Vehicular Technology Conference (Fall VTC 2001), Vol. 1, pp. 187—191, Atlantic City, NJ, October 7-11, 2001.
Demirkol, M.F. and Ingram, M.A., “Control using capacity constraints for interfering MIMO links”, submitted to IEEE International Symposium on Personal, Indoor and Mobile Radio Communications), Lisbon, Portugal, on Sep. 15-18, 2002.
Jiang, J. S. and Ingram, M. A., “Path models and MIMO capacity for measured indoor channels at 5.8GHz,” to appear in the Proceedings of the Conference on Computational Electromagnetics and Antenna Technology (ANTEM), Montreal, Quebec, July 31-August 2, 2002.
Kuo-Hui Li and Ingram, M.A., “Space-time-block-coded OFDM systems with transmit beamformers for high-speed indoor wireless communications,” Proceedings of the 52nd IEEE Fall Vehicular Technology Conference (Fall VTC 2000), Vol. 5, pp. 2473 –2477, Boston, MA, September 24-28, 2000.
Li, K-H., Ingram, M.A., and Rausch, E.O., “Multibeam antennas for indoor wireless communications,” IEEE Transactions on Communications.
Li, K-H, Ingram, M.A., ``Impact of clustering in statistical propagation models on link capacity,'' to the IEEE Transactions on Communications.
Jo, J-H., Ingram, M.A., and Jayant, N., “Angle clustering in indoor space-time channels based on ray-tracing,” Proceedings of the 54th IEEE Fall Vehicular Technology Conference (Fall VTC 2001), Vol. 4, pp. 2067--2071, Atlantic City, NJ, October 7-11, 2001.
Current Grants and Research Contracts:
“Broadband MIMO OFDM Wireless Access,” National Science Foundation, with Profs. Stuber, Barry, Li and McLaughlin, and Dr. Pratt, 10/01-9/06, Synchronization and channel estimation for measured vector OFDM signals over high-mobility channels; also MIMO channel measurements and modeling
“High speed indoor wireless Prototype,” Yamacraw Research Center, 7/99-6/02. A large team project to develop a prototype for a high-speed wireless link. Prof. Ingram has been responsible for the design and implementation of the multibeam beamformer and dual-beam selection algorithm for a space-time block coded OFDM link.
“Array-to-array propagation measurements,” Georgia Tech Broadband Institute, 7/00-6/02. Measurement, modeling, and capacity analysis of indoor MIMO links, uses synthesized arrays at both ends of the link.
“Ad hoc networks of array-to-array links” Georgia Tech Broadband Institute, with Professor Sivakumar, 7/00-6/02. Theoretical and simulation-based study of algorithms to control the antenna element weights of interfering closed-loop MIMO links, and MAC algorithms for ad hoc networks of nodes with switched-beam antennas.
Share with your friends: |