Hydra Multihop Testbed
Hydra is a flexible wireless network testbed developed in the WNCG at UT Austin. It’s development was a collaborative effort between research groups led by Prof. Scott Nettles and Prof. Robert Heath. Hydra is unique because it supports both a software defined PHY including MIMO and OFDM functions, as well as a software defined MAC protocol. The PHY is implemented using GNU radio and the Ettus USRP Clasic; the MAC is implemented in Click! Hydra has an IEEE 802.11n PHY supporting at present 2×2 operation along with the DCF MAC protocol.
All Hydra source code is available for download. Updated September 2014: Downloads are available here. Note that this software was developed with the original USRP and may require substantial redevelopment to work with the newer USRPs or GNU radio releases.
Several of our recent publications are summarized below.
R. Daniels, C. Caramanis, and R. W. Heath, Jr., “Adaptation in Convolutionally-Coded MIMO-OFDM Wireless Systems through Supervised Learning and SNR Ordering,” IEEE Trans. on Veh. Tech., vol. 59, no. 1, 2010, pp. 114-126.
This paper proposes a new methodology for link adaptation based on a concept from machine learning known as supervised learning. The work in this paper was inspired by our observation that link adaptation in real MIMO-OFDM systems is hard. Supervised learning provides a way to implement link adaptation in a MIMO-OFDM including spatial mode adaptation, modulation selection, and code rate selection through performance-related observations. The approach described in this paper has been implemented on Hydra.
Wonsoo Kim, M. O. Khan, K. T. Truong, Soon-Hyeok Choi, R. Grant, H. K. Wright, K. Mandke, R. Daniels, R. W. Heath, Jr., and S. Nettles and R. W. Heath, Jr., “An Experimental Evaluation of Rate Adaptation for Multi-Antenna Systems,” Proc. of IEEE INFOCOM, April 19-25, 2009, Rio de Janeiro, Brazil, pp. 2313 – 2321.
This paper proposes extensions of two well-known link adaptation algorithms: Receiver-Based AutoRate (RBAR) and Auto Rate Fall back (ARF) to link adaptation in MIMO-OFDM. The challenge in extending both of these algorithms is that MIMO systems also have a spatial dimension. If there is a packet failure in a MIMO system, it is not clear if the rate for the current spatial configuration should be changed, or if the spatial mode should be adjusted. For example, a packet failure may mean switching from spatial multiplexing to space-time block coding, rather than just decreasing the per-stream data rate. We implemented these algorithms on Hydra and showed how they have some limitations. This motivates more sophisticated approaches to link adaptation.
R. Daniels and R. W. Heath, Jr., “An Online Learning Framework for Link Adaptation in Wireless Networks,” Proc. of the Information Theory and Applications Workshop, San Diego, CA, February 8-13, 2009, pp. 138-140.
This paper proposes an online approach to machine learning for rate adaptation. Compared with supervised learning, this online framework uses real-time measurements to update the rate-adaptation classifier. There are many interesting tradeoffs with online implementations including exploration vs. exploitation (how often should the system explore new rates to keep the mappings up to date), and database management (how should the database tradeoff new data versus diverse data). Online learning has been implemented and demonstrated on Hydra.
R. Daniels, K. Mandke, S. Nettles and R. W. Heath, Jr., “Throughput/Delay Measurements of Limited Feedback Beamforming in Indoor Wireless Networks,” Proc. of IEEE Global Telecommunications Conf., pp. 1-6, New Orleans, LA, USA, Nov. 30 – Dec. 4, 2008.
This paper verifies experimentally the exponential relationship between throughput loss and delay, as derived in some of our earlier work. It uses Hydra to measure goodput for different delays. The fit between theory and practice is extremely good.
K. Mandke, R. Daniels, Soon-Hyeok Choi, S. Nettles and R. W. Heath, Jr., “Physical Concerns for Cross-Layer Prototyping and Wireless Network Experimentation,” Proc. of the Second ACM International Workshop on Wireless Network Testbeds, Experimental Evaluation and Characterization, pp. 11-18, Montreal, Canada, September 10, 2007.
This paper summarizes some of the challenges faced when implementing a cross-layer prototype. It highlights specific features of wireless networking protocols impacted by temporal scaling, measurement reciprocity, and cross-layer adaptation. The paper also provides recommendations to guide researchers in setting up interesting and useful wireless experiments. Three concerns for wireless experimenentation are addressed, namely: ambient interference, RF hardware profiling, and fading properties of the wireless channel.
K. Mandke, Soon-Hyeok Choi, Gibeom Kim, R. Grant, R. Daniels, Wonsoo Kim, R. W. Heath, Jr., and S. Nettles, “Early Results on Hydra: A Flexible MAC/PHY Multihop Testbed,” Proc. of IEEE Vehicular Tech. Conf. , pp. 1896-1900, Dublin, Ireland, April 23 – 25, 2007.
This is one of the key Hydra references. It describes the Hydra architecture in detail including hardware and software. Of particular interest, I presented a demo of Hydra during the conference presentation.
We won the Grand Prize in the 2008 WinCool Demo Contest, held in conjunction with the Third International Workshop on Wireless Network Testbeds, Experimental Evaluation, and Characterization, on conjunction with ACM MobiCom 2008. The prize was one for a demonstration of online learning for rate adaptation using Hydra.
We have been fortunate to have several outstanding sponsors support the development of the Hydra prototype including the National Science Foundation under grant CNS-626797, Office of Naval Research (ONR) under grant number N00014-05-1-0169, US Army Research Laboratory under grant number W911-NF-08-1-0438, and the DARPA IT-MANET program, Grant W911NF-07-1-0028. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the aforementioned sponsors.