Download as pdf or txt
Download as pdf or txt
You are on page 1of 4

IMPLEMENTATION OF COOPERATIVE COMMUNICATIONS USING SOFTWARE DEFINED RADIOS

Michael Knox, Elza Erkip


Department of Electrical and Computer Engineering Polytechnic Institute of New York University Brooklyn, NY, USA AbstractCooperative communications leverages the spatial diversity available in a wireless network enabling multiple radio nodes work together to improve the overall system performance. When a destination receiver combines the signal from an originating source with the associated signals from relay nodes, significant improvements in the bit error rate performance can be achieved. This paper details the measured bit error rate performance of a three-node cooperative communication system operating in a software defined radio testbed. The measured performances of several types of cooperative physical layer protocols are compared to similar systems operating over a single wireless link. The measured results include cooperative systems operating with a maximum ratio combining technique and two cooperative coded systems using hard decision decoding.
Index Termscooperative communications, coding, software defined radio, WARP, WARPLAB cooperative

Wireless Open-Access Radio Platform (WARP) developed by Rice University [8]-[11]. This paper describes the PHY layer cooperative communications testbed housed at the Wireless Information Systems Laboratory (WISL) located at the Brooklyn campus of Polytechnic Institute of New York University. The goal of the reported experiments is to study effects of the wireless environment and cooperation protocols in the measured error rates. The basic laboratory configuration consists of a threenode cooperative communication network implemented using the WARP SDR hardware. The cooperative schemes presented were implemented with variations of decode-andforward cooperative protocol [3]. In Section II of this paper, the operation of the three-node cooperative communication network is discussed and details of the WARP system architecture including packet structure and modulation type are provided as background. Section III describes the measurement results of a cooperative scheme based on a type of maximum ratio combining (MRC). Section IV details the measurement results of a cooperative coded scheme using hard decision decoding operating in a controlled multipath environment and section V provides the results for a cooperative coded scheme using hard decision decoding operating under the influence of Rayleigh fading.
This work is supported in part by the National Science Foundation (NSF) under awards 0722868 and 0708989, the Wireless Internet Center for Advanced Technology (WICAT), the New York State Center for Advanced Technology in Telecommunications (CATT), and Polytechnic Angel funds.

I.

INTRODUCTION

Wireless communication systems suffering from signal fading, multipath distortion and multi-user interference can improve system performance when multiple wireless nodes work cooperatively to overcome the negative channel effects. Cooperative communication is a technique where three or more active nodes operating in a common wireless network share their resources to jointly transmit messages while improving system performance through their inherent spatial diversity. Following [1]-[3], there have been numerous theoretical and simulation studies examining the potential benefits of cooperative communications. Developing and testing cooperative networks using actual radio hardware can verify the concepts that cooperative protocols can improve system performance over traditional methods using point-to-point transmission. Over the past several years, cooperative communication implementations at the medium access control (MAC) layer have been reported [5]-[8]. However, these commercial systems do not provide access to the physical layer (PHY). These MAC implementations operate on commercial radio platforms configured with open source drivers. Recently, several PHY layer implementations have also been reported operating on software defined radio (SDR) platforms such as the GNU/Universal Software Radio Peripheral (USRP) and

II. DESCRIPTION OF THE COOPERATIVE COMMUNICATION NETWORK AND OPERATION In any wireless system, it is desired to maintain a high quality link between the Source (S) and the Destination (D) nodes but practical issues such as multipath, shadowing and distance may decrease the actual performance below acceptable limits. With the help of a cooperating Relay (R) node, performance can be improved by combining signals from this second path via the relay with those transmitted from the originating source. To implement cooperative combining at the destination, one needs access to the physical layer as available in an open-source SDR platform. The implemented cooperative communication system described in this paper and shown in Figure 1 is configured

978-1-4244-4296-6/10/$25.00 2010 IEEE

5618

ICASSP 2010

Figure 1. WARP SDR system configuration for Cooperative Communication performance testing

with three separate SDR nodes connected to a common system controller. In this system, data transmission between the Source and Destination nodes occur over two unique paths where one path provides a direct connection between the Source and Destination and the second path provides a two-step connection through the Relay node. In the simplest form of cooperation, the destination would only receive from the relay provided the connection has better average end-to-end error rates. However, as demonstrated in theory [1][3] and will be exhibited via experiments in this paper, combining the signals from both paths will further enhance the overall system performance. The implemented cooperative system operates under a Time Division Multiplexing (TDM) protocol where the Source transmits data during the first time slot and the Relay transmits during the second. The data transmitted during the first time slot is received by both the Relay and Destination nodes. During the second time slot, the Relay processes the Source data and retransmits to the Destination node. There are several techniques for data retransmission at the Relay including amplify and forward and decode and forward [4]. The system described here uses variations on the decode and forward scheme. The cooperative system described in this paper uses PCbased MATLAB control of the WARP radios referred to as WARPLAB. The WARPLAB environment allows rapid prototyping and algorithm development of the cooperative system. Under WARPLAB control, raw IQ data is transferred from the radio to the PC for demodulation and bit recovery. In this case, raw data transfer occurs over a 100Base-T Ethernet connection and tends to be the bottleneck for data throughput. When using WARP, higher data throughput can be achieved when all radio functions are performed inside the FPGA but this mode of operation requires the radio programmer to have extensive knowledge of Mathworks Simulink, Xilinx System Generator and Xlinix EDK which often results in a steeper learning curve for the novice.

The WARP platform uses direct-downconversion receiver architecture. The slight differences in the RF carriers between radio nodes result in a frequency offset in the recovered baseband IQ waveforms. If the RF carriers are relatively close to one another, the selected DQPSK modulation provides a simple technique for bit recovery without the need for RF carrier synchronization. Under the control of WARPLAB, each radio operates using an internal FPGA clock rate of 40 MHz. The baseband modulated symbols are over-sampled by a factor of eight resulting in a 5MHz symbol rate. Each radio implements a transmit data buffer of 16384 complex samples, resulting in a packet size of 2048 symbols. A preamble is placed at the start of each waveform for fine timing synchronization of the received signals. The preamble is BPSK modulated and contains a Barker sequence of length 5 followed by eight symbols of no data. The transmitted preamble had an amplitude level that was +3dB higher than the data portion of the packet in order to improve the likelihood of proper synchronization to the actual start of the packet. The complex IQ waveforms are filtered with root-raised cosine functions and downloaded into the transmit buffers in preparation for signal transmission. Coarse timing synchronization is controlled by the WARPLAB program when an initial sync pulse is transmitted over the Ethernet connection from the host PC to each radio node. Once the sync is received, the respective radio begins transmission of the symbol data stored in the transmit buffers. At approximately the same time, the Destination and Relay nodes fill their respective receive data buffers. For this first time slot, the received data is then downloaded into the host PC for fine timing synchronization, demodulation and bit recovery. During the second time slot, the data from the Relay node is processed and uploaded to the WARP transmit buffers. The Relay sends the retransmission following a second sync pulse from the host PC. After receiving the signal from the Relay, the two separate transmissions are cooperatively combined. MAXIMUM RATIO COMBINING IN A DEMODULATEAND-FORWARD COOPERATIVE SYSTEM The first cooperative system measured on the testbed was implemented using a process similar to Maximum Ratio Combining (MRC) found in smart antenna systems. The measurements are made over a wireless channel with all three radios physically separated by approximately 10 feet. In this cooperative system, the Source and Relay signals are combined at the Destination based on their received signal strength. The Destination uses a measurement of the received energy in the preamble to determine the weighting of the signal addition. During the first time slot, the Source transmits a packet of uncoded data to the Relay and Destination. The Destination temporarily stores the received I and Q data from the first time slot. The Relay demodulates the Source packet, modulates again and re-transmits the same uncoded data during the second time slot. The Destination measures the Received Signal Strength (RSS) using the III.

5619

preamble from both received packets. As the system is using DQPSK modulation, the measured differential phase at each symbol is combined using a weighted average based on the measured RSS using the following equation
D ( i ) =
RSSS ( j ) + RSS R ( j ) RSS S ( j )

S ( i ) +

RSS S ( j ) + RSS R ( j )

RSS R ( j )

R ( i )

(1)

where D(i) is the calculated phase of the ith symbol in the jth packet at the Destination, S(i) is the measured phase of the ith symbol from the Source, R(i) is the measured phase of the ith symbol from the Relay, RSSS(j) is the measured signal strength of the jth packet from the Source, and RSSR(j) is the measured signal strength of the jth packet from the Relay. It is expected that the measured differential phase will be improperly estimated for a path with low received energy (low SNR), therefore, the above calculation places more weight to the path with a larger received energy (higher SNR). Once the differential phases from the two paths are combined, the composite phase angle is demodulated using a standard DQPSK demodulator. The measured BER as a function of transmit power is shown in Figure 2. Each BER measurement point contains 1250 packets representing over 5 million data bits. The BER is measured against the transmitted power level as it becomes difficult to accurately measure the received SNR. As SNR measurements would include undesired wireless interference as well as distortion from the RF and digital converter components, it is difficult to independently measure the signal and noise especially under low SNR conditions. Using transmitted power as a substitute for SNR will allow the relative slopes in the BER curves to be preserved. For this measurement, the transmitter power level was adjusted so that the output power at the Source and Relay are approximately equal. As shown in the figure, the BER for the non-cooperative direct (S-D) path has the poorest performance across the measured range of transmit power levels. The BER performance using only the Source-RelayDestination path (multi-hop) has shown a slight improvement over the direct path. The highest performance is achieved using the MRC cooperative technique. Using the weighted combination of the two paths, cooperation achieves a large improvement in measured BER when compared to the performance of either individual path. It is important to note that even if the performance of the relay path is lower than the direct path, the MRC cooperative technique would still result in an increase in system performance. IV. COOPERATIVE CODING WITH HARD DECISION DECODING UNDER CONTROLLED MULTIPATH For the second cooperative system studied, the same threenode system, described in section III above, was configured with a cooperative coded protocol similar to [4]. For cooperative coding, the Source transmits the punctured coded stream and the Relay transmits the remaining parity. As a comparison, a traditional convolutional encoder transmits data and all parity over a single channel. For our

Figure 2. Measured BER for the direct, multi-hop and MRC for demodulate-and-forward cooperative techniques

experimental set-up, the original uncoded data bits are transmitted from the Source during the first time slot. The Relay then encodes the received data according to the selected -rate channel encoder and transmits only the parity bits during the second time slot. The Destination multiplexes the two received data streams into a single stream and then decodes the combined data using the standard convolutional decoder. Note that the symbol rate was fixed in all cases and therefore the frequency and time resources remain unchanged between the two systems. For the implementation described in this paper, the channel encoder was a -rate systematic encoder with feedback having a constraint length of 5 and a generator polynomial matrix of (37 33). The measured results for the three-node cooperative coded system are shown in Figure 3. The measurements show the BER performance as a function of transmit power for the direct-uncoded case, the direct-coded case and the cooperative coded system. The direct uncoded and coded performances were measured over the single S-D path and did not make use of the Relay node. The direct-coded and cooperative-coded systems use the same convolutional encoder/decoder structure. The results shown in figure 3 use hard decision decoding and further improvements in performance have also been shown using soft decision decoding [10]. As shown in Figure 3, the cooperative coded system shows a large improvement in measured BER as compared to the direct-coded path using the same encoding/decoding scheme. It is important to note that these measurements in sections III and IV were taken in the laboratory over a wireless channel under controlled test conditions where the dynamic effects of varying multipath and other wireless interference were greatly reduced. Under these conditions, the system performance was mainly influenced by noise present in the channel and system components. In the next section, Rayleigh fading is introduced into the cooperative system to study the effects in the experimental platform.

5620

Figure 3. Measured BER performance for a direct-uncoded, directcoded and cooperative coded system with hard decision decoding

Figure 4. Measured BER performance for a direct-uncoded, direct-coded and cooperative coded system with hard decision decoding operating in Rayleigh fading PHY layer implementations of cooperative communication techniques, whose benefits have been well-established in theory and simulations. Our ongoing work includes experimentation with cooperative protocols that use multiple relays, such as distributed space-time coding protocols, as well as implementing the cooperative protocols directly into the FPGA platform.

V. COOPERATIVE CODING WITH HARD DECISION DECODING OPERATING IN RAYLEIGH FADING For the last configuration examined in this paper, Rayleigh fading was introduced into the cooperative coded system described above. There are several popular techniques to introduce a controlled amount of multipath fading in a laboratory environment including the use of commerciallyavailable channel emulators or metallic reflectors placed in the surrounding environment [11]. For the measurements presented here, fading was introduced on a packet by packet basis by varying the transmitted power level at each node. The WARP nodes have variable gain adjustments at RF and baseband that are used to modify the transmitted power following a Rayleigh distribution. To precisely control fading characteristics and eliminate undesired radio interference, the WARP nodes were cabled together with a set of RF power splitter and coaxial cables. Figure 4 shows the measured results of the three-node cooperative coded system with Rayleigh fading. As mentioned, each data point represents the measured BER over 1250 packets. In this case, the average power level is fixed for each displayed point but the packet transmit power level is varied according to a Rayleigh distribution. The S-D, S-R and R-D links utilize independent fading statistics. It is important to note that the measurements shown in figure 4 are made under different channel conditions as those in figures 2 and 3 and should not be compared in an absolute terms. As shown in the figure 4, the cooperative coded system exhibits a steeper slope in the BER curve as a result of the spatial diversity found in the cooperative network. This spatial diversity results in a larger improvement in the measured BER when compared to the direct link having the same encoder/decoder. VI. CONCLUSION
In this paper, the measured results of several cooperative communications systems were described and compared to the performance of traditional wireless systems operating over a single link using the same radio hardware. It was shown, in agreement with the theory, that the spatial diversity provided by a cooperative communications system can improve the BER performance under of wide range of test conditions. The results represent one of the few

VII. REFERENCES
A. Sendonaris, E. Erkip, and B. Aazhang, User Cooperation Diversity - Part I: System Description, IEEE Trans. on Communications, vol. 51, pp. 19271938, November 2003. [2] A. Sendonaris, E. Erkip, and B. Aazhang, User Cooperation Diversity - Part II: Implementation Aspects and Performance Analysis, IEEE Trans. on Communications, vol. 51, pp. 19391948, November 2003. [3] J. N. Laneman, D. N. C. Tse, and G. W. Wornell, "Cooperative Diversity in Wireless Networks: Efficient Protocols and Outage Behavior," IEEE Trans. Inform. Theory, vol. 50, no. 12, pp. 30623080, Dec. 2004. [4] A. Stefanov and E. Erkip, Cooperative Coding for Wireless Networks, IEEE Trans. on Communications, vol. 52, pp. 14701476, September 2004. [5] T. Korakis, S. Narayanan, A. Bagri, S. Panwar, Implementing a Cooperative MAC Protocol for Wireless LANs, Proceedings of the IEEE ICC 2006, Istanbul, Turkey, July 2006. [6] S. Singh, E. Siddiqui, T. Korakis, P. Liu, and S. Panwar, A Demonstration of Cooperative Coding Schemes using the WARP platform, in Mobicom - WinTECH, (San Francisco, CA, USA), Sept 2008. [7] A. Sharma, V. Gelara, S. Singh, T. Korakis P. Liu and Panwar, S, "Implementation of a cooperative MAC protocol using a software defined radio platform," 16th IEEE Workshop on Local and Metropolitan Area Networks (LANMAN), Sept. 2008. [8] T. Korakis, M. Knox, E. Erkip and S. Panwar, "Cooperative network implementation using open-source platforms", IEEE Communications Magazine, February 2009. [9] P. Murphy, A. Sabharwal, and B. Aazhaing, Building a Cooperative Communications System, under review, available at http://warp.rice.edu/trac/ wiki/JSAC CooperativeComm. [10] M. Knox, E. Erkip and K. Singh, "Cooperative Coding Implementation at the Physical Layer," IEEE Wireless and Optical Communications Conference (WOCC), May 2009. [11] G. Bradford and J. Laneman, "An Experimental Framework for the Evaluation of Cooperative Diversity," 43rd Annual Conference on Information Sciences and Systems (CISS), May 2009. [1]

5621

You might also like