Blind Iterative Channel Estimation and Detection for LDPC-Coded Cooperation Under Multi-User Interference

Transcription

1 Blind Iterative Channel Estimation and Detection for LDPC-Coded Cooperation Under Multi-User Interference Don Torrieri*, Amitav Mukherjee, Hyuck M. Kwon Army Research Laboratory* University of California Irvine Wichita State University Nov. 19, 2008 Don Torrieri*, Amitav Mukherjee, Hyuck M. Blind Kwon Iterative (2008) Channel Estimation and Detection for LDPC-Coded Nov. Cooperation 19, 2008 Under 1 / 19 Mul

2 Outline 1 Background 2 System Model Coded cooperation Transmitter/Receiver 3 CSI estimation methods 4 Results 5 Conclusions Don Torrieri*, Amitav Mukherjee, Hyuck M. Blind Kwon Iterative (2008) Channel Estimation and Detection for LDPC-Coded Nov. Cooperation 19, 2008 Under 2 / 19 Mul

3 Background Cooperative diversity is achieved when two or more users jointly relay their information by means of coordinated transmissions, resulting in a simulated spatial diversity from the perspective of the destination. Coded cooperation[1] [2]: Each user divides its codeword into 2 (or greater) segments: a part of it s coded bits are transmitted by the user itself to the final destination (BS), and the remaining portion is relayed via its partner. The two users exchange roles in the next transmission interval, and so forth. Segmentation of codeword can be done by puncturing or by using product codes. Don Torrieri*, Amitav Mukherjee, Hyuck M. Blind Kwon Iterative (2008) Channel Estimation and Detection for LDPC-Coded Nov. Cooperation 19, 2008 Under 3 / 19 Mul

4 Background However, coherent demodulation and iterative decoding requires channel state information! Don Torrieri*, Amitav Mukherjee, Hyuck M. Blind Kwon Iterative (2008) Channel Estimation and Detection for LDPC-Coded Nov. Cooperation 19, 2008 Under 4 / 19 Mul

5 Background However, coherent demodulation and iterative decoding requires channel state information! Issues of channel estimation and multiple-user interference (MUI) in cooperative communications have largely been neglected. Don Torrieri*, Amitav Mukherjee, Hyuck M. Blind Kwon Iterative (2008) Channel Estimation and Detection for LDPC-Coded Nov. Cooperation 19, 2008 Under 4 / 19 Mul

6 Background However, coherent demodulation and iterative decoding requires channel state information! Issues of channel estimation and multiple-user interference (MUI) in cooperative communications have largely been neglected. Are conventional CSI estimation methods applicable to coded cooperation? Don Torrieri*, Amitav Mukherjee, Hyuck M. Blind Kwon Iterative (2008) Channel Estimation and Detection for LDPC-Coded Nov. Cooperation 19, 2008 Under 4 / 19 Mul

7 Background However, coherent demodulation and iterative decoding requires channel state information! Issues of channel estimation and multiple-user interference (MUI) in cooperative communications have largely been neglected. Are conventional CSI estimation methods applicable to coded cooperation? What is the gain in performance offered by iterative processing, if any? Don Torrieri*, Amitav Mukherjee, Hyuck M. Blind Kwon Iterative (2008) Channel Estimation and Detection for LDPC-Coded Nov. Cooperation 19, 2008 Under 4 / 19 Mul

8 System Model Proposed cooperative network example Figure: Cooperative network architecture. Don Torrieri*, Amitav Mukherjee, Hyuck M. Blind Kwon Iterative (2008) Channel Estimation and Detection for LDPC-Coded Nov. Cooperation 19, 2008 Under 5 / 19 Mul

9 System Model Coded cooperation Figure: Coded cooperation time line in a TDD setting. Partners alternate their transmissions within a cooperation interval. No feedback required between partners. Don Torrieri*, Amitav Mukherjee, Hyuck M. Blind Kwon Iterative (2008) Channel Estimation and Detection for LDPC-Coded Nov. Cooperation 19, 2008 Under 6 / 19 Mul

10 System Model Coded cooperation Channel code chosen: Rate-compatible low-density parity-check (LDPC) codes with optimized puncturing for short block lengths (N < 5000) [3]. Systematic 2-step search algorithm implemented to construct punctured code: 1 Grouping: Distribute all variable nodes into groups G 0, G 1,..., G k, where G k indicates that the punctured variable node information can be recovered in k decoder iterations. 2 Sorting: Determine the order of puncturing of variable nodes within each group. Claim: Puncturing carried out to minimize the number of decoder iterations required for recovery outperforms random puncturing. Don Torrieri*, Amitav Mukherjee, Hyuck M. Blind Kwon Iterative (2008) Channel Estimation and Detection for LDPC-Coded Nov. Cooperation 19, 2008 Under 7 / 19 Mul

11 System Model Transmitter/Receiver Figure: Transmitter and receiver of a user. Don Torrieri*, Amitav Mukherjee, Hyuck M. Blind Kwon Iterative (2008) Channel Estimation and Detection for LDPC-Coded Nov. Cooperation 19, 2008 Under 8 / 19 Mul

12 System Model Transmitter/Receiver A receiver iteration is defined as 20 LDPC decoder iterations, followed by i max = 10 EM internal iterations, and then a single demodulator metric generation. Estimate of the fading coefficient at iteration i + 1 is ĥ (j) (i+1) = 1 N t/2 y (k) x (j) N t /2 (i) (k) (1) k=1 where x (j) (i) (k) = E z y,ˆθ (j) [x(k)] = E [x k ] = E x y,ˆθ(j) xk y,ˆθ (j) [x(k)]. (i) (i) (i) Interference-plus-noise PSD I 0 estimated as Î (j) 0,(i+1) = 1 N t /2 N t/2 k=1 y(k) ĥ(j) (i+1) x (j) (i) (k) 2. (2) Don Torrieri*, Amitav Mukherjee, Hyuck M. Blind Kwon Iterative (2008) Channel Estimation and Detection for LDPC-Coded Nov. Cooperation 19, 2008 Under 9 / 19 Mul

13 System Model Transmitter/Receiver The interference with PSD N j is assumed to be active or inactive with a duty cycle d. The time-varying total interference-plus-noise PSD is modeled as { N I 0 = 0 if interfering users are inactive (3) N 0 + N j if interfering users are active Simplifying approximation: MUI is modeled as additive Gaussian noise with a bandlimited white PSD N j at the same power level as the unfaded desired signal, i.e., N j = ME s, and all interfering users are active or inactive at the same time. Don Torrieri*, Amitav Mukherjee, Hyuck M. Blind Kwon Iterative (2008) Channel Estimation and Detection for LDPC-Coded Nov. 19, Cooperation 2008 Under 10 / 19 Mul

14 CSI estimation methods Initial estimation Three methods compared for the initial CSI estimates: 1 Multiplexed Pilot-assisted Channel Estimation (M-PACE): multiplex or embed pilot symbols at known locations throughout the transmitted codeword. 2 Superimposed or Overlaid PACE (O-PACE): periodic, non-random pilot sequence superimposed on to all data code symbols within the codeword. 3 Blind (no-pilot): CSI estimated directly from received symbols using heuristic formula. No-pilot and O-PACE codewords have the same number of information symbols. M-PACE and O-PACE allocate the same total power for pilot symbols per frame. Don Torrieri*, Amitav Mukherjee, Hyuck M. Blind Kwon Iterative (2008) Channel Estimation and Detection for LDPC-Coded Nov. 19, Cooperation 2008 Under 11 / 19 Mul

15 CSI estimation methods Figure: Transmitted codeword with data and pilot power allocations for (a) M-PACE (b) O-PACE (c) proposed method without pilot symbols. Shaded regions in (a) and (b) represent the pilot symbol locations. Don Torrieri*, Amitav Mukherjee, Hyuck M. Blind Kwon Iterative (2008) Channel Estimation and Detection for LDPC-Coded Nov. 19, Cooperation 2008 Under 12 / 19 Mul

16 CSI estimation methods M-PACE: initial channel coefficients are obtained using a simple average of the received pilot symbols per fading block, then obtain Î (0) 0,(i max ) from (2) with x (0) (i max )(k) set to the known pilot symbols. O-PACE: initial fading-coefficient estimate per fading block obtained using an approximate least-squares method as ĥ (0) o,i max = N 1 t/2 Ep /E s N t /2 k=1 y (k) p (k). (4) Blind: simple heuristic method to obtain initial CSI from the received data: Î (0) ĥ(0) (i max ) = 1 N t /2 0,(i max ) (D = max N t/2 k=1 ĥ(0) (i max ) y(k) (5) 2, f ĥ(0) (i max ) 2) (6) Don Torrieri*, Amitav Mukherjee, Hyuck M. Blind Kwon Iterative (2008) Channel Estimation and Detection for LDPC-Coded Nov. 19, Cooperation 2008 Under 13 / 19 Mul

17 CSI estimation methods Two possible configurations exist for the iterative coded-cooperation system. Case-I: Additional (j max 1) receiver iterations at either partner or destination only. Case-II: Additional (j max 1) receiver iterations at both partner and destination. Case II is expected to provide the best decoding performance, albeit at the cost of increased latency (processing delay). Don Torrieri*, Amitav Mukherjee, Hyuck M. Blind Kwon Iterative (2008) Channel Estimation and Detection for LDPC-Coded Nov. 19, Cooperation 2008 Under 14 / 19 Mul

18 Results Table: Rate-Compatible LDPC Parameters Rate (Column,Row) Weights (3,6) (3,6) Length Punctures Channel model: correlated, fading-block size of F = 4 symbols, information-bit rate of 100 kb/s. No. of interfering pairs: One Don Torrieri*, Amitav Mukherjee, Hyuck M. Blind Kwon Iterative (2008) Channel Estimation and Detection for LDPC-Coded Nov. 19, Cooperation 2008 Under 15 / 19 Mul

19 Results Interference environment BER Blind methods 0.4 db from PACE at BER = 10 3 Figure: BER versus E b /N 0 for the blind and PACE methods. on Torrieri*, Amitav Mukherjee, Hyuck M. Blind Kwon Iterative (2008) Channel Estimation and Detection for LDPC-Coded Nov. 19, Cooperation 2008 Under 16 / 19 Mul

20 Throughput Results Blind case II : highest throughput (within 10% of perfect CSI) M-PACE and O-PACE : 17% lower throughput than perfect CSI Figure: Throughput versus E b /N 0 for the proposed and PACE methods. on Torrieri*, Amitav Mukherjee, Hyuck M. Blind Kwon Iterative (2008) Channel Estimation and Detection for LDPC-Coded Nov. 19, Cooperation 2008 Under 17 / 19 Mul

21 Conclusion Conclusions Blind channel-coefficient and interference-plus-noise PSD estimation method based on expectation-maximization was proposed for LDPC-coded cooperative communications. Don Torrieri*, Amitav Mukherjee, Hyuck M. Blind Kwon Iterative (2008) Channel Estimation and Detection for LDPC-Coded Nov. 19, Cooperation 2008 Under 18 / 19 Mul

22 Conclusions Conclusion Blind channel-coefficient and interference-plus-noise PSD estimation method based on expectation-maximization was proposed for LDPC-coded cooperative communications. The importance of estimating the interference-plus-noise PSD was shown by the noticeable improvement in LDPC decoding performance under multi-user interference. Don Torrieri*, Amitav Mukherjee, Hyuck M. Blind Kwon Iterative (2008) Channel Estimation and Detection for LDPC-Coded Nov. 19, Cooperation 2008 Under 18 / 19 Mul

23 Conclusions Conclusion Blind channel-coefficient and interference-plus-noise PSD estimation method based on expectation-maximization was proposed for LDPC-coded cooperative communications. The importance of estimating the interference-plus-noise PSD was shown by the noticeable improvement in LDPC decoding performance under multi-user interference. The proposed blind estimation method has a decoding performance superior to superimposed PACE and within 0.6 db of M-PACE. Don Torrieri*, Amitav Mukherjee, Hyuck M. Blind Kwon Iterative (2008) Channel Estimation and Detection for LDPC-Coded Nov. 19, Cooperation 2008 Under 18 / 19 Mul

24 Conclusions Conclusion Blind channel-coefficient and interference-plus-noise PSD estimation method based on expectation-maximization was proposed for LDPC-coded cooperative communications. The importance of estimating the interference-plus-noise PSD was shown by the noticeable improvement in LDPC decoding performance under multi-user interference. The proposed blind estimation method has a decoding performance superior to superimposed PACE and within 0.6 db of M-PACE. The proposed method provides the highest information throughput out of the three CSI estimation methods considered in this paper. Don Torrieri*, Amitav Mukherjee, Hyuck M. Blind Kwon Iterative (2008) Channel Estimation and Detection for LDPC-Coded Nov. 19, Cooperation 2008 Under 18 / 19 Mul

25 Conclusions Thank You A. Stefanov and E. Erkip, Cooperative coding for wireless networks, IEEE Trans. Commun., vol. 52, no. 9, pp , Sep T. E. Hunter and A. Nosratinia, Diversity through coded cooperation, IEEE Trans. Wireless Commun., no. 2, pp , Feb J. Ha, J. Kim, D. Klinc, and S. W. McLaughlin, Rate-compatible punctured low-density parity-check codes with short block lengths, IEEE Trans. Inform. Theory, vol. 52, pp , Feb D. Torrieri, A. Mukherjee, and H. M. Kwon, Doubly iterative LDPC-coded DS-CDMA receivers with coherent detection, EM channel estimation, and no pilot symbols, to appear, IEEE MILCOM, P. Hoeher and F. Tufvesson, Channel estimation with superimposed pilot sequence, in Proc. IEEE GLOBECOM, pp , Don Torrieri*, Amitav Mukherjee, Hyuck M. Blind Kwon Iterative (2008) Channel Estimation and Detection for LDPC-Coded Nov. 19, Cooperation 2008 Under 19 / 19 Mul