1. Tsinghua National Laboratory for Information Science and Technology,
Spectrum-Efficient Superimposed Pilot Design Based on Structured Compressive Sensing for Downlink Large-Scale MIMO Systems
Linglong Dai, Zhen Gao, Zhaocheng Wang, and Zhixing Yang
Tsinghua National Laboratory for Information Science and Technology,
Department of Electronic Engineering, Tsinghua University, Beijing , China

2. Contents 1 Technical Background 2 Proposed Solution 3
Simulation Results 4
Conclusions

3. 5G and its key technologies
Key requirement of 5G: ~1000 folds of increase in data traffic
Three main directions

4. Technical background of 5G - Large-scale MIMO
What is large-scale MIMO and why ?
Use hundreds of antennas at the BS to simultaneously serve a set of users
Increase the spectral and power efficiency by orders of magnitude
Conventional MIMO
M:2~8, K:1~4 (LTE-A)
Massive MIMO
M: ~100~1000, K: 16~64
T. L. Marzetta, “Non-cooperative Cellular Wireless with Unlimited Numbers of Base Station Antennas,” IEEE
Transactions on Wireless Communications, vol. 9, no. 11, pp , Nov (2013 IEEE Marconi Prize)

5. Technical background of 5G - Large-scale MIMO
Feb 2012, Rice university & Bell labs, Argos, 64 antennas, 15 users, 85 bit/s/Hz, 1/64 power consumption
Sep 2013, Rice university & Bell labs, ArgosV2, 96 antennas, 32 users
July 2013, Linköping & Lund University, 128 antennas, 36 users

6. Challenges for large-scale MIMO
Prospective
Considered as the promising key technology for 5G [Vargas’13]
Challenges [Rusek’13]
Channel measurement and modeling
Design of large antenna arrays
Pilot pattern design under pilot contamination
Channel feedback
Low-complexity precoding/signal detection

7. Orthogonal pilots in conventional MIMO systems
What will happen if Tx=128
Orthogonal pilots (Tx=2) for LTE/LTE-A
New
theory
Different channels are distinguished by orthogonal pilots 7

8. Compressive sensing (CS)
Sparse signals with high dimension can be perfectly reconstructed from measurements with low dimension by solving an optimization problem

9. Application of compressive sensing
A new fundamental theory against the conventional Shannon-Nyquist theory
Successful application in many fields
Single pixel camera(Rice University)
MRI (MIT)
Hyper spectral imager (Yale University)
DNA array sensor (UIUC)
CS for wireless networks
Sparse channel estimation
positioning
Multiple access
Cognitive radios 9

10. Application of compressive sensing
Example of CS application CS
Image separation
Image is sparse in a certain domain

11. Contents 1 Technical Background 2 Proposed Solution 3
Simulation Results 4
Conclusions

12. Spatial-temporal correlation of MIMO channels
Common Sparse Support
Common sparsity holds for different CIRs for adjacent transmit antennas in several adjacent time slots

13. Non-orthogonal pilots at the transmitter
Conventional orthogonal pilots：
Proposed superimposed pilots (non-orthogonal)：

14. Simultaneous channel reconstruction at the receiver
Received pilots
Channel is sparse !

15. Simultaneous channel reconstruction at the receiver
Structured CS exploits spatial-temporal correlation of MIMO channels to improve the channel estimation accuracy and reduce the pilot overhead
Adjacent antennas
Adjacent time slots

16. Proposed structured SP (SSP) algorithm

17. Contents 1 Technical Background 2 Proposed Solution 3
Simulation Results 4
Conclusions

18. Simulation Results Simulation setup Transmit antennas: N=64
OFDM size: 4096
CP length: 256
System bandwidth: 10MHz
Channel model: ITU-VB
Pilot number: 800
Overall pilot overhead: 19.53%
平均每发射天线需12.5导频！平均仅占0.31%！

19. Simulation Results
Pilot number per antenna: 12.5! Only 0.31%!

20. Contents 1 Technical Background 2 Proposed Solution 3
Simulation Results 4
Conclusions

21. Conclusions
This paper focuses on the downlink training and channel estimation for large-scale MIMO systems.
In contrast to standardized orthogonal pilots with the prohibitive overhead increasing with the number of transmit antennas, the proposed superimposed pilot design based on structured CS can efficiently solve the pilot overhead problem.
At the receiver, the proposed SSP algorithm can exploit the spatial and temporal correlations of large-scale MIMO channels for simultaneous recovery of multiple channels.
Moreover, the proposed superimposed pilot design and the corresponding channel estimator can be applied in the uplink too, and conventional small-scale MIMO can also adopt the proposed scheme to reduce the pilot overhead and improve the channel estimation performance.