An MMSE Strategy at Relays With Partial CSI for a Multi-Layer Relay Network
Summary (4 min read)
Introduction
- They showed that this creates a distributed space-time code and achieves the same diversity as that of a MIMO system at high total transmitted power.
- In some systems, the relays may be able to exchange information amongst themselves before transmission.
- Unlike [18]–[23], which adopted MMSED, the authors minimize MSE at the relays and obtain relay precoder matrices.
A. Contribution
- The contributions in this paper are: For the multi-layer relay network shown in Fig. 1, the authors propose a novel MMSER relaying strategy, which does not require forward CSI in the transmitting nodes.
- The authors obtain closed form solutions for the optimumMMSER relay precoders for both cooperative and non-cooperative relays for arbitrary number of layers of relays while also including leaked signals.
- The authors enhance theMMSED strategies (though these enhancements may not be optimal) proposed in [18], [19], and [21] to work in this multi-layer network for meaningful comparison with MMSER.
- The authors show using simulations that combining more number of leaked signals improves the performance of MMSER, which outperforms/approaches that of MMSED schemes that use global CSI.
B. Notation and Organization
- Denotes the identity matrix and is the vector of elements .
- Represents a circularly symmetric complex Gaussian random variable with real and imaginary parts having mean 0 and variance .
- Is a diagonal matrix with diagonal elements .
- Thereafter, in Section III, the MMSER strategy is presented in detail and the precoders for the relay layers are derived.
- Finally, in Section VII, the authors summarize and conclude the paper.
II. SYSTEM MODEL AND THE PROPOSED SCHEME
- The authors use and to represent the links and the channel coefficients respectively, with the first two subscripts denoting the transmitter and the next two the receiver.
- Let be the set of all links and be the set of links from to .
- In the MMSER- strategy, S transmits in phase 0, and receive and store for later use; transmits in phase 1, and receive and store for later use and so on till phase , when D receives from .
- Let us denote the transmitted, received, and noise signals by , and respectively with subscript and superscript on them denoting layer and phase respectively.
- The cost function given in (2) is motivated by the following facts: Somewhat similar to the principle behind regenerative relaying (DF), the relays are desired to transmit a signal that is close to the source symbol or its scaled version.
III. MMSER PRECODER
- Khajehnouri and Sayed [18] minimized the MSE at the destination and found the precoder matrix for when with no power constraint.
- Krishna et al. [19] derived a non-diagonal precoder matrix for cooperative relays with average power constraint .
- In both (3) and (4), the authors have changed symbol notation to be consistent with this paper.
- Let us call these non-cooperative systems, which use (3) and (4), as MMSED-Khajehnouri/Krishna (MMSED-KK) and MMSED-Lee (MMSED-L) respectively.
- In their strategy, the authors minimize the MSE at the relays resulting in precoders that do not depend on forward CSI, and use leaked signals to improve the performance.
IV. EXTENSION OF MMSED SCHEMES
- Derivation of precoders for MMSED schemes, MMSED-KK and MMSED-L ((3) and (4) give the diagonal elements of the precoders of these schemes for ), is complicated when .
- The system in [23] has two layers, i.e., , and the precoders are obtained iteratively.
- The authors enhance the MMSED strategies (though these enhancements may not be optimal) proposed in [18], [19], and [21] to obtain closed-form solutions to precoders in a multi-layer network for meaningful comparison with the MMSER- system proposed in this paper.
- The authors call these systems E-MMSED, for Enhanced-MMSED systems, and find their precoders to be dependent on global CSI as shown in (17) and (18).
A. E-MMSED Strategy
- For a meaningful comparison, the authors consider the total number of phases as , and the total power transmitted as to be the same as that used by MMSER.
- S transmits times in as many phases; all the relays average their received signals and transmit times in as many phases, also known as The authors assume.
- Thus, Watts and Watts respectively, with , the total power available.
- The authors also assume that the noise is uncorrelated, i.e., where when and when is the Kronecker delta function.
- Therefore, the authors get and a diagonal element of the precoder of E-MMSED-KK as (17) 2) Precoder of E-MMSED-L: Similarly, to get the diagonal elements of the precoder of E-MMSED-L, they replace the signal power transmitted and noise variance as in E-MMSED-KK into (4) to get its diagonal element as (18).
B. Selection of
- Let us now select the best and use it while comparing its performance with MMSER- system.
- As it is hard to derive BER, the authors obtain SNR at D for any and attempt to select the value of that maximizes it.
- Substituting (16) into (19), the authors get where (20) are respectively, the signal and noise components of the received signal in phase , without considering the noise that is added at D.
- The authors do not take the noise added at D with these components, as it is not considered while deriving optimum precoder in MMSED.
- This also reflects in the BER plots shown in Section VI in Fig. 4 that for different values of , BER does not change.
VI. SIMULATION RESULTS
- The authors compare the BER performance of the proposed MMSER scheme with existing schemes and their enhanced E-MMSED schemes.
- The authors run Monte-Carlo simulations for both the cooperative and non-cooperative relays cases.
- First, the authors study the behavior of MMSER, E-MMSED-KK, and E-MMSED-L by varying the following parameters: total number of layers, , number of transmissions by S, and power allocated to S.
- This is used to analyze the performance of MMSER and MMSED schemes and select the best values of and for E-MMSED in the simulations.
A. Simulation Parameters
- The authors select from the Gray coded quadrature phase shift keying constellation with unit variance and use MMSE decoders at D, derived in Section V. Hence, the variance of the channel coefficient is given by .
- In all the simulations, whenever the authors need to increase the number of layers , they do it by inserting them between S and D keeping the distance constant.
- For , Jing and Hassibi [1] proved that Jing-Hassibi Scheme (JHS) achieves maximum SNR at D when power is equally divided between S and the relays; i.e., .
- Extending this for any , the authors use in all simulations for EJHS.
B. Summary of Results
- —Figs. 2 and 3 show respectively how the performance of MMSER-1 worsens and that of MMSER-2 improves as increases.
- This is used to select the best and for comparison with MMSER- . — Fig. 6 shows a comparison of BERs of MMSER-1 and MMSER-2 with MMSED for the single-layer case , when the relays cooperate.
- — Figs. 7 and 8 show that the BER performance of MMSER outperforms that of E-MMSED-KK and approaches that of E-MMSED-L when is increased from 1 to with and 4 respectively, when the relays do not cooperate.
C. Usefulness of Leaked Signals,
- Fig. 2 shows SNR at D and BER plots of MMSER-1 when is varied.
- It can be observed that as increases, the SNR at D decreases and BER of MMSER-1 increases.
- Fig. 3 shows BER plots of MMSER-2 for varying .
- Unlike MMSER-1, the performance of MMSER-2 improves as the number of layers increases.
- This is because MMSER-2 uses leaked signals.
D. Selection of and for E-MMSED-KK and L
- Fig. 4 shows the performance of E-MMSED-KK and E-MMSED-L, for various values of the number of transmissions of S. As was shown in (23) and (24), the BER plots also corroborate the fact that the performance does not vary with .
- Hence, the authors use in all the simulations of E-MMSED.
- Another parameter that needs to be fixed is the power allocated to S , and the relays .
- The authors find these using simulations as shown in Fig. 5, where they have used and Watt.
- For E-MMSED-KK, it can be seen that 20% of total power or achieves low BER for dB and almost same BER for dB than other power allocations.
E. Comparison: Single Layer Case
- Fig. 6 shows plots of BER in a single layer network, when relays cooperate.
- MMSED schemes use global CSI optimally while MMSER- schemes use backward CSI alone.
- When there is only one relay , MMSER-1 performs exactly same as that of MMSED.
- The performance of MMSER-2 and MMSED-K with leaked signals are also identical.
- Total power used in simulations is Watts. and MMSED-K with leaked signals perform better than the MMSER-1 and MMSED-K schemes.
F. Comparison: Multi-Layer Case
- Finally, the authors consider 3 and 4 layer systems with number of relays in each layer to be and a total transmitted power of Watt for comparing BER performance of the proposed system MMSER, with E-MMSED systems.
- Ideally, the authors would like to compare theMMSER scheme with theMMSED scheme.
- The authors MMSER- scheme performs better than E-MMSED-KK scheme even though the authors do not use forward CSI, and worse compared to the E-MMSED-L scheme.
- The comparisons for the multi-layer case in Figs. 7 and 8 also show much more gain (than in the single-layer case) using global CSI in terms of the gap between MMSER-1 and the E-MMSED-L schemes.
- Using more leaked signals reduces this gap significantly.
VII. SUMMARY AND CONCLUSION
- The authors have considered the AF relaying protocol for a multi-layer cooperative system, and proposed a precoder design method MMSER which minimizes the MSE at each relay instead of the MSE at the destination that is considered in earlier works.
- The authors believe that their MMSER schemes provide an interesting method for AF precoder design for multi-layer relay system.
- The authors found the evaluation of the achieved MSE at the destination for their MMSER schemes challenging, and thus resorted to simulation based study.
- Even when , i.e., when the unconstrained estimate has lower power, the authors amplify the estimate to use the full sum transmit power for each layer.
- Here each of the sub-matrices shown in (1) are constrained to be diagonal.
Did you find this useful? Give us your feedback
Citations
1 citations
Cites background from "An MMSE Strategy at Relays With Par..."
...These assumptions are justified since, in some systems, the relays may be able to exchange information among themselves before transmission [32]....
[...]
References
[...]
33,341 citations
7,348 citations
6,621 citations
4,385 citations
4,311 citations
Related Papers (5)
Frequently Asked Questions (12)
Q2. What future works have the authors mentioned in the paper "An mmse strategy at relays with partial csi for a multi-layer relay network" ?
Analytical performance evaluation remains a valuable future work. Also, and are given by respectively, which can be found using Table I. Now, expanding ( 6 ), the authors get ( A4 ) Substituting ( A1 ) and ( A4 ) into ( 7 ), they get ( A5 ) Now, they will derive for the case when the relays are cooperative. From ( A13 ) and ( A15 ), ( A5 ) becomes ( A16 ) Differentiating ( A16 ) w. r. t. the conjugate of the precoder matrix diagonal element, and simplifying, the authors get ( A17 ) From complementary slackness, they get or ( A18 ) Equation ( A17 ) can be written in matrix form as ( A19 ) where, and are as defined in ( 10 ), ( 11 ), and ( 12 ) respectively. APPENDIX PROOF OF CLAIM 2: SNR OF E-MMSED Using ( 20 ), in ( 21 ) can be expanded as ( A21 ) using the diagonal properties of and.
Q3. how many vectors would be received from previous transmissions?
In various phases, the layer would have received vectors each of size from previous transmissions, starting from phase till phase , where .
Q4. What is the optimum precoder for a non-cooperative case?
For the non-cooperative case, the optimum precoder , is made from the optimum vectors, , by noting that these vectors give the th diagonal elements of all submatrices,, that make up the precoder.
Q5. What is the sum transmit power constraint at each layer?
The sum transmit power constraint at each layer allows for optimal allocation of power among the relays depending on the quality of the estimate at each relay.
Q6. What is the optimum precoder matrix for MMSER?
In their strategy, the authors minimize the MSE at the relays resulting in precoders that do not depend on forward CSI, and use leaked signals to improve the performance.
Q7. How does Jing-Hassibi Scheme achieve maximum SNR?
Jing and Hassibi [1] proved that Jing-Hassibi Scheme (JHS) achieves maximum SNR at D when power is equally divided between S and the relays; i.e., .
Q8. What are the parameters used to study the behavior of MMSER?
the authors study the behavior of MMSER, E-MMSED-KK, and E-MMSED-L by varying the following parameters: total number of layers, , number of transmissions by S, and power allocated to S.
Q9. What is the power of the relays?
The authors take (3) and replace , the power transmitted by S, with and , the power transmitted by the relays, with , as the authors allocate fractions of powers to them due to their multiple transmissions.
Q10. What is the optimum precoder vector for a relay?
When the relays do not cooperate, the optimum precoder vector is given by(9)where(10)(11) and... . . ....(12)Here and represent the th diagonal elements of and respectively.
Q11. What is the information required for MMSER precoders?
B. Information Required for MMSER Precoders From (8) and (9), the authors can see that the precoders of MMSER depend on two correlation matrices and .
Q12. What is the optimum value of the decoder?
The optimum can be obtained as(27)where(28)... . . .... (29)The authors note that, prime is used over in correlation matrix to identify it to be a scalar unlike in (A2).