2021 Fiscal Year Research-status Report
Capacity-Achieving Encoding, Detection and Decoding in Generalized Multiuser Multiple-Input-Multiple-Output Wireless Communication Beyond 5G
| Project/Area Number |
21K14156
|
|---|
| Research Institution | Japan Advanced Institute of Science and Technology |
|---|
Principal Investigator |
LIU Lei 北陸先端科学技術大学院大学, 先端科学技術研究科, 助教 (80870906)
|
|---|
| Project Period (FY) |
2021-04-01 – 2025-03-31
|
| Keywords | Capacity optimal / Coded linear systems / Memory AMP / Low complexity / Encoding / Iterative detection / Decoding / Compressed sensing |
|---|
| Outline of Annual Research Achievements |
We published the first paper on low-complexity and capacity-achieving encoding and decoding for discrete massive MIMO in IEEE TIT (see [1]), which was evaluated as a potential solution for 6G communications. Furthermore, we made a major breakthrough in the matrix limitation of AMP and high-complexity limitation of orthogonal/vector AMP, firstly built a universal MAMP framework, which unifies the existing message-passing algorithms, and proposed a low-complexity and minimum mean square error (MMSE)-optimal MAMP algorithm (see [2]).
[1] Lei Liu et al, “Capacity optimality of AMP in coded systems,” IEEE Transactions on Information Theory, vol. 67, no. 7, 4929-4445, 2021.
[2] Lei Liu et al, “Memory approximate message passing,” IEEE International Symposium on Information Theory (ISIT), pp.1379-1384, Australia, 2021.
|
| Current Status of Research Progress |
Current Status of Research Progress
1: Research has progressed more than it was originally planned.
Reason
Besides the achievements above, we have made a great step forward toward the final target. We focus on the fundamental open problem: How to achieve the information-theoretic limit (i.e., constrained capacity) of a large linear system with a non-IID sensing (i.e., right-unitarily-invariant) matrix and non-Gaussian signaling? As a result, a capacity achieving coded OAMP is designed based on matched FEC coding [1]. Note: This work is an extension of our previous work “Capacity optimality of AMP in coded systems", which shows the capacity optimality of a coded AMP for IID sensing matrix. For non-IID sensing matrices, AMP does not work well. This work is also an extension of our previous work "Capacity-achieving MIMO-NOMA: Iterative LMMSE detection", which shows the Gaussian capacity-achieving of Turbo LMMSE with Gaussian signaling. However, for practical non-Gaussian signaling, Turbo is not capacity optimal anymore!
[1] L. Liu, S. Liang, and L. Ping, "Capacity optimality of OAMP: Beyond IID sensing matrices and Gaussian signaling," arXiv preprint: arXiv:2108.08503, Aug. 2021.
|
| Strategy for Future Research Activity |
Most existing works on capacity achieving OAMP are for point-to-point channels. At present, it is still a lack of rigorous analysis on the information-theoretical limits of OAMP/VAMP for generalized multi-user (GMU)-MIMO systems. Meanwhile, the characterization of the capacity region in GMU-MIMO is an intractable problem, considering the distinct rate requirements of different users. Another key challenge is to design the framework of a practical transceiver that can achieve the optimal sum capacity of GMU-MIMO and meet the different users’ rate requirements, especially for a large number of users. Therefore, this motivates us to design a practical framework based on OAMP/VAMP to achieve the sum capacity of generalized GMU-MIMO.
To address the above challenges in GMU-MIMO, we will propose a unified framework to accurately characterize the capacity and design a capacity optimal transceiver of GMU- MIMO, jointly considering encoding, modulation, detection, and decoding. It is complexity prohibited to design transceivers for a completely asymmetrical GMU-MIMO that all users may have different rates. Therefore, group asymmetry is developed to make a good tradeoff between implementation complexity and rate allocation.
|
| Causes of Carryover |
The difference will be used to pay the "Article Costs" in the next fiscal year.
|
