Page 26 - ITU KALEIDOSCOPE, ATLANTA 2019
P. 26

2019 ITU Kaleidoscope Academic Conference




             18
                                                              ergodic capacity of FD-NOMA can be given as [7].
             16
                                                                                  M   N n+1 t+1
                                                                                                (  )
                                                                                  X X X X        1
                                                                       C ≈ π log e             e  ¯ γ i,j  ×
             14                                                                2
                                                                                  i=1 j=1 k=1 s=1          (2)
            System capacity (Bit/s/Hz)  10 8                  where M, N denote the transmitter number and
                                                                                       1
             12
                                                                                         )
                                                                           p
                                                                             b k a s e
                                                                                          ,
                                                                                 (−b k b s ¯γ i,j
                                                                        a k
                                                                                respectively.
                                                              receiver number,
                                                                                               We denote α i,k
                                                              as the FD self-interference from transmitter i to
              6
                                                              receiver j.
                                                                     ρα i,j  Moreover, ¯γ i,j can be given by ¯γ i,j =
                                                                                , with α i,j , ρ, η, α i,k the NOMA
              4                                               ρ( P N  α i,l +ηα i,k )+1
                                                                 l=i+1
                                                              power coefficient, signal to noise ratio (SNR), FD
              2                                               coefficient and its corresponding FD self-interference.
                                                              With regard to a k , b k , a s , b s , we have a k =  θ k −θ k−1 , b k =
              0                                                                                      π
              5          10         15         20        25   cot θ k−1 −cot θ k  , a s =  θ s −θ s−1 , b s =  cot θ s−1 −cot θ s , and 0 ≤
                                                                   π              π             π
                                 SNR (dB)                                  π                 π
                                                              θ 1 < ...θ k < ... , 0 ≤ θ 1 < ...θ s < ... . On the contrary,
                                                                           2                 2
           Figure 6 – System capacity comparison between      the half-duplex (HD) based NOMA capacity expression
           HD-OMA, HD-NOMA, FD-OMA, and FD-NOMA.              can be obtained while removing the FD self-interference
                                                              part. In literature, capacity expressions for HD-OMA
           and successive interference cancellation (SIC) for  and FD-OMA have been investigated a lot, we omitted
           simultaneously encoding and decoding multiple user’s  the derivations here.
           information[7]. In NOMA studies, we generally assumed  In this comparison, we paired 3 users of the NOMA
           that SIC can perfectly eliminate the interference from  scheme with normalized channel noise. The allocated
           other users within the same resource block that are  NOMA powers are [4, 2, 1]. We average the allocated
           with inferior channel conditions.  This is an ideal  power for the OMA scheme, i.e., the power allocation
           assumption, which is almost impossible. Moreover, SIC  for OMA user is  4+2+1  = 3.5.  The comparison
                                                                                  2
           is a time-consuming and of great complexity method,  results are given as in Fig.  6.  We can find that
           which might even be beyond the processing ability of  combining the FD and NOMA always has better
           current electronic devices. In future studies, some novel  capacity performance compared to the half duplex (HD)
           encoding methods besides SIC might be needed.      and OMA schemes. However, due to self-interference,
                                                              the merit of FD-NOMA is reduced.
           Compared to massive MIMO and NOMA, FD enables
           synchronous transmission and reception.  FD offers  4.2  Redesigning the networking architecture
           even lower latency and better capacity performance      technologies
           [7]. The weakness of FD lies in the self-interference
           [14] generated by its own transmitter.  In order to  The current driving force of wireless evolution is
           deal with this problem, we may employ interference  from the data-centric with the aim to connect more
           cancellation devices at the receiver side.  This is  devices and provide even faster transmission speeds
           not a simple job because digital-domain cancellation  for the devices.  As we know, massive connected
           can be successfully implemented only when up to its  devices and their faster transmission speeds bring in
           effective dynamic range of the analog-to-digital (ADC)  traffic overload, especially to wireless networks such
           (suppose the FD terminal uses a B-bit ADC, the     as base stations (BS). Nevertheless, users do not
           range is about 6.02(ENOB − 2) dB [15]. Additionally,  care about how and where the data comes from
           this cancellation implementation generally has multiple  but only the quality of service (QoS) and quality
           stages.  Due to the consumed processing time of    of experience (QoE) of its service.  In order to
           these multiple stages, latency is increased. In future,  solve this problem while catering to the driving
           self-interference cancellation algorithms and some fast  force’s shift from data-centric to information-centric,
           processing devices can be some good topics for the FD  the information-centric networking (ICN) and edge
           studies. Besides, combining these 5G NR technologies  computing technologies receive increasing attention [16,
           can further improve performance, which is another  17, 18, 19].
           interesting topic for future studies.              In Fig.   7, we compare the network throughput
                                                              between  the  edge  computing-assisted  ICN  and
           We   compare  the  capacity  performance  between  conventional  network  TCP    technologies  while
           half  duplex-NOMA     (HD-NOMA),      FD-NOMA,     increasing the subscriber numbers with per subscriber’s
           HD-orthogonal  multiple  access  (HD-OMA)    and   transmission rate 10 MBit/s.  As depicted by this
           FD-OMA by considering the Relay channel model.     figure,  with subscriber numbers growing,   curves
           According to previous work, the achievable system  of edge computing-assisted ICN remain constant




                                                           – 6 –
   21   22   23   24   25   26   27   28   29   30   31