Page 80 - Kaleidoscope Academic Conference Proceedings 2020

2020 ITU Kaleidoscope Academic Conference

2. THREE MAIN TRENDS

In this section, three main trends of future networks are
introduced. The problems of existing protocol are analyzed
under these trends.

2.1 Further Rises in MTC
The first trend is that the main participants of
communications vary from human to machines. Although
massive machine-type communications (mMTC) is included
in 5G, the developments of more massive and critical MTC Figure 1 – Key enablers of future access protocols
will still be in demand towards 2030 and beyond [1]. Unlike
human communications, the potential users can be massive, 3.1 Contention-based NOMA
and the reliability requirement can be very high. However,
the classical random access protocol requires a handshaking NOMA is a very effective technology to increase spectrum
procedure, which is not suitable for the high-efficiency efficiency when there is a near-far effect. Due to the
transmissions of massive potential users or low-latency complexity limitation, a non-orthogonal power domain has
transmissions of high-mobility networks. not been fully utilized especially for the uplink in current
protocols. NOMA still acts as an important role in the future,
2.2 Uplink-dominated System as the complexity limitation is expected to be solved by
advanced algorithms and powerful hardware in the future.
The second trend is that the overall performance is becoming
dominated by uplink instead of downlink transmissions. For It is complex and inefficient to implement accurate power
many MTC applications, uplink is the main bottleneck [9]. control and resource allocation when there are massive users
Moreover, in massive multiple input multiple output (MIMO) or the latency requirement is high. Therefore, a scheme
systems, time division duplex (TDD) is much easier to allowing users to transmit freely is in high demand. To
realize; it uses uplink-downlink reciprocity to obtain acquire this convenience for end devices, the transmission
downlink channel information. Direct downlink channel itself is inevitably to be non-orthogonal. In this case, extra
estimation is very inefficient as the overheads of downlink sensing and local power control can be required to utilize the
pilots increase with the antenna number of the base station power domain [12], but it is not suitable for low-cost and
(BS) [10]. Therefore, the pilot in the uplink becomes very low-power devices. To support a more flexible transmission
important to obtaining the channel information and not relying on any sensing and power control, a joint use of
effectively using the capability of massive antennas of the power domain, code domain and spatial domain should be
BS. In the current protocol, the uplink pilot, or demodulation considered as in Figure 1(a).
reference signal (DMRS), is orthogonal among users. It
limits the number of pilots and is not suitable for contention- 3.2 Data Features
based transmission.
To get rid of pilots or reduce the pilot overheads, data
2.3 Decentralized Structure features should be used. There are two mainstream ways to
realize this. One is to utilize the prior knowledge and
The last trend is from centralization to decentralization. This statistical information of data [5] [13], e.g. the constellation
trend has many aspects, including: (1) The distributed shape, correlation matrix, constant modulus, etc. This
antenna or cell-free design for ubiquitous connectivity, (2) method is compatible with existing protocols, and the
device-to-device transmission not relying on the central modification to existing standards is relatively small. The
controller, and (3) decentralized information management other is a data-driven method which uses deep learning (DL).
and control, e.g. blockchain [11]. Although network The end-to-end auto-encoder is one important application of
centralization brought us many good aspects like easy DL in the physical (PHY) layer, and [14] shows that pilot-
management and global control, the costs should not be free transmission can also be realized by an auto-encoder.
neglected, including coverage, latency and privacy risk.
These costs greatly limit the performance and credibility of During the exploitation of data features, novel waveform
the networks. potentially arises, [15], [16]. Discrete Fourier transform
spreading orthogonal frequency division multiplexing (DFT-
3. NOVEL ACCESS TECHNOLOGIES s-OFDM) plays an important role in 5G for its low peak to
average power ratio (PAPR). However, it makes the data
This section analyzes four novel access technologies shown feature become hard to use. One solution is real Fourier
in Figure 1. The basic idea and advantages of them are related transform spreading OFDM (RFRT-s-OFDM) [15].
discussed. As shown in Figure 1 (b), this novel waveform maintain the
data feature. Channel equalization and time/frequency offset
correction can be done using the data features in RFRT-s-

– 22 –