Professional Documents
Culture Documents
Near-Field MIMO Communications For 6G Fundamentals Challenges Potentials and Future Directions
Near-Field MIMO Communications For 6G Fundamentals Challenges Potentials and Future Directions
Near-Field MIMO Communications For 6G Fundamentals Challenges Potentials and Future Directions
Authorized licensed use limited to: Kookmin University. Downloaded on July 20,2023 at 01:19:32 UTC from IEEE Xplore. Restrictions apply.
In this article, we systematically investigate the Near-field Far-filed
recent near-field communication techniques for
6G. The key features of this article can be summa- Spherical wave Planar wave
rized as follows:
• To begin with, the fundamental differences Rayleigh
between far-field and near-field communica- distance
⋯
tions are explained. Comparatively speaking,
the planar wavefront in the far-field can steer
the signal energy toward a specific physical
angle. On the contrary, the near-field spherical
wavefront achieves energy focusing on both
angle and distance. Moreover, the Rayleigh
y-axis [m]
y-axis [m]
distance that quantifies the boundary between
far-field and near-field regions is introduced,
and its derivation is explained in detail. Based
on this derivation, we further extend the clas-
sical Rayleigh distance, for MIMO channels
with a direct base station (BS)-user equipment x-axis [m] x-axis [m]
(UE) link, to the one for RIS-aided communi-
cations, where a cascaded channel is utilized FIGURE 1. Far-field planar wavefront vs. near-field spherical wavefront. The plots at the bottom illustrate the normal-
for presenting the BS-RIS-UE link. ized received signal energy in the physical space achieved by near-field beamfocusing (bottom left) and far-field
• Additionally, we investigate the emerging near- beamsteering (bottom right).
field communication techniques by classifying
them into two types: techniques addressing as planar waves. In contrast, when the BS-UE dis-
the challenges and those exploiting the poten- tance is shorter than the Rayleigh distance, the UE
tials in near-field regions. On one hand, as is located in the near-field region of the BS. In this
most techniques specific to far-field often suf- region, EM waves impinging on the BS array must
fer from severe performance loss in the near- be accurately modeled as spherical waves [5].
field area, the first type of techniques aims to More precisely, the planar wave is a long-dis-
compensate for this loss, such as near-field tance approximation of the spherical wave. In
channel estimation and beamforming. On far-field regions, the phase of EM waves can be
the other hand, the second kind of study has elegantly approximated by a linear function of the
revealed that the nature of near-field spherical antenna index through Taylor expansion. This con-
wavefront can also be exploited to provide cise linear phase forms a planar wavefront only
new possibilities for capacity improvement. related to an incident angle. Accordingly, by the
The principles, recent progress, pros, and utilization of planar wavefronts, far-field beamform-
cons of these two categories of research are ing can steer the beam energy toward a specific
discussed in detail. angle over different distances, which is also called
• Finally, several open problems and future beamsteering, as shown in the bottom right of
research directions for near-field communi- Fig. 1. Unfortunately, this concise linear phase fails
cations are pointed out. For example, the to thoroughly reveal the information of spherical
improvement of Rayleigh distance consid- waves. In near-field regions, the phase of spherical
ering various communication metrics need waves should be accurately derived based on the
to be analyzed, artificial intelligence (AI) is physical geometry, which is a nonlinear function of
expected to enable high-performance near- the antenna index. The information of the incident
field transmissions with low complexity, and angle and distance in each path between BS and
hybrid far- and near-field communications UE is embedded in this nonlinear phase. Exploiting
also require in-depth study. the extra distance information of spherical wave-
fronts, near-field beamforming is able to focus the
FundAmentAls oF neAr-FIeld communIcAtIons beam energy on a specific location, where energy
In this section, we first present the differences focusing on both the angle and distance domain is
between far-field and near-field communications. achievable, as shown in the bottom left of Fig. 1.
Then we identify the principle to determine the Due to this property, beamforming in the near-field
boundary between the far-field and near-field is also called beam focusing.
regions in several typical application scenarios. The differences between far-field planar wave-
fronts and near-field spherical wavefronts bring
FAr-FIeld communIcAtIons vs. neAr-FIeld communIcAtIons several challenges and potentials to wireless commu-
The critical characteristics of far-field and near-field nications, which are detailed in the following sections.
communications are shown in Fig. 1. We consid-
er an uplink communication scenario, while the rAyleIgh dIstAnce
discussions in this article are also valid for down- The most crucial premise for near-field communi-
link scenarios. The BS is equipped with an ELAA. cations is quantifying the boundary between the
A widely adopted metric to determine the bound- far-field and near-field regions (i.e., the Rayleigh
ary between far-field and near-field regions is distance). Generally, the classical Rayleigh dis-
the Rayleigh distance, also called the Fraunhofer tance is proportional to the square of the array
distance [4]. When the communication distance aperture and the inverse of the wavelength. Its
between the BS and UE (BS-UE distance) is larger derivation can be summarized as follows [4].
than the Rayleigh distance, the UE is located in the The true phase of the EM wave impinging on a
far-field region of the BS. Then EM waves imping- BS antenna has to be calculated based on the
ing on the BS array can be approximately modeled accurate spherical wave model. In far-field sce-
Authorized licensed use limited to: Kookmin University. Downloaded on July 20,2023 at 01:19:32 UTC from IEEE Xplore. Restrictions apply.
The true phase of the EM SIMO/MISO MIMO RIS Aperture 𝐷𝐷
Wavelength 𝜆𝜆 Wavelength 𝜆𝜆 Wavelength 𝜆𝜆
wave impinging on a BS
antenna has to be calculat-
Communication
ed based on the accurate scenarios 𝑟𝑟
spherical wave model. In far- RIS
BS BS UE
field scenarios, this phase is UE
BS UE
Authorized licensed use limited to: Kookmin University. Downloaded on July 20,2023 at 01:19:32 UTC from IEEE Xplore. Restrictions apply.
Therefore, we can conceive the intuition that the
grids should be sparse far away from the ELAA The BS ELAA is first parti-
but dense near the ELAA. Without considering Grid aligned with location 1 tioned into multiple sub-ar-
this intuition, the codebook in [6] is hard to use Grid aligned with location 2
to attain satisfactory channel estimation perfor-
rays. The UE is assumed to
mance in the entire near-field region. To this end, be located in the far-field
y-axis [m]
by minimizing the largest coherence among code- area of each small sub-array
words in the near-field codebook, the authors in
[5] mathematically prove this intuition (i.e., the but within the near-field
angle space could be uniformly divided, while the range of the ELAA. Then one
distance space should be non-uniformly divided).
As shown in Fig. 3, the shorter the distance, the
TTD line is inserted between
denser the grid. With the help of this non-uniform each sub-array and the
codebook, a polar-domain sparse channel repre- radio frequency (RF) chain
sentation and corresponding CS-based algorithms x-axis [m]
are proposed in [5] to accomplish accurate chan- to realize frequency-depen-
nel estimation in both near- and far-field areas. FIGURE 3. Near-field codebook with non-uniform grids. dent phase shifts.
Near-Field Beam Split
Challenge: In THz wideband systems, ELAA TTD line is inserted between each sub-array and
might encounter a beam split phenomenon, also the radio frequency (RF) chain to realize frequen-
known as beam squint and spatial-wideband cy-dependent phase shifts. Finally, the frequen-
effect. Existing THz beamforming architecture cy-dependent phase variations across different
often employs analog phase shifters (PSs) [7], sub-arrays induced by spherical wavefronts are
which usually tune the same phase shift for signals compensated by the inserted TTD line. As a result,
at different frequencies. Nonetheless, the actual beams over the working band are focused at the
phase of the EM wave is the product of the signal target UE location [9].
propagation delay and the frequency-dependent In conclusion, the first solution [8] follows
wave number. As a result, the signal propagation the PS-based beamforming, which is easy to
delay can be compensated through a phase shift implement, but the achievable performance is
adequately only for a narrowband signal. Phase unsatisfactory. The second scheme [9] can near-
errors are introduced for the other frequencies, ly eliminate the near-field beam split effect but
thus causing the beam split effect. In fact, the requires the implementation of TTD lines. In fact,
impact of beam split on far-field and near-field although deploying TTD lines by optical fibers
propagations also differs. has been demonstrated in the optical domain,
In far-field, beam split leads to the fact that this kind of deployment is nontrivial to be extend-
beams at different frequencies are transmitting ed to THz ELAA communications. Fortunately,
toward different angles, as shown in the left of recent advances in graphene-based plasmonic
Fig. 4. For near-field beam split, however, beams waveguides provide low-cost solutions for imple-
are focused at both different angles and various menting TTD lines at high frequencies [7].
distances due to the split of spherical waves, as
shown in the right of Fig.4. Both far-field and near- Potentials for Near-Field Communications
field beam splits severely reduce the received sig- Unlike the aforementioned works for dealing with
nal energy of frequency components misaligned the performance degradation in the near-field,
with the user location. Over the years, extensive some recent studies have surprisingly revealed
works have been proposed to mitigate far-field that 6G networks can also benefit from near-field
beam split by tuning frequency-dependent phase propagation. In this section, we discuss those
shifts with planar wavefronts through true-time- studies exploiting the potentials of near-field prop-
delay-based (TTD-based) beamforming instead of agation to improve communication performance.
PS-based beamforming. Unfortunately, due to the
discrepancy between planar and spherical waves, Single-User Capacity Improvement
these schemes addressing the far-field beam split Potential: The spatial multiplexing gain of MIMO
no longer work well in the near-field, posing chal- communications considerably increases with the
lenges to THz ELAA communications. transition from far-field regions to near-field regions.
Recent progress: Recently, a few efforts have In far-field MIMO communications, the line-of-
tried to overcome the near-field beam split effect. sight (LoS) channel can be represented by a rank-
In [8], a variant of chirp sequence is utilized to one matrix, where the spatial degrees of freedom
design the phase shifts for flattening the beam (DoFs) are very limited. In contrast, the near-field
focusing gain across frequencies with the sacrifice LoS channel can be rank-sufficient derived from the
of the maximum beam focusing gain. This meth- geometric relationship under the spherical propa-
od can slightly alleviate the near-field beam split gation model. The increased rank indicates dramati-
effect, but its spectral efficiency degrades as well cally improved spatial DoFs in the near-field region.
when the bandwidth is very large, as the beams Precisely, based on the expansion of prolate sphe-
are still generated by PSs. To this end, a phase-de- roidal wave functions, it is proved in [10] that near-
lay focusing (PDF) method is proposed in [9] field spatial DoFs are proportional to the product
exploiting TTD-based beamforming. To further of the BS and UE array apertures and inversely pro-
illustrate, the BS ELAA is first partitioned into mul- portional to the BS-UE distance. This conclusion is
tiple sub-arrays. The UE is assumed to be located further improved in [11] by meticulously designing
in the far-field area of each small sub-array but the beamfocusing vectors of the BS and UE arrays.
within the near-field range of the ELAA. Then one As shown in Fig. 5, the DoFs increase from 1 to 20
Authorized licensed use limited to: Kookmin University. Downloaded on July 20,2023 at 01:19:32 UTC from IEEE Xplore. Restrictions apply.
Fortunately, near-field beam
focusing enjoys the capa-
bility of energy focusing
on the joint angle-distance
domain. Hence, near-field
SDMA could generate beams
with spherical wavefronts to
simultaneously serve users
located at similar angles but
different distances.
FIGURE 4. This figure illustrates the far-field beam split effect (left) and the near-field beam split effect (right). Far-field beam split makes beams at different
frequencies transmit toward different directions, while near-field beam split makes beams at different frequencies be focused on various locations.
when the BS-UE distance decreases from 350 m multI-user cApAcIty Improvement
to 10 m. Thanks to the increased DoFs, the near-
field LoS path enables simultaneous transmission Potential: Near-field propagation is also able
of multiple data streams by MIMO precoding, as to improve capacity in multi-user (MU) com-
opposed to the rank-one far-field LoS channel sup- munications. To increase the spectral efficiency
porting only one data stream. The increased spa- in MU-MIMO communications, space-division
tial DoFs can be exploited as an additional spatial multiple access (SDMA) is widely considered to
multiplexing gain, which offers a new possibility for distinguish users through orthogonal or near-or-
significant capacity enhancement. thogonal spatial beams. Thus, multiple users can
Recently, some novel precoding architectures share the same time and frequency resources. For
have been proposed to leverage these extra near- far-field SDMA, utilizing beamsteering to gener-
field DoFs for MIMO capacity enhancement [12, ate beams with planar wavefronts can distinguish
13]. First, distance-aware precoding (DAP) was users at different angles. A downside is that users
developed in [12]. Unlike classical hybrid precod- located at similar angles will severely interfere
ing with a fixed and limited number of RF chains with each other, and thus cannot simultaneous-
(e.g., 2 or 4 fixed RF chains), the DAP archi- ly access the network through far-field SDMA.
tecture could flexibly adjust the number of RF Fortunately, near-field beam focusing enjoys
chains to match the distance-related DoFs, which the capability of energy focusing on the joint
is achieved by deploying a selection network to angle-distance domain. Hence, near-field SDMA
configure each RF chain as active or inactive. For could generate beams with spherical wavefronts
instance, in the far-field region, only one RF chain to simultaneously serve users located at similar
is activated for data transmission. When communi- angles but different distances, as shown in Fig. 6.
cation distance is decreasing to 10–20 m, around The distance information of spherical wavefronts
20 activated RF chains are enough to adapt to the supplies a new utilizable dimension for multi-user
DoFs, as shown in Fig. 5. By doing so, the number access; thus, near-field SDMA can also be regard-
of transmitted data streams dynamically match- ed as location-division multiple access (LDMA).
es the DoFs. Simulations demonstrate the DAP Recent progress: Taking advantage of the capa-
could significantly increase the spectral efficien- bility of beamfocusing, the authors in [14] studied
cy, while its energy efficiency is comparable with the near-field multi-user transmission consider-
hybrid precoding. To avoid the utilization of extra ing fully digital precoding, hybrid precoding, and
RF chains, another effort to harvest the potential transmissive reconfigurable metasurface (RMS).
spatial multiplexing gain in near-field areas is the By optimizing the sum rate in multi-user systems
widely spaced multi-subarray (WSMS) precoding through alternating optimization, all considered
[13]. In this architecture, the sub-arrays are widely precoding architectures can naturally generate
spaced to enlarge the array aperture, artificially beams with spherical wavefronts to distinguish
creating the expansion of the near-field region. users located at similar angles but different dis-
Compared to classical hybrid precoding, the num- tances. The simulation results demonstrate that
ber of sub-arrays and the sub-array spacing should near-field propagation has the potential of
be additionally designed in the WSMS architec- enhancing multi-user capacity.
ture. To this end, [13] first assumed planar-wave
propagation within each sub-array and spher- Future reseArch dIrectIons
ical-wave propagation across different sub-ar- In this section, several future research directions
rays similar to [9]. Then [13] jointly optimized for near-field communications are pointed out.
the number of sub-arrays, their spacing, and the
precoding matrix for maximizing the achievable neAr-FIeld communIcAtIon theory
rate. Simulations demonstrate that WSMS could Improvement of Rayleigh Distance: As a widely
achieve nearly 200 percent higher spectral effi- adopted quantification of near-field range, Rayleigh
ciency than classical hybrid precoding. distance is attained in terms of phase discrepan-
Authorized licensed use limited to: Kookmin University. Downloaded on July 20,2023 at 01:19:32 UTC from IEEE Xplore. Restrictions apply.
cy. For communication metrics directly affected 103
by phase discrepancy, such as channel estimation Except for near-field propa-
accuracy, Rayleigh distance can accurately cap- gation, the spatial non-sta-
ture the degradation of these metrics when apply- 102 tionarity effect is another
ing far-field transmission schemes in the near-field
Degrees of Freedom
region. On the contrary, some metrics are directly fundamental characteristic
influenced by other factors instead of phase dis- 101 of ELAA compared to 5G
crepancy; for example, capacity is determined by
beamforming gain and channel rank. Accordingly, massive MIMO, where differ-
classical Rayleigh distance probably cannot cap- ent scatterers and users are
100
ture the performance loss of these metrics well. visible to different portions
To this end, several recent works have endeavored
to improve classical Rayleigh distance in terms of of the ELAA. This effect leads
some vital communication metrics. For instance,
10-1
0 50 100 150 200 250 300 350 400 450 500 to the fact that only a part
an effective Rayleigh distance (ERD) is derived in Distance between BS and UE [m]
[9] for the accurate description of beamforming of the ELAA can receive the
FIGURE 5. The spatial DoF increases in the near-field region. (Simulation
gain loss and capacity loss. Nevertheless, ERD is spherical EM waves radiated
codes can be found at http://oa.ee.tsinghua.edu.cn/dailinglong/publi-
only valid for MISO communications, while more by a scatterer or a user.
cations/publications.html.)
discussion should be made to improve Rayleigh
distance in more practical scenarios under more
general metrics (e.g., channel rank and energy effi- field) communication scenario. Additionally, it is
ciency in MIMO and RIS systems). worth mentioning that the Rayleigh distance is
proportional to frequency. Thus, in an ultra-wide-
Near-Field Transmission Technologies band or frequency-hopping system with a very
AI-Aided Near-Field Communications: Differ- large frequency span, its near-field range varies
ent from far-field communications, the transmis- dramatically across the bandwidth. Chances are
sion algorithms for near-field communications are that the signal components at low frequencies
more complex. To be specific, since extra grids on may operate in far-field regions, while those at
the distance domain are required, as mentioned high frequencies with larger Rayleigh distances
earlier [5], the size of near-field codebooks is are propagating in the near-field areas, which also
usually much larger than that of far-field code- contributes to hybrid-field communications. Con-
books, leading to high-complexity channel estima- sequently, the above factors make hybrid-field
tion and beamforming. Moreover, the nonlinear communications practical and crucial in future
phase characteristics of spherical waves make the 6G networks. Thus, hybrid-field transmission tech-
design of near-field beam training and precod- niques handling both far-field and near-field signal
ing algorithms more complicated than that in far- components deserve in-depth study.
field areas. AI-based transmission methods are Spatial Non-Stationarity Effect on Near-Field
promising to address these problems since they Communications: Except for near-field propaga-
can mine the features of near-field environments tion, the spatial non-stationarity effect is another
through nonlinear neural networks. Currently, fundamental characteristic of ELAA compared
there are plenty of works elaborating on AI-based to 5G massive MIMO, where different scatter-
far-field transmissions (references are not provid- ers and users are visible to different portions of
ed here since the number of references is limited the ELAA. This effect leads to the fact that only
in this magazine), while AI-based near-field trans- a part of the ELAA can receive the spherical EM
missions have not been well studied. waves radiated by a scatterer or a user. The angu-
RIS-Aided Near-Field Communications: Com- lar power spectral and average received power
pared to MIMO communications, the near-field rapidly vary over the ELAA. Recently, there have
propagation becomes even more dominant been intensive works dealing with the non-station-
and complex in RIS-aided systems. In MIMO arity effect and near-field propagation simultane-
communications, based on the spherical propa- ously [6]. However, the impact of non-stationarity
gation model, the EM waves form spherical equi- on other emerging near-field communications has
phase surfaces at the receiver. On the contrary, not been well studied, such as RIS-aided systems
in RIS-aided systems, the phase of received EM and hybrid-field communications.
waves is accumulated by the propagation delays
through the BE-RIS and RIS-UE links. Based on Hardware Development
the geometry relationship, the equiphase surfaces To verify the effectiveness of near-field trans-
become ellipses in the near-field range instead mission technologies, hardware developments
of spherical. Accordingly, the research on beam- and over-the-air experiments are of great signifi-
focusing [15], channel estimation, and multiple cance. For example, for alleviating the near-field
access techniques taking into account this ellips- beam split effect, TTD lines need to be meticu-
es-equiphase property are required for RIS-aided lously designed in the THz domain. The hardware
near-field communications. developments of WSMS and DAP architectures
Hybrid Far- and Near-Field Communications: are worth carrying out to exploit the near-field
In practical systems, communication environments spatial DoFs. Furthermore, implementing these
usually exist with both far-field and near-field sig- techniques still has to overcome several hardware
nal components. First, in multi-user systems with impairment issues, including in-phase/quadrature
multi-path channels, some users and scatterers imbalance, low-efficiency power amplifier at high
may be far away from the BS while others are frequency, and so on. All these challenges should
located in the near-field region of the BS, which be carefully addressed to enable the implementa-
constitutes a hybrid far- and near-field (hybrid- tion of 6G near-field communications.
Authorized licensed use limited to: Kookmin University. Downloaded on July 20,2023 at 01:19:32 UTC from IEEE Xplore. Restrictions apply.
FIGURE 5. Near-field beamfocusing is able to serve multiple users in the same angle.
Authorized licensed use limited to: Kookmin University. Downloaded on July 20,2023 at 01:19:32 UTC from IEEE Xplore. Restrictions apply.