江波杏子-mdi是什么
Five Disruptive Technology Directions for 5G
ABSTRACT: New research directions will lead to fundamental changes in the design
of future 5th generation (5G) cellular networks. This paper describes five technologies
that
could
lead
to
both
architectural
and
component
disruptive
design
changes:
device
-
centric architectures, millimeter Wave, Massive
-
MIMO, smarter devices, and
native
support
to
machine
-
2
-
machine.
The
key
ideas
for
each
technology
are
described,
along
with
their
potential
impact
on
5G
and
the
research
challenges
that
remain.
UCTION:
5G is coming. What technologies will define it? Will 5G be just an evolution of 4G, or
will
emerging
technologies
cause
a
disruption
requiring
a
wholesale
rethinking
of
entrenched cellular principles? This paper focuses on potential disruptive technologies
and their implications for 5G. We classify the impact of new technologies, leveraging
the Henderson
-
Clark model [1], as follows:
changes at both the node and the architectural level, e.g., the introduction of
codebooks and signaling support for a higher number of antennas. We refer to these as
evolutions in the design.
tive changes in the design of a class of network nodes, e.g., the introduction
of a new waveform. We refer to these as component changes.
tive changes in the system architecture, e.g., the introduction of new types of
nodes or new functions in existing ones. We refer to these as architectural changes.
tive changes that have an impact at both the node and the architecture levels.
We refer to these as radical changes.
We focus on disruptive (component, architectural or radical) technologies, driven by
our belief that the extremely higher aggregate data rates and the much lower latencies
required
by
5G
cannot
be
achieved
with
a
mere
evolution
of
the
status
quo.
We
believe that the following five potentially disruptive technologies could lead to both
architectural and component design changes, as classified in Figure 1.
-
centric architectures.
The base
-
station
-
centric architecture of cellular systems may change in 5G. It may be
time to reconsider the concepts of uplink and downlink, as well as control and data
channels,
to
better
route
information
flows
with
different
priorities
and
purposes
towards
different
sets
of
nodes
within
the
network.
We
present
device
-
centric
architectures in Section II.
eter Wave (mmWave).
While
spectrum
has
become
scarce
at
microwave
frequencies,
it
is
plentiful
in
the
mmWave
realm.
Such
a
spectrum
‘el
Dorado’
has
led
to
a
mmWave
‘gold
rush’
in
which
researchers
with
diverse
backgrounds
are
studying
different
aspects
of
mmWave
transmission.
Although
far
from
fully
understood,
mmWave
technologies
have
already
been
standardized
for
short
-
range
services
(IEEE
802.11ad)
and
deployed for niche applications such as small
-
cell backhaul. In Section III, we discuss
the potential of mmWave for a broader application in 5G.
e
-
MIMO.
Massive
-
MIMO1
proposes
utilizing
a
very
high
number
of
antennas
to
multiplex
messages for several devices on each time
-
frequency resource, focusing the radiated
energy
towards
the
intended
directions
while
minimizing
intra
-
and
inter
-
cell
interference. Massive
-
MIMO may require major architectural changes, in particular in
the design of macro base stations, and it may also lead to new types of deployments.
We discuss massive
-
MIMO in Section IV
.
r devices.
2G
-
3G
-
4G cellular networks were built under the design premise of having complete
control at the infrastructure side. We argue that 5G systems should drop this design
assumption
and
exploit
intelligence
at
the
device
side
within
different
layers
of
the
protocol
stack,
e.g.,
by
allowing
Device
-
to
-
Device
(D2D)
connectivity
or
by
exploiting
smart
caching
at
the
mobile
side.
While
this
design
philosophy
mainly
requires a change at the node level (component change), it has also implications at the
architectural level. We argue for smarter devices in Section V
.
support for Machine
-
to
-
Machine (M2M) communication
A
native2
inclusion
of
M2M
communication
in
5G
involves
satisfying
three
fundamentally
different
requirements
associated to
different
classes of low
-
data
-
rate
services: support of a massive number of low
-
rate devices, sustainment of a minimal
data rate in virtually all circumstances, and very
-
low
-
latency data transfer. Addressing
these requirements in 5G requires new methods and ideas at both the component and
architectural level, and such is the focus of Section VI.
-
CENTRIC ARCHITECTURES
Cellular
designs
have
historically
relied
on
the
axiomatic
role
of
‘cells’
as
fundamental units within the radio access network. Under such a design postulate, a
device obtains service by establishing a downlink and an uplink connection, carrying
both
control
and
data
traffic,
with
the
base
station
commanding
the
cell
where
the
device
is
located.
Over
the
last
few
years,
different
trends
have
been
pointing
to
a
disruption of this cell
-
centric structure:
1.
The base
-
station density is increasing rapidly, driven by the rise of heterogeneous
networks.
While
heterogeneous
networks
were
already
standardized
in
4G,
the
architecture was not natively designed to support them. Network densification could
require
some
major
changes
in
5G.
The
deployment
of
base
stations
with
vastly
different transmit powers and coverage areas, for instance, calls for a decoupling of
downlink and uplink in a way that allows for the corresponding information to flow
through different sets of nodes [5].
2.
The
need
for
additional
spectrum
will
inevitably
lead
to
the
coexistence
of
frequency bands with radically different propagation characteristics within the same
system. In this context, [6] proposes the concept of a ‘phantom cell’ where the data
and control planes are separated: the control information is sent by high
-
power nodes
at microwave frequencies whereas the payload data is conveyed by low
-
power nodes
at mm
-
Wave frequencies. (cf. Section III.)
3.A
new
concept
termed
centralized
baseband
related
to
the
concept
of
cloud
radio
access
networks
is
emerging
(cf.
[7]),
where
virtualization
leads
to
a
decoupling
between a node and the hardware allocated to handle the processing associated with
this node. Hardware resources in a pool, for instance, could be dynamically allocated
to different nodes depending on metrics defined by the network operator.
Emerging
service
classes,
described
in
Section
VI,
could
require
a
complete
redefinition
of
the
architecture.
Current
works
are
looking
at
architectural
designs
ranging
from
centralization
or
partial
centralization
(e.g.,
via
aggregators)
to
full
distribution (e.g., via compressed sensing and/or multihop).
Cooperative
communications
paradigms
such
as
CoMP
or
relaying,
which
despite
falling
short
of
their
initial
hype
are
nonetheless
beneficial
[8],
could
require
a
redefinition
of
the
functions
of
the
different
nodes.
In
the
context
of
relaying,
for
instance,
recent
developments
in
wireless
network
coding
[9]
suggest
transmission
principles that would allow recovering some of the losses associated with half
-
duplex
relays. Moreover, recent research points to the plausibility of full
-
duplex nodes for
short
-
range communication in a not
-
so
-
distant future.
The use of smarter devices (cf. Section V) could impact the radio access network. In
particular, both D2D and smart caching call for an architectural redefinition where the
center
of
gravity
moves
from
the
network
core
to
the
periphery
(devices,
local
wireless
proxies,
relays).
Based
on
these
trends,
our
vision
is
that
the
cell
-
centric
architecture
should
evolve
into
a
device
-
centric
one:
a
given
device
(human
or
machine) should be able to communicate by exchanging multiple information flows
through
several
possible
sets
of
heterogeneous
nodes.
In
other
words,
the
set
of
network
nodes
providing
connectivity
to
a
given
device
and
the
functions
of
these
nodes in a particular communication session should be tailored to that specific device
and
session.
Under
this
vision,
the
concepts
of
uplink/downlink
and
control/data
channel should be rethought (cf. Figure 2).
While
the
need
for
a
disruptive
change
in
architectural
design
appears
clear,
major
research efforts are still needed to transform the resulting vision into a coherent and
realistic
proposition.
Since
the
history
of
innovations
(cf.
[1])
indicates
that
architectural changes are often the drivers of major technological discontinuities, we
believe that the trends above might have a major influence on the development of 5G.
ETER WA
VE COMMUNICATION
Microwave
cellular
systems
have
precious
little
spectrum:
around
600
MHz
are
currently in use, divided among operators [10]. There are two ways to gain access to
more microwave spectrum:
1.
To
repurpose
or
refarm
spectrum.
This
has
occurred
worldwide
with
the
repurposing of terrestrial TV spectrum for applications such as rural broadband access.
Unfortunately, repurposing has not freed up that much spectrum, only about 80 MHz,
and at a high cost associated with moving the incumbents.
share spectrum utilizing, for instance, cognitive radio techniques. The high hopes
initially placed on cognitive radio have been dampened by the fact that an incumbent
not fully willing to cooperate is a major obstacle to spectrum efficiency for secondary
users.
ther,
it
appears
that
a
doubling
of
the
current
cellular
bandwidth
is
the
best
-
case
scenario
at
microwave
frequencies.
Alternatively,
there
is
an
enormous
amount of spectrum at mmWave frequencies ranging from 3 to 300 GHz. Many bands
therein seem promising, including most immediately the local multipoint distribution
service at 28
-
30 GHz, the license
-
free band at 60 GHz, and the E
-
band at 71
-
76 GHz,
81
-
86 GHz and 92
-
95 GHz. Foreseeably, several tens of GHz could become available
for
5G,
offering
well
over
an
order
-
of
-
magnitude
increase
over
what
is
available
at
present. Needless to say, work needs to be done on spectrum policy to render these
bands available for mobile cellular.
3.
Propagation
is
not
an
insurmountable
challenge.
Recent
measurements
indicate
similar
general
characteristics
as
at
microwave
frequencies,
including
distance
-
dependent pathloss and the possibility of non
-
line
-
of
-
sight communication. A
main
difference
between
microwave
and
mmWave
frequencies
is
the
sensitivity
to
blockages:
the
results
in
[11],
for
instance,
indicate
a
pathloss
exponent
of
2
for
line
-
of
-
sight propagation but 4 (plus an additional power loss) for non
-
line
-
of
-
sight.
MmWave cellular research will need to incorporate sensitivity to blockages and more
complex channel models into the analysis, and also study the effects of enablers such
as higher density infrastructure and relays. Another enabler is the separation between
control and data planes, already mentioned in Section II.
Antenna arrays
are
a key
feature in
mmWave systems.
Large
arrays can
be used to
keep the antenna aperture constant, eliminating the frequency dependence of pathloss
relative
to
omnidirectional
antennas
(when
utilized
at
one
side
of
the
link)
and
providing
a
net
array
gain
to
counter
the
larger
thermal
noise
bandwidth
(when
utilized at both sides of the link). Adaptive arrays with narrow beams also reduce the
impact
of interference,
meaning that mmWave
systems
could
more often operate in
noise
-
limited
rather
than
interference
-
limited
conditions.
Since
meaningful
communication
might
only
happen
under
sufficient
array
gain,
new
random
access
protocols are needed that work when transmitters can only emit in certain directions
and
receivers
can
only
receive
from
certain
directions.
Adaptive
array
processing
algorithms are required that can adapt quickly when beams are blocked by people or
when some device antennas become obscured by the user’s own body.
MmWave systems also have distinct hardware constraints. A major one comes from
the
high
power
consumption
of
mixed
signal
components,
chiefly
the
analog
-
to
-
digital
(ADC)
and
digital
-
to
-
analog
converters
(DAC).
Thus,
the
conventional microwave architecture where every antenna is connected to a high
-
rate
ADC/DAC is unlikely to be applicable to mmWave without a huge leap forward in
semiconductor
technology.
One
alternative
is
a
hybrid
architecture
where
beamforming
is
performed
in
analog
at
RF
and
multiple
sets
of
beamformers
are
connected to a small number of ADCs or DACS; in this alternative, signal processing
algorithms are needed to steer the analog beamforming weights. Another alternative is
to connect each RF chain to a 1
-
bit ADC/DAC, with very low power requirements; in
this
case,
the
beamforming
would
be
performed
digitally
but
on
very
noisy
data.
There are abundant research challenges in optimizing different transceiver strategies,
analyzing their capacity, incorporating multiuser capabilities, and leveraging channel
features such as sparsity.
A
data
rate
comparison
between
technologies
is
provided
in
Fig.
3,
for
certain
simulation settings, in terms of mean and 5% outage rates. MmWave operation is seen
to provide very high rates compared to two different microwave systems. The gains
exceed the 10x spectrum increase because of the enhanced signal power and reduced
interference thanks to directional beamforming at both transmitter and receiver.
IV
.MASSIVE MIMO
Massive
MIMO
(also
referred
to
as
‘Large
-
Scale
MIMO’
or
‘Large
-
Scale
Antenna
Systems’) is a form of multiuser MIMO in which the number of antennas at the base
station is much larger than the number of devices per signaling resource [14]. Having
many
more
base
station
antennas
than
devices
renders
the
channels
to
the
different
devices
quasi
-
orthogonal
and
very
simple
spatial
multiplexing/de
-
multiplexing
procedures
quasi
-
optimal.
The
favorable
action
of
the
law
of
large
numbers
smoothens out frequency dependencies in the channel and, altogether, huge gains in
spectral efficiency can be attained (cf. Fig. 4).
In the context of the Henderson
-
Clark framework, we argue that massive
-
MIMO has
a disruptive potential for 5G:
At
a
node
level,
it
is
a
scalable
technology.
This
is
in
contrast
with
4G,
which,
in
many respects, is not scalable: further sectorization therein is not feasible because of
(i) the limited space for bulky azimuthally
-
directive antennas, and (ii) the inevitable
angle
spread
of
the
propagation;
in
turn,
single
-
user
MIMO
is
constrained
by
the
limited number of antennas that can fit in certain mobile devices. In contrast, there is
almost no limit on the number of base station antennas in massive
-
MIMO provided
that time
-
division duplexing is employed to enable channel estimation through uplink
pilots.
It
enables
new
deployments
and
architectures.
While
one
can
envision
direct
replacement of macro base stations with arrays of low
-
gain resonant antennas, other
deployments
are
possible,
e.g.,
conformal
arrays
on
the
facades
of
skyscrapers
or
arrays
on
the
faces
of
water
tanks
in
rural
locations.
Moreover,
the
same
massive
-
MIMO principles that govern the use of collocated arrays of antennas apply
also to distributed deployments in which a college campus or an entire city could be
covered with a multitude of distributed antennas that collectively serve many users (in
this
framework,
the
centralized
baseband
concept
presented
in
Section
II
is
an
important architectural enabler).
While very promising, massive
-
MIMO still presents a number of research challenges.
Channel estimation is critical and currently it represents the main source of limitations.
User
motion
imposes
a
finite
coherence
interval
during
which
channel
knowledge
must be acquired and utilized, and consequently there is a finite number of orthogonal
pilot sequences that can be assigned to the devices. Reuse of pilot sequences causes
pilot
contamination
and
coherent
interference,
which
grows
with
the
number
of
antennas
as
fast
as
the
desired
signals.
The
mitigation
of
pilot
contamination
is
an
active
research
topic.
Also,
there
is
still
much
to
be
learned
about
massive
-
MIMO
propagation,
although
experiments
thus
far
support
the
hypothesis
of
channel
quasi
-
orthogonality.
From
an
implementation
perspective,
massive
-
MIMO
can
potentially be realized with modular low
-
cost low
-
power hardware with each antenna
functioning semi
-
autonomously, but a considerable development effort is still required
to
demonstrate
the
cost
-
effectiveness
of
this
solution.
Note
that,
at
the
microwave
frequencies
considered
in
this
section,
the
cost
and
the
energy
consumption
of
ADCs/DACs are sensibly lower than at mmWave frequencies (cf. Section III).
From the discussion above, we conclude that the adoption of massive
-
MIMO for 5G
could
represent
a
major
leap
with
respect
to
today’s
state
-
of
-
the
-
art
in
system
and
component design. To justify these major changes, massive
-
MIMO proponents should
further
work
on
solving
the
challenges
emphasized
above
and
on
showing
realistic
performance
improvements
by
means
of
theoretical
studies,
simulation
campaigns,
and testbed experiments.
V
.SMARTER DEVICES
Earlier
generations
of
cellular
systems
were
built
on
the
design
premise
of
having
complete
control
at
the
infrastructure
side.
In
this
section,
we
discuss
some
of
the
possibilities that can be unleashed by allowing the devices to play a more active role
and, thereafter, how 5G’s design should account for an increase in device smartness.
We focus on three different examples of technologies that could be incorporated into
smarter devices, namely D2D, local caching, and advanced interference rejection.
V
.1 D2D
In voice
-
centric systems it was implicitly accepted that two parties willing to establish
a call would not be in close proximity. In the age of data, this premise might no longer
hold,
and
it
could
be
common
to
have
situations
where
several
co
-
located
devices
would
like
to
wirelessly
share
content
(e.g.,
digital
pictures)
or
interact
(e.g.,
video
gaming
or
social
networking).
Handling
these
communication
scenarios
via
simply
connecting through the network involves gross inefficiencies at various levels:
1.
Multiple
wireless
hops
are
utilized
to
achieve
what
requires,
fundamentally,
a
single
hop.
This
entails
a
multifold
waste
of
signaling
resources,
and
also
a
higher
latency. Transmit powers of a fraction of a Watt (in the uplink) and several Watts (in
the
downlink)
are
consumed
to
achieve
what
requires,
fundamentally,
a
few
milliWatts. This, in turn, entails unnecessary levels of battery drain and of interference
to all other devices occupying the same signaling resources elsewhere.
2.
Given that the pathlosses to possibly distant base stations are much stronger than
the direct
-
link ones, the corresponding spectral efficiencies are also lower. While it is
clear
that
D2D
has
the
potential
of
handling
local
communication
more
efficiently,
local
high
-
data
-
rate
exchanges
could
also
be
handled
by
other
radio
access
technologies such as Bluetooth or Wi
-
Fi direct. Use cases requiring a mixture of local
and nonlocal content or a mixture of low
-
latency and high
-
data
-
rate constraints (e.g.,
interaction
between
users
via
augmented
reality),
could
represent
more
compelling
reasons for the use of D2D. In particular, we envision D2D as an important enabler
for
applications
requiring
low
-
latency
3
,
especially
in
future
network
deployments
utilizing baseband centralization and radio virtualization (cf. Section I).
From a research perspective, D2D communication presents relevant challenges:
fication
of
the
real
opportunities
for
D2D.
How
often
does
local
communication
occur?
What
is
the
main
use
case
for
D2D:
fast
local
exchanges,
low
-
latency applications or energy saving?
ation of a D2D mode with the uplink/downlink duplexing structure.
of D2D
-
enabled devices, from both a hardware and a protocol perspective,
by providing the needed flexibility at both the PHY and MAC layers.
ing
the
true
net
gains
associated
with
having
a
D2D
mode,
accounting
for
possible extra overheads for control and channel estimation.
y, note that, while D2D is already being studied in 3GPP as a 4G add
-
on2, the
main focus of current studies is proximity detection for public safety [15]. What we
discussed here is having a D2D dimension natively supported in 5G.
V
.2 Local Caching
The
current
paradigm
of
cloud
computing
is
the
result
of
a
progressive
shift
in
the
balance between data storage and data transfer: information is stored and processed
wherever
it
is
most
convenient
and
inexpensive
because
the
marginal
cost
of
transferring it has become negligible, at least on wireline networks [2]. For wireless
devices though, this cost is not always negligible. The understanding that mobile users
are subject to sporadic ‘abundance’ of connectivity amidst stretches of ‘deprivation’ is
hardly new, and the natural idea of opportunistically leveraging the former to alleviate
the
latter
has
been
entertained
since
the
1990s
[3].
However,
this
idea
of
caching
massive amounts of data at the edge of the wireline network, right before the wireless
hop, only applies to delay
-
tolerant traffic and thus it made little sense in voice
-
centric
systems. Caching might finally make sense now, in data
-
centric systems [4].
Thinking
ahead,
it
is
easy
to
envision
mobile
devices
with
truly
vast
amounts
of
memory.
Under
this
assumption,
and
given
that
a
substantial
share
of
the
data
that
circulates wirelessly corresponds to the most popular audio/video/social content that
is in vogue at a given time, it is clearly inefficient to transmit such content via unicast
and
yet
it
is
frustratingly
impossible
to
resort
to
multicast
because
the
demand
is
asynchronous.
We
hence
see
local
caching
as
an
important
alternative,
both
at
the
radio access network edge (e.g., at small cells) and at the mobile devices, also thanks
to enablers such as mmWave and D2D.
V
.3 Advanced Interference Rejection
In
addition
to
D2D
capabilities
and
massive
volumes
of
memory,
future
mobile
devices
may
also
have
varying
form
factors.
In
some
instances,
the
devices
might
accommodate
several
antennas
with
the
consequent
opportunity
for
active
interference
rejection
therein,
along
with
beamforming
and
spatial
multiplexing.
A
joint
design
of
transmitter
and
receiver
processing,
and
proper
control
and
pilot
signals, are critical to allow advanced interference rejection. As an example, in Fig. 5
we
show
the
gains
obtained
by
incorporating
the
effects
of
nonlinear,
intra
and
inter
-
cluster interference awareness into devices with 1, 2 and 4 antennas.
While this section has been mainly focused on analyzing the implications of smarter
devices at a component level, in Section II we discussed the impact at the radio access
network architecture level. We regard smarter devices as having all the characteristic
of
a
disruptive
technology
(cf.
Section
I)
for
5G,
and
therefore
we
encourage
researchers to further explore this direction.
SUPPORT FOR M2M COMMUNICATION
Wireless communication is becoming a commodity, just like electricity or water [13].
This commoditization, in turn, is giving rise to a large class of emerging services with
new types of requirements. We point to a few representative such requirements, each
exemplified by a typical service:
1.A massive number of connected devices. Whereas current systems typically operate
with,
at
most,
a
few
hundred
devices
per
base
station,
some
M2M
services
might
require over 104 connected devices. Examples include metering, sensors, smart grid
components, and other enablers of services targeting wide area coverage.
high link reliability. Systems
geared at critical control, safety, or production,
have been dominated by wireline connectivity largely because wireless links did not
offer
the
same
degree
of
confidence.
As
these
systems
transition
from
wireline
to
wireless, it becomes necessary for the wireless link to be reliably operational virtually
all the time.
latency
and
real
-
time
operation.
This
can
be
an
even
more
stringent
requirement than the ones above, as it demands that data be transferred reliably within
a given time interval. A typical example is Vehicle
-
to
-
X connectivity, whereby traffic
safety can be improved through the timely delivery of critical messages (e.g., alert and
control).
Fig. 5 provides a perspective on the M2M requirements by plotting the data rate vs.
the device population size. This cartoon illustrates where systems currently stand and
how the research efforts are expanding them. The area R1 reflects the operating range
of
today’s
systems,
outlining
the
fact
that
the
device
data
rate
decreases
as
its
population increases. In turn, R2 is the region that reflects current research aimed at
improving the spectral efficiency. Finally, R5 indicates the region where operation is
not feasible due to fundamental physical and information
-
theoretical limits.
Regions R3 and R4 correspond to the emerging services discussed in this section:
R3 refers to massive M2M communication where each connected machine or sensor
transmits
small
data
blocks
sporadically.
Current
systems
are
not
designed
to
simultaneously
serve
the
aggregated
traffic
accrued
from
a
large
number
of
such
devices. For instance, a current system could easily serve 5 devices at 2 Mbps each,
but not 10000 devices each requiring 1 Kbps. R4 demarks the operation of systems
that require high reliability and/or low latency, but with a relatively low average rate
per
device.
The
complete
description
of
this
region
requires
additional
dimensions
related to reliability and latency.
There are services that pose simultaneously more than one of the above requirements,
but the common point is that the data size of each individual transmission is small,
going down to several bytes. This profoundly changes the communication paradigm
for the following reasons:
Existing coding methods that rely on long codewords are not applicable to very short
data
blocks.
Short
data
blocks
also
exacerbate
the
inefficiencies
associated
with
control and channel
estimation
overheads. Currently, the
control
plane is robust but
suboptimal
as
it
represents
only
a
modest
fraction
of
the
payload
data;
the
most
sophisticated
signal
processing
is
reserved
for
payload
data
transmission.
An
optimized design should aim at a much tighter coupling between the data and control
planes.
As mentioned in Section II, the architecture needs a major redesign, looking at new
types of nodes. At a system level, the frame
-
based approaches that are at the heart of
4G
need
rethinking
in
order
to
meet
the
requirements
for
latency
and
flexible
allocation of resources to a massive number of devices. From the discussion above,
and from
the related architectural
consideration in
Section
II,
and referring one last
time to the Henderson
-
Clark model, we conclude that a native support of M2M in 5G
requires radical
changes at
both
the node and the architecture level.
Major research
work
remains
to
be
done
to
come
up
with
concrete
and
interworking
solutions
江波杏子-mdi是什么
江波杏子-mdi是什么
江波杏子-mdi是什么
江波杏子-mdi是什么
江波杏子-mdi是什么
江波杏子-mdi是什么
江波杏子-mdi是什么
江波杏子-mdi是什么
本文更新与2021-01-20 05:35，由作者提供，不代表本网站立场，转载请注明出处：https://www.bjmy2z.cn/gaokao/536857.html