Professional Documents
Culture Documents
Received November 15, 2016, accepted December 3, 2016, date of publication December 15, 2016, date of current version January 23,
2017.
Digital Object Identifier 10.1109/ACCESS.2016.2640270
ABSTRACT Caching popular contents at mobile devices can potentially improve the quality of service
for mobile users and relieve traffic burden of base station in cellular networks. In this paper, we jointly
consider the resource allocation, the cached contents, and the distance between two devices for the optimal
device pairing problem in centralized and distributed cases, where the BS is the central controller in the
centralized case. The joint optimization problem of device-to-device (D2D) caching with channel allocation
is formulated as a weighted four-uniform hypergraph model. The optimal solution for the problem is
4-D maximum weighted matching (4-DMWM), which is NP-hard unfortunately. To approach the 4-DMWM
with low-complexity, we adopt the greedy algorithm and the squareIMP algorithm in the centralized case.
Moreover, distributed algorithms are also designed for the caching problem in both synchronous and
asynchronous cases. The simulation results will illustrate that the squareIMP algorithm can be used to get a
better transmission rate with the complexity of O(n5 ), while the greedy algorithm can be used in the case with
stringent latency requirement for centralized 4-DMWM problem. The sum rate of distributed asynchronous
algorithm is close to the centralized greedy algorithm with the complexity of O(n2 ) for each device.
However, the simulation result of the synchronous algorithm is slightly lower than the centralized algorithm,
where each device performs O(n2 ) computational operations in each iteration. Therefore, the algorithms
proposed in this paper can be used in different cases for solving optimal D2D pairing and channel allocation
problem.
INDEX TERMS D2D caching, weighted 4-uniform hypergraph, 4-dimensional maximum weighted
matching (4-DMWM), centralized algorithm, distributed algorithm.
2169-3536
2016 IEEE. Translations and content mining are permitted for academic research only.
VOLUME 4, 2016 Personal use is also permitted, but republication/redistribution requires IEEE permission. 9213
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
L. Miao et al.: 4-DMWM Approach for Caching-Based Optimal D2D Pairing and Channel Allocation
networks and device-to-device (D2D) networks respectively. which is shown to achieve the 1/2 optimal performance
In [7], the caching performance was studied in terms of in [22].
outage probability and average delivery rate in wireless Usually, BS in centralized case is required to collect the
small-cell networks, where the cache-enabled BSs are dis- whole information of the network, which introduces much
tributed according to a Poisson point process. In [8], the overhead. However, coordination may not be possible with no
throughput-outage tradeoff performance of wireless networks central controller exists in some self-organized networks [14].
was investigated by exploiting clustered device caching via Moreover, the underlay caching and interference manage-
D2D communications. The caching and pushing problem ment algorithms often bring in much extra computation over-
with a finite receiver buffer was studied in [9] and [10]. There head for the central controller. Therefore, we use distributed
are many studies using matching algorithm for resource allo- algorithms to solve the caching based optimal D2D pairing
cation. In OFDMA systems, the problem of optimal subcar- and channel allocation problem. In the distributed algorithm,
rier allocation can be solved by H-matching approach to min- we consider synchronous and asynchronous scenarios for the
imize outage probability in [11]. To offload not only traffic caching based optimal D2D pairing and channel allocation
but also computation overhead, some distributed algorithms problem. A bipartite graph matching problem can be solved
are adopted by pushing computation and storage to network in [23]. Inspired by [23], we convert the weighted 4-uniform
edges for the caching problem in D2D network [12], [13]. hypergraph to a weighted bipartite graph. After the trans-
A belief propagation based distributed algorithm was used to formation, the 4-DMWM problem can be formulate as a
solve the cache placement problem in [14]. In the distributed weighted bipartite graph matching problem with a little per-
system, the appropriate channel allocation can achieve a low formance loss in the distributed synchronous system. To solve
outage probability with low computation complexity [15]. the weighted matching problem, belief propagation (BP) is
To the best of our knowledge, however, these studies have an effective algorithm that explained in [24] and [25]. In the
not considered the joint optimization problem of content distributed synchronous system, a distributed synchronous
caching, resource allocation and device pairing to improve algorithm based on BP is proposed for the weighted bipar-
the throughput in D2D networks. tite graph matching problem, so that we obtain the approxi-
In this paper, we consider the problem of caching based mate solution for the 4-DMWM problem. In the distributed
optimal D2D pairing and channel allocation. In the con- synchronous algorithm, each device performs only O(n2 )
sidered system, each mobile device caches some popular computational operations in each iteration. Moreover, in the
contents. 3GPP Proximity Services (ProSe) discovery can distributed asynchronous algorithm, the 4-DMWM problem
provide registration of both services offered by devices, and based on the weighted 4-uniform hypergraph can be trans-
application requests of users in [16]. In the same way, a formed into a maximum independent set (MIS) problem. We
device can get the content caching information of users in search for each device and its neighbors to find the maximum
D2D network. When some users request the contents, the sys- independent set of devices one by one in the distributed
tem will determine which device provides help and allocate asynchronous system. After that, we get the 4-DMWM result
the channel at the same time, so as to maximize the system with the complexity of O(n2 ) for each device. The distributed
throughput. Inspired by [17], the considered joint optimiza- algorithms in this paper are both low-complexity for the
tion problem is first formulated as a weighted 4-uniform caching based optimal D2D pairing and channel allocation
hypergraph model. In this context, the optimal solution for the problem.
caching based D2D pairing and channel allocation problem is The main contributions of this paper are summarized as
the 4-dimensional maximum weighted matching (4-DMWM) follows:
in the weighted 4-uniform hypergraph. Unfortunately, the • We propose the D2D pairing and channel allocation
multi-dimensional matching problem is NP-hard in general. algorithm scheme based on the squareIMP and a greedy
Therefore, we adopt some approximation algorithms in this algorithm to solve the caching based optimal D2D pair-
paper with low complexity in centralized and distributed ing and channel allocation problem in the centralized
manners. manner.
Recently proposed cloud radio access networks (Cloud- • In the distributed algorithm, there are synchronous and
RAN) is a typical centralized cloud computing-based archi- asynchronous scenarios.
tecture [18], while fog radio access networks (Fog-RAN) – In the distributed synchronous system, a distributed
is a distributed architecture that leverages caching capabil- synchronous algorithm based on BP is proposed
ities at the wireless edge nodes for cellular networks [19]. for optimal D2D pairing and caching with channel
In [20] we focused our attention on the centralized 4-DMWM allocation with low complexity.
problem under the control of BS. In the centralized case, – In distributed asynchronous system, a low complex-
based on the squareIMP algorithm proposed in [21], we first ity distributed asynchronous algorithm is proposed
propose high performance algorithm with the complexity to solve the caching based optimal D2D pairing and
of O(n5 ), which nearly achieves the optimal throughput of the channel allocation problem.
exhaustive search. Moreover, to fulfill the stringent latency The rest of this paper is organized as follows.
requirement, we further proposed a greedy based algorithm, Section II presents the system model and problem
formulation. The hypergraph based centralized 4-DMWM In our network, many pairs of devices send requests at
algorithm for caching based optimal D2D pairing and chan- the same time. Therefore, we protect the device pairs from
nel allocation is proposed in Section III. The distributed interference by considering the resource allocation. Since
synchronous and asynchronous algorithm for the caching the underlay interference management algorithms often have
problem are discussed in Section IV. Section V illustrates much extra computation overhead for the BS, a hierarchi-
the simulation results. Finally, Section VI concludes the cal overlay spectrum sharing and relay selection scheme
paper. is proposed to avoid serious interference between cellular
devices in [13]. The BSs allocate interleaved resource blocks
to the entire D2D network at first. The interleaved subcar-
rier allocation scheme is adopt in both IEEE 802.16 and
LTE-A standards [26], [27]. The BS allocates the resource
blocks in different coherence bandwidths, so that different
resource blocks undergo independent channel fading. After
that the D2D network will obtain the full frequency diversity
according to [28]. We assume that the index set of resource
blocks is denoted by B, where B = {1, . . . , B}. Let hi be the
channel gain on the resource block b between the devices. The
channel gain follows a circular symmetric Gaussian distribu-
tion [29]. Suppose the resource block has a unit bandwidth.
The channel between two devices in D2D communications is
Rayleigh fading channel. In this case, we have:
yi = hi xi + zi , i = 1, 2, . . . , (1)
where wr,l,b is the channel capacity between requester r and A. HYPERGRAPH FORMULATION
helper l by using the resource block b, and A is a set of A hypergraph can be expressed as HG = (V, E) where V is
these elements (r, l, b) in D2D links. Therefore, we focus on the set of vertices and E is the set of hyperedges. A weighted
maximizing the sum-rate for the caching based optimal D2D hypergraph means each hyperedge e ∈ E has a weight w(e).
pairing with channel allocation. A hyperedge can be represented as a nonempty subset of
the vertices. The number of vertices is the cardinality of the the 4-DMWM problem can be solved by finding a w-MIS
hyperedge. If each hyperedge has the same cardinality n, the solution in graph G. The w-MIS problem of an ordinary
hypergraph is n-uniform. Therefore, an ordinary graph is a weighted graph G is NP-hard. Therefore, we need some
2-uniform hypergraph. It means each edge has a subset of two approximation algorithms to find the w-MIS result.
vertices. Content requester, content helper, resource block
and content item of each D2D link can be represented as 1) SquareIMP ALGORITHM
4 vertices in a hyperedge. Each hyperedge has a weight that The squareIMP algorithm needs polynomial time to find an
indicate the channel capacity of the D2D link. Therefore, the independent set A, where there is a k-claw free graph.
caching based optimal D2D pairing with channel allocation A k-claw W is an induced subgraph that consists of a center
problem can be formulated as a weighted 4-uniform hyper- node and talons TW in an undirected graph. The center node
graph, as illustrated in Fig. 2. is connected to an independent set of k talons. A graph that is
A matching of an ordinary graph G is a subset of the edges k-claw free means it possesses no k-claws induced subgraph.
in G such that no edges of the matching are incident. It means Set the center set of a claw W as ZW = W − TW [21]. In the
that each vertex is covered at most once in these matching 4-DMWM problem, the weighted graph G is a 5-claw free
edges. Similarly, a hypergraph matching of a hypergraph HG graph. To realize the squareIMP algorithm, we search for each
is a subset of the hyperedges in HG such that no hyper- claw in the 5-claw free graph to update A by improving w2 (A)
edges contain the same vertex. The cardinality of a matching in Algorithm 2. We define
is the number of matching edges. If a matching has the N (TW , A) = {u ∈ A : ∃v ∈ TW such that u, v ∈ Eor u = v}.
largest cardinality of all possible matching results, it is called
the maximum matching. The multi-dimensional maximum If the TW of a claw W can improve w2 (A), we update the set
weighted matching of a weighted hypergraph HG is a subset A as follows.
of the matching hyperedges in HG. In this case, our target A ←− A ∪ TW \ N (TW , A).
is to solve the 4-dimensional maximum weighted matching
(4-DMWM) problem. Therefore, the caching based optimal Therefore, the set A is the expected set of D2D links for the
D2D pairing problem with channel allocation is to find the w-MIS problem.
4-DMWM solution in the weighted 4-uniform hypergraph.
2) GREEDY ALGORITHM
B. CENTRALIZED 4-DMWM BASED D2D PAIRING The greedy algorithm is a simple algorithm to find an inde-
AND CHANNEL ALLOCATION ALGORITHM pendent set A. In the greedy algorithm, we search for all
In this paper, we formulate the 4-DMWM problem for the vertices to choose the heaviest independent set of vertices in
caching based optimal D2D pairing and channel allocation. the graph G, as shown in Algorithm 3.
However, the 4-DMWM problem is a NP-hard problem. The 4-DMWM is the subset of mutually disjoint hyper-
Therefore, we propose a caching joint optimal D2D pairing edges. The set A corresponds to a subset of hyperedges in
and channel allocation algorithm to maximize the sum-rate hypergraph. The maximum weighted independent set A is a
with low-complexity. set of vertices in graph G. The index set of A can be expressed
The MIS problem of graph G is to find a maximum as A = {1, . . . , A}. Each vertex a of A corresponds to a D2D
collection of mutually non-adjacent vertices. Since the hyper- link for the caching based optimal D2D pairing and channel
graph matching is closely related to the MIS problem, the allocation.
4-DMWM problem based on the weighted 4-uniform hyper- TABLE 1. Complexity of Algorithm
graph can be transformed into a MIS problem. First, we
create a graph G = (V 0 , E 0 ) where each vertex is a sub-
set of 4 vertices in a hyperedge and the edge E 0 captures
all intersections between the vertices V 0 . In our weighted
4-uniform hypergraph, the hyperedges of HG correspond to
the vertices of graph G. Therefore, each vertex of G has a
weight that indicate the channel capacity of the D2D link. We
define the maximum weighted matching set as A, where A is
the set of the (r, l, b, f ) combinations. The w-MIS is to find We compare the complexity of the optimal, squareIMP,
an independent set A with maximum w(A). Therefore, the greedy and random algorithm, as shown in Table 1. The
hypergraph HG can be transformed into an ordinary weighted squareIMP algorithm is more achievable than the exhausted
graph G = (V 0 , E 0 , w). Each vertex of weighted graph G is an search with the complexity of O(n5 ). The greedy algorithm
element of potential combinations in Fig. 2. The edge of G and random algorithm are much easier to implement than the
indicates non-empty set intersections between two vertices. other algorithms.
In the D2D network, each vertex represents a D2D link and As shown in Algorithm 1, we maximize the sum-rate for
an edge means that two D2D links have one or more same the caching based optimal D2D pairing and channel allo-
elements in the weighted 4-uniform hypergraph. In this case, cation in D2D network. In the D2D pairing and channel
Algorithm 1 D2D Pairing and Channel Allocation Algorithm Algorithm 3 Greedy Algorithm
Input: M , B, F, α Input: graph G = (V 0 , E 0 , w)
Output: Srate Output: A
Construct HG = (V, E, w) with M , B, F A=∅
Map HG to graph G Vsort =sorted(V 0 , ’descend’)
if α = 1 then for v in Vsort do
A = msq (G) //Algorithm2 N (A, V) = {u ∈ V : ∃v ∈ A such that {u, v} ∈ E 0 or
else u = v}
if α = 0 then if N (A, v) = ∅ then
A = mgr (G) //Algorithm3 A←A∪v
end if end if
end if P end for
Srate (A) = (wi )
i∈A
FIGURE 3. The bipartite graph for the caching based optimal D2D pairing and channel allocation in the distributed
synchronous algorithm.
and distributed caching content statistics, etc. To avoid heavy which contains P a factor ψi (xEi ) for each vertex i ∈ V.
overhead of global network information and maximize the ψi (xEi ) = 1 if e∈E xe ≤ 1, and 0 otherwise. It is noted
sum-rate of the D2D network, BP based distributed syn- that i is referred to both the vertices of BG and factors of p.
chronous algorithm for 4-DMWM problem is proposed by Moreover, e is referred to both the edges of BG and variables
transforming the hypergraph HG into a weighted bipartite of p. The factor ψi (xEi ) means that at most one edge incident
graph BG. In the BP based distributed synchronous algorithm, to vertex i can be assigned the value 1. If xe = 1, e. Therefore,
the weighted matching can be formulated as follows. the edges {e | xe = 1} construct a matching in BG and
X p(x) = 0 otherwise.
max we xe ; The max-sum algorithm passes messages between vari-
e∈E ables and the factors at each iteration t for the caching based
X
s.t. xe ≤ 1 for all i ∈ V, optimal D2D pairing and channel allocation. For the p in (10),
e∈Ei each variable is a member of exactly two factors. An estimate
xe ∈ {0, 1} for all e ∈ E, (9) x̂ is the output of the MAP of p. (i, j) and e denotes the same
edge in this case. Set xeh1 , xeh2 , xeh3 , xeh4 as the 4 edges in
where Ei is the set of edges incident to vertex i in bipartite bipartite graph corresponding to a hyperedges of the hyper-
graph. The linear programming relaxation can be realized graph HG. Then we present the max-sum update equations
by replacing the constraint xe ∈ {0, 1} with the constraint adapted for the p in (10). At last, BP based distributed caching
0 ≤ xe ≤ 1 for each e ∈ E in the matching problem [24]. algorithm can find the matching result for the distributed
BP is a method to estimate the marginal distributions 4-DMWM problem based on the weighted 4-uniform hyper-
with respect to the variables. The method associates with a graph, as shown in Algorithm 4.
bipartite graph BG called the factor graph. With this graph, The computational complexity is measured by the number
BP iteratively passes belief messages along the edges of of calculations in the BP based distributed caching algorithm
the graph that represent estimates of the marginal distribu- with a graph of size n. Each vertex of bipartite graph BG
tions with respect to the variables [30]. The Max-product sends a vector of at most size n to each of the neighbor
form of belief propagation allows us to find the most likely vertices in the other partition in each iteration of the BP based
state of a probability distribution [24]. The max-sum algo- distributed synchronous algorithm. Each vertex has at most
rithm is equivalent to the max-product algorithm, but work n − 1 neighbors. Thus, total number of messages exchanged
in log space. Therefore, it can avoid potential underflow in each iteration are O(n2 ). Each vertex performs O(n) basic
problem. The max-sum algorithm works by iteratively pass- computational operations to compute each element in a mes-
ing messages between the variables and factors. In order to sage vector of at most size n. That is, each vertex performs
apply max-sum algorithm, we formulate maximum weighted O(n2 ) computational operations to compute a message vector
matching on bipartite graph BG as a MAP estimation prob- in each iteration. Since each vertex sends at most n message
lem by constructing a suitable probability distribution. In vectors, the total cost is O(n3 ) per iteration for all vertices.
our max-sum algorithm based caching problem, set a binary By sharing computing tasks, each device does fewer calcula-
variable xe ∈ {0, 1} with each edge e ∈ E. The probability tions when running the distributed algorithm.
The channel between two devices is independent Rayleigh with B = 2, F = 3. Observed from Fig. 4, the sum-rate
fading channel. In this paper, we consider the influence of increases with the number of devices. It is clearly shown that
physical distance, resource block and content item during the the 4-DMWM based squareIMP algorithm can achieve an
communication. We set the number of resource blocks and highly approximate performance compared with the optimal
content items as B and F, respectively. Suppose the transmis- approach. The optimal approach is NP-hard which is com-
sion time of each resource block is 20ms. Our simulation has putational prohibited. Therefore, squareIMP algorithm can
the same result when we choose other parameters. In our sim- guarantee the performance of sum-rate with the complexity
ulation, we analyze different schemes for the caching based of O(n5 ) in D2D network. Moreover, the result of greedy
optimal D2D pairing and channel allocation. The detailed based algorithm is slightly lower than that of squareIMP
simulation parameters are presented in Table 2. based with a much lower computational complexity O(n2 ).
Besides, the matching result of greedy based algorithm is
TABLE 2. Simulation Parameters better than that of random algorithm. Therefore, squareIMP
algorithm can achieve a higher rate requirements, while the
greedy algorithm is a practical method with stringent latency
requirement.
FIGURE 6. Sum-rate versus number of contents in 4-DMWM based greedy FIGURE 8. Sum-rate of centralized and distributed algorithm by Content
and squareIMP algorithm with B=2. Caching scheme 1 with B=3, F=4.
FIGURE 7. Sum-rate of centralized and distributed algorithm with M=6, FIGURE 9. Sum-rate of centralized and distributed algorithm by Content
B=3, F=4. Caching scheme 2 with B=3, F=4.
algorithms in distributed system. The detailed algorithms are number of resource blocks is 3 and there are 4 content items.
present at Section IV. The sum-rate of different algorithms Each content is cached twice in the devices of the network
varies with the different cell sizes. In Fig. 7, we compare in scheme 2. The sum-rate also increases with the num-
the performance of sum-rate among the centralized and dis- ber of devices in different algorithms with different caching
tributed algorithms with different cell sizes when we set schemes. The simulation result of both the synchronous and
M = 6, B = 2, F = 3. Observed from Fig. 7, the sum- asynchronous distributed algorithm are lower than that of
rate decreases with the radius of the circular cell. Fig. 7 also centralized 4-DMWM based algorithm. It is obvious that
reveals that these centralized and distributed algorithms have the sum-rate of BP based distributed synchronous algorithm
similar sum-rate when the radius is larger than a certain size. is lower than the centralized greedy algorithm. However,
We assume that the file size is 120 kbits. Given a larger size of the complexity of distributed synchronous algorithm is only
file, long transmission time is required to ensure the reliable O(n2 ) for each device in a iteration. Moreover, the sum-rate of
transmission. distributed asynchronous algorithm is close to the centralized
Two different content caching schemes are used in these greedy algorithm and the computational complexity is only
algorithms, as shown in Fig. 8 and Fig. 9. We fix the buffer O(n2 ) for each device. In our simulation, we consider the
size as 2 for each devices in the simulation of scheme 1. channel estimation is perfect. The optimality of the pro-
However, in scheme 2, the same amount of contents are posed algorithm relies on the assumption of perfect channel
cached in different numbers of devices. In this case, the estimation.
VI. CONCLUSIONS [10] Y.-W. Lu, W. Chen, and H. V. Poor, ‘‘Content pushing based on physical
In this paper, we studied the caching based optimal D2D layer multicasting and request delay information,’’ in Proc. IEEE Global
Commun. Conf. (GLOBECOM), Washington, DC, USA, 2016.
pairing and channel allocation problem. This joint opti- [11] B. Bai, W. Chen, K. B. Letaief, and Z. Cao, ‘‘Diversity-multiplexing
mization problem is formulated as a weighted 4-uniform tradeoff in OFDMA systems: An H-matching approach,’’ IEEE
hypergraph model, where the 4-DMWM is the optimal solu- Trans. Wireless Commun., vol. 10, no. 11, pp. 3675–3687,
Nov. 2011.
tion. However, the 4-DMWM problem is NP-hard. Under [12] M. Maddah-Ali and U. Niesen, ‘‘Decentralized coded caching attains
the control of the BS, 4-DMWM based squareIMP and order-optimal memory-rate tradeoff,’’ IEEE/ACM Trans. Netw., vol. 23,
greedy algorithm was proposed to approach the optimal no. 4, pp. 1029–1040, Aug. 2015.
[13] B. Bai, W. Chen, K. B. Letaief, and Z.-G. Cao, ‘‘Distributed WRBG
solution with low-complexity. The squareIMP algorithm matching approach for multiflow two-way D2D networks,’’ IEEE Trans.
achieves a better performance approximate to the optimal Wireless Commun., vol. 15, no. 4, pp. 2925–2939, Apr. 2016.
approach with the complexity of O(n5 ). The greedy algo- [14] J. Liu, B. Bai, J. Zhang, and K. B. Letaief, ‘‘Content caching at the
rithm, however, is much more useful in practical caching wireless network edge: A distributed algorithm via belief propagation,’’
in Proc. IEEE Int. Conf. Commun. (ICC), Kuala Lumpur, Malaysia,
based D2D communications with stringent latency require- May 2016, pp. 1–6.
ment. In the distributed system, we proposed algorithms [15] B. Bai, W. Chen, K. B. Letaief, and Z.-G. Cao, ‘‘Low complexity outage
in the scenarios of synchronous and asynchronous, respec- optimal distributed channel allocation for vehicle-to-vehicle communi-
cations,’’ IEEE J. Sel. Areas Commun., vol. 29, no. 1, pp. 161–172,
tively. The sum-rate of distributed asynchronous algorithm Jan. 2011.
is closer to the centralized greedy algorithm with the com- [16] L. Militano, G. Araniti, M. Condoluci, I. Farris, and A. Iera, ‘‘Device-to-
plexity of O(n2 ) for each device. Moreover, the simula- device communications for 5G Internet of Things,’’ EAI Endorsed Trans.
Internet Things, vol. 1, no. 1, pp. 1–15, Oct. 2015.
tion result of BP based distributed synchronous algorithm [17] B. Bai, L. Wang, Z. Han, W. Chen, and T. Svensson, ‘‘Caching based
is slightly lower than the centralized greedy algorithm in socially-aware D2D communications in wireless content delivery net-
the distributed synchronous system and the complexity is works: A hypergraph framework,’’ IEEE Wireless Commun., vol. 23,
no. 4, pp. 74–81, Aug. 2016.
O(n2 ) per iteration for each device. The distributed algorithms [18] J.-H. Tang, T. Q. S. Quek, and W. P. Tay, ‘‘Joint resource segmentation and
have very low complexity in each devices. Therefore, the transmission rate adaptation in Cloud RAN with Caching as a Service,’’
algorithms of this paper can be used in different scenarios in Proc. SPAWC, Jul. 2016, pp. 1–6.
to solve the optimal D2D pairing and channel allocation [19] R. Tandon and O. Simeone, ‘‘Cloud-aided wireless networks
with edge caching: Fundamental latency trade-offs in fog radio
problem. access networks,’’ in Proc. Int. Symp.Inf. Theory (ISIT), Jul. 2016,
pp. 2029–2033.
[20] L. Miao, B. Bai, and W. Chen, ‘‘4-DMWM approach for caching based
Acknowledgment optimal D2D pairing and channel allocation,’’ submitted to Proc. ICC
Part of this work has been submitted to IEEE ICC 2017. 2017.
[21] P. Berman, ‘‘A d/2 approximation for maximum weight independent set
in d-claw free graphs,’’ Nordic J. Comput., vol. 7, no. 3, pp. 178–184,
2000.
REFERENCES [22] D. E. Drake and S. Hougardy, ‘‘A simple approximation algorithm for
[1] C.-H. Yu, K. Doppler, C. B. Ribeiro, and O. Tirkkonen, ‘‘Resource the weighted matching problem,’’ Inf. Process. Lett., vol. 85, no. 4,
sharing optimization for device-to-device communication underlaying pp. 211–213, Feb. 2003.
cellular networks,’’ IEEE Trans. Wireless Commun., vol. 10, no. 8, [23] B. Bai, W. Chen, Z. Cao, and K. B. Letaief, ‘‘Max-matching diversity in
pp. 2752–2763, Aug. 2011. OFDMA systems,’’ IEEE Trans. Commun., vol. 58, no. 4, pp. 1161–1171,
[2] C. Xu et al., ‘‘Efficiency resource allocation for device-to-device underlay Apr. 2010.
communication systems: A reverse iterative combinatorial auction based [24] S. Sanghavi, D. Malioutov, and A. Willsky, ‘‘Belief propagation and LP
approach,’’ IEEE J. Sel. Areas Commun., vol. 31, no. 9, pp. 348–358, relaxation for weighted matching in general graphs,’’ IEEE Trans. Inf.
Sep. 2013. Theory, vol. 57, no. 4, pp. 2203–2212, Apr. 2011.
[3] X. Wang, M. Chen, T. Taleb, A. Ksentini, and V. C. M. Leung, [25] F. R. Kschischang, B. J. Frey, and H.-A. Loeliger, ‘‘Factor graphs and
‘‘Cache in the air: Exploiting content caching and delivery techniques the sum-product algorithm,’’ IEEE Trans. Inf. Theory, vol. 47, no. 2,
for 5G systems,’’ IEEE Commun. Mag., vol. 52, no. 2, pp. 131–139, pp. 498–519, Feb. 2001.
Feb. 2014. [26] Evolved Universal Terrestrial Radio Access: Physical Channels Modula-
[4] U. Niesen, D. Shah, and G. W. Wornell, ‘‘Caching in wireless networks,’’ tion (R9), document TS 36.211 V9.0.0, 3GPP, 2009.
IEEE Trans. Inf. Theory, vol. 58, no. 10, pp. 6524–6540, Oct. 2012. [27] Air Interface for Fixed and Mobile Broadband Wireless Access Systems,
[5] M.-Y. Ji, G. Caire, and A. F. Molisch, ‘‘Fundamental limits of caching IEEE Standard 802.16-2009, 2009.
in wireless D2D networks,’’ IEEE Trans. Inf. Theory, vol. 62, no. 2, [28] B. Bai, W. Chen, K. B. Letaief, and Z.-G. Cao, ‘‘Outage exponent:
pp. 849–869, Feb. 2016. A unified performance metric for parallel fading channels,’’ IEEE Trans.
[6] K. Shanmugam, N. Golrezaei, A. G. Dimakis, A. F. Molisch, and Inf. Theory, vol. 59, no. 3, pp. 1657–1677, Mar. 2013.
G. Caire, ‘‘FemtoCaching: Wireless content delivery through dis- [29] B. Bai, W. Chen, K. B. Letaief, and Z. Cao, ‘‘A unified matching frame-
tributed caching helpers,’’ IEEE Trans. Inf. Theory, vol. 59, no. 12, work for multi-flow decode-and-forward cooperative networks,’’ IEEE
pp. 8402–8413, Dec. 2013. J. Sel. Areas Commun., vol. 30, no. 2, pp. 397–406, Feb. 2012.
[7] E. Baştuǧ, M. Bennis, M. Kountouris, and M. Debbah, ‘‘Cache-enabled [30] S. Rangan and R. Madan, ‘‘Belief propagation methods for inter-
small cell networks: Modeling and tradeoffs,’’ EURASIP J. Wireless cell interference coordination,’’ in Proc. IEEE INFOCOM, Apr. 2011,
Commun. Netw., pp. 1–11, Feb. 2015. pp. 2543–2551.
[8] M.-Y. Ji, G. Caire, and A. F. Molisch, ‘‘Wireless device-to-device caching [31] P. Bianchi, W. Hachem, and F. Iutzeler, ‘‘A coordinate descent primal-dual
networks: Basic principles and system performance,’’ IEEE J. Sel. Areas algorithm and application to distributed asynchronous optimization,’’
Commun., vol. 34, no. 1, pp. 176–189, Jul. 2015. IEEE Trans. Autom. Control, vol. 61, no. 10, pp. 2947–2957, Oct. 2016.
[9] W. Chen and H. V. Poor, ‘‘Joint pushing and caching with a finite [32] Y. Li, X. F. Lei, P. Z. Fan, and D. G. Chen, ‘‘An SCMA-based uplink
receiver buffer: Optimal policies and throughput analysis,’’ in Proc. inter-cell interference cancellation technique for 5G wireless systems,’’
IEEE Int. Conf. Commun. (ICC), Kuala Lumpur, Malaysia, May 2016, in Proc. WCSP, Nanjing, China, Oct. 2015, pp. 1–5.
pp. 1–6.
LU MIAO received the B.S. degree in communica- WEI CHEN (S’05–M’07–SM’13) received the
tion engineering electronic engineering from Jilin B.S. degree (Hons.) in operations research and
University, Changchun, China, in 2015. She is cur- the Ph.D. degree (Hons.) in electronic engineering
rently pursuing the master’s degree in electronic from Tsinghua University, Beijing, China, in 2002
engineering from Tsinghua University, Beijing, and 2007, respectively. From 2005 to 2007, he was
China. Her research interests include D2D caching a Visiting Research Staff with The Hong Kong
and maximum weighted matching problem. University of Science and Technology. Since 2007,
he has been with the Department of Electronic
Engineering, Tsinghua University, where he has
been prompted to Full Professor since 2012, as a
special case of early promotion. After the human resource reform in Tsinghua
University, he is elected as a Tenured Full Professor of the new research and
teaching track in 2015. He also serves as an University Council Member, the
Director of the Academic Degree Administration Office, Tsinghua Univer-
sity, and the Deputy Head of the EE Department.
He visited the University of Southampton, U.K., in 2010, Telecom
BO BAI (S’09–M’11) received the B.S. degree ParisTech, France, in 2014, and Princeton University, in 2015 and in 2016.
(Hons.) from the Department of Communica- He was supported by the National 973 Youth Project, the NSFC Excellent
tion Engineering, Xidian University, X’ian, China, Young Investigator Project, the 10000-Talent Program, the New Century
in 2004, and the Ph.D. degree from the Department Talent Program of the Ministry of Education, and the Beijing Nova Program.
of Electronic Engineering, Tsinghua University, He holds guest/chair professor positions of some Universities in China.
in 2010, Beijing, China. From 2009 to 2012, he His research interests are in the areas of wireless communications and
was a Visiting Research Staff (a Research Assis- information theory. He is a member of the Beijing Youth Federation. He
tant from 2009 to 2010 and a Research Associate served as a Tutorial Co-Chair of the IEEE ICC 2013, a TPC Co-Chair of
from 2010 to 2012) with the Department of Elec- the IEEE VTC 2011-Spring, and the Symposium Co-Chair of the IEEE ICC,
tronic and Computer Engineering, The Hong Kong ICCC, CCNC, Chinacom, and WOCC. He serves as an Editor of the IEEE
University of Science and Technology. He is currently an Assistant Professor TRANSACTIONS ON COMMUNICATIONS, the IEEE TRANSACTIONS EDUCATION, the
with the Department of Electronic Engineering, Tsinghua University. He has IEEE WIRELESS COMMUNICATIONS LETTERS, the IET Communications, China
also obtained the support from Backbone Talents Supporting Project of Communications, a Co-Chair of the Communications Theory Symposium in
Tsinghua University. In 2015 and 2016, he visited the City University of the IEEE Globecom 2017, and a Vice Director of the Youth Committee of
Hong Kong as a Senior Research Fellow. He received the Honor of Outstand- the China Institute of Communications.
ing Graduates of Shaanxi Province. He also received the Honor of Young Dr. Chen was a recipient of the First Prize of the 14th Henry Fok
Academic Talent of Electronic Engineering in Tsinghua University. Ying-Tung Young Faculty Award, the Yi-Sheng Mao Beijing Youth Sci-
He has authored over 60 papers in major IEEE journals and flagship ence & Technology Award, the 2010 IEEE Comsoc Asia Pacific Board Best
conferences. His research interests include hot topics in wireless communica- Young Researcher Award, the 2009 IEEE Marconi Prize Paper Award, the
tions and networking, fog/cloud-RAN, network big data, and random graph. 2015 CIE information theory new star award, the Best Paper Awards at the
He is the IEEE ComSoc Member. He received the Student Travel Grant at IEEE ICC 2006, the IEEE IWCLD 2007, the IEEE SmartGirdComm 2012,
the IEEE Globecom09. He was also invited as the Young Scientist Speaker and the First Prize in the 7th Young Faculty Teaching Competition in Beijing.
at the IEEE TTM 2011. He was a recipient of the IEEE ICC 2016 Best Paper He received the 2011 Tsinghua Raising Academic Star Award and the
Award. He served in the Wireless Communications Technical Committee and 2012 Tsinghua Teaching Excellence Award. He holds the honorary titles of
the Signal Processing and Communications Electronics Technical Commit- Beijing Outstanding Teacher and Beijing Outstanding Young Talent. He is
tee. He has served as a TPC Member of several IEEE ComSoc conferences, the Champion of the First National Young Faculty Teaching Competition,
such as ICC, Globecom, WCNC, VTC, ICCC, and ICCVE. He has also and a winner of National May 1st Medal.
served as a Reviewer for a number of major IEEE journals and conferences.