This article has been accepted for inclusion in a future issue of this journal.

Content is final as presented, with the exception of pagination.

IEEE SYSTEMS JOURNAL 1

Supporting Mobile Cloud Computing in Smart Cities

via Randomized Algorithms
Daniela Mazza, Adela Pages-Bernaus, Daniele Tarchi, Senior Member, IEEE, Angel Alejandro Juan,
and Giovanni Emanuele Corazza, Senior Member, IEEE

AbstractSmart cities represent rich and dynamic environ- on the environment, cannot be neglected in a sustainable city
ments in which a multitude of smart mobile devices (SMDs) in- management [5].
teract among them by sharing data. SMDs require from fast access The increasing trends of both urbanization and mobile-
to online services, but they offer limited computing capabilities
and battery lifetime. SMDs make frequent use of computation of- connected people have driven the society toward the definition of
floading, delegating computing-intensive tasks to the cloud instead a geographic information system for smart cities that is charac-
of performing them locally. In such a large-scale and dynamic terized by the presence of a multitude of smart devices, sensors,
environment, there might be thousands of SMDs simultaneously and processing nodes aiming to improve urban services.
executing processes and, therefore, competing for the allotment of Smart mobile devices (SMDs) are becoming an essential part
remote resources. This arises the need for a smart allocation of these
resources. Accordingly, this paper proposes a biased-randomized of social life and the most effective and convenient communi-
algorithm to support efficient and fast link selection. This algo- cation tool, which is not bounded in time nor place. In recent
rithm is able to provide real-time near-optimal solutions that years, the increasing request of ubiquitous services has also led
outperform solutions obtained through existing greedy heuristics. to a sharp increase in the so-called cloud computing, where re-
Furthermore, it overcomes the responsiveness limitations of exact mote servers can be exploited for remote storage and processing
optimization methods.
services. This is even more important in smart city scenarios:
Index TermsBiased-randomized algorithms, computation of- due to their complexity related to the number of users, hetero-
floading, HetNets, mobile cloud computing (MCC), smart city. geneous services, and specific requirements, smart cities can
significantly benefit from a cloud computing infrastructure as
I. INTRODUCTION it can be used to improve system performance and the battery
life of user devices through decreased work loads. Indeed, it
CCORDING to the flagship publication of the United Na-
A tions World Urbanization Prospect [1], more than one
half of the world population is currently living in urban areas,
is through information and communication technologies that
smart cities are truly turning smart [6], in particular by exploit-
ing SMDs, which together with cloud computing constitute the
and it is expected that about 70% of global population will be mobile cloud computing (MCC) framework [7]. A third element
city inhabitants by 2050. At the same time, the Visual Net- is the wireless network, allowing to link resource-constrained
working Index [2] outlines that mobile-connected devices have devices to centralized large data centers located inside the cloud.
already outnumbered the amount of people in the world. Includ- As suggested by Batty, to understand cities, we must view
ing machine-to-machine modules, there will be over 10 billion them not simply as places in space but as systems of networks
of such devices by 2018. Furthermore, an 11-fold increase in the and flows [8]. Thus, we can consider the MCC as a framework
overall mobile data traffic over the next five years is expected. that is the technological nervous system allowing the networks
Innovative smart city projects have been adopted in the po- and flows of the city to achieve a better urban way of life.
litical agenda of many governments as a key program to en- Moreover, the pervasiveness of wireless technologies has led to
able a vision where municipalities can use technology to meet the presence of heterogeneous networks where multiple types
sustainability goals, boost local economies, and improve ur- of access nodes operate simultaneously in the same city area
ban services. These projects focus on strengthening network [9], [10].
infrastructures, optimizing traffic, and transportation flows, de- In this context, one of the main challenges is to provide solu-
creasing energy consumption, and supplying innovative ser- tions able to jointly optimize the activities of data transfer (by
vices [3], [4]. In addition, a green concept of technology, in- exploiting wireless heterogeneous networks or HetNets) and
tended to mitigate and reverse the effects of human activity data processing (by delegating computing-intensive tasks to the
cloud in the MCC framework) [11], [12]. This strategy, com-
Manuscript received February 19, 2016; revised May 10, 2016; accepted May
28, 2016. This work was supported in part by the Spanish Ministry of Economy monly called cyberforaging or computation offloading, allows
and Competitiveness under Grant TRA2013-48180-C3-P and FEDER. us to overcome SMD limitations concerning limited battery
D. Mazza, D. Tarchi, and G. E. Corazza are with the Department of Electri- power and computational capacities. Moreover, it plays a key
cal, Electronic, and Information Engineering, University of Bologna, 40126
Bologna, Italy (e-mail: daniela.mazza6@unibo.it; daniele.tarchi@unibo.it; role in a smart environment where wireless communication is of
giovanni.corazza@unibo.it). utmost importance, particularly in mobility and traffic control
A. Pages-Bernaus and A. A. Juan are with the Department of Computer domains [13].
ScienceIN3, Open University of Catalonia, 08018 Barcelona, Spain (e-mail:
apagesb@uoc.edu; ajuanp@uoc.edu). Although computation offloading can significantly increase
Digital Object Identifier 10.1109/JSYST.2016.2578358 data-processing capabilities of mobile users, it is challenging

1937-9234 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications standards/publications/rights/index.html for more information.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

2 IEEE SYSTEMS JOURNAL

to achieve an efficient coordination among the entire set of scribe the problem and the optimization model in Section III. In
requesting devices. In fact, and in order to make the operation Section IV, the biased randomization techniques are described.
possible, the computation offloading entails data transmission Then, in Section V, we present the numerical results. Finally, in
among different devices and the cloud. If a large number of Section VI, the main findings and conclusions are highlighted.
citizens use the wireless resource to delegate computation to the
cloud, the efficiency of the access node can be severely affected:
its channel capacity has to be shared among all the devices, II. RELATED WORKS
leading to a reduced throughput for each user. This can lead to The cyberforaging mechanism for achieving efficient compu-
a crucial increase in energy usage and data transmission times, tation offloading for MCC has been discussed in the literature.
representing an inefficient use of offloading operations [14]. In [14], a game-theoretic approach is proposed, showing that
In this paper, we present a probabilistic algorithm based on the game always admits a Nash equilibrium. In [10], the authors
biased-randomization techniques [15] to solve the link selection propose a scheme for optimizing the quality of service (QoS)-
problem. In this problem setting, the most promising node allow- aware energy efficiency of the Base Stations composing the
ing the potential increase in system efficiency has to be selected. environment (macro and small cells), and in [17], the authors
The biased-randomization techniques work by introducing a bi- investigate an energy-efficient offloading policy for transcod-
ased or oriented random effect on the possible solutions of a ing as a service in a generic mobile cloud system. The cloud
problem, allowing to choose the best solution from a set of pos- system consists of a dispatcher at the front end, and a set of
sible alternatives that are close to the global optimal. Hence, a service engines at the back end. In [18], the authors exploit
set of alternative solutions is generated in short computational opportunistic communications to facilitate information dissem-
times. ination in the emerging mobile social networks. They propose
We will compare the solution provided by our probabilistic three algorithms for selecting the target set with only k users,
method with the optimal solution provided by exact methods. for minimizing the mobile data traffic over cellular networks.
However, these methods cannot be used in practice, since they In the literature, there are several works focused on the of-
require excessive computation times. Furthermore, we com- floading operation from the point of view of a single mobile user,
pare our approach to a greedy algorithm [16] that implements e.g., in [19] and [20]. In addition, the focus of several works con-
a selfish behavior in allocating resources to users. Such three cerning computation offloading is related to the development of
techniquesthe proposed probabilistic algorithm, the greedy application models aiming to extract offloading friendly parts of
heuristic, and the exact methodare based on the definition codes from existing applications [21][23]. In [24], the authors
of a proper objective function taking into account different re- present a dynamic offloading algorithm based on Lyapunov op-
quirements and characteristics of the considered scenario. In timization for selecting the software components to be remotely
particular, this paper will focus on energy consumption and executed given an available wireless network connectivity.
time delays as key performance indicators for comparing the However, none of the above explicitly considers the interac-
performance of the three approaches. tion among users. This paper addresses this problem by con-
The proposed randomized algorithm allows us to achieve sim- sidering entire populations instead of each user individually,
ilar solutions to those provided by the optimal method, although aiming at reducing overall community or social costs. In con-
in much less computing time. Moreover, the proposed random- trast, in [16], the same authors proposed a utility function model
ized algorithm clearly outperforms the greedy heuristic while derived from economics, in which the greedy heuristic summa-
employing similar computing times once it is parallelized, given rized in Section IV-B was used for minimizing the collective cost
that the modifications added do not increase the computational for a community in mobile clouds. Our approach in this paper is
complexity of the greedy heuristic. These characteristics make focused on a global perspective, corresponding to a centralized
it an efficient alternative for real-life applications. vision that is more effective for a community. Hereby, the same
All in all, in this paper, we propose a model establishing the application (viewed as an entity made out of data and operation)
relationship among all the devices (the cloud infrastructure, Het- is performed by many users.
Net nodes, and SMDs), both in terms of computing and commu- With respect to biased randomization techniques, probabilis-
nication. From this model, a cost function is derived. The goal is tic algorithmssimilar to the one presented herehave been
to minimize the overall energy consumption by selecting the best widely used to solve many combinatorial optimization prob-
communication link for each SMD. We show that this cost func- lems such as sequencing and scheduling problems [25], vehicle
tion cannot be optimally solved promptly, given that we have to routing problems [26], quadratic assignment problems [27], lo-
apply it in real time. Therefore, we present a biased-randomized cation and layout problems [28], covering, clustering, packing,
algorithm that allows us to find a near-optimal solution in short and partitions problems [29].
computing times. Thus, our proposed approach allows us to In particular, as described in [30], greedy randomized adap-
reach near-optimal solutions in real time. Moreover, we show tive search procedure (GRASP)-like algorithms have been
that it outclasses a greedy heuristic solution, used as a bench- applied to solve a wide set of problems, among others: schedul-
mark, presented by the same authors in [16], and summarized ing, routing, logic, partitioning, location, graph theory, as-
in this paper for completeness. signment, manufacturing, transportation, telecommunications,
The remainder of this paper is organized as follows. We biology and related fields, automatic drawing, power systems,
first analyze the problem-related works in Section II and de- and very large scale integration design.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

MAZZA et al.: SUPPORTING MOBILE CLOUD COMPUTING IN SMART CITIES VIA RANDOMIZED ALGORITHMS 3

deployment of relays, distributed antennas, and small cellular

base stations (e.g., microcells, picocells, and femtocells). These
deployments have occurred both indoors (e.g., in residential
homes and offices) and outdoors (e.g., in amusement parks or
traffic intersections).
Such network deployment is typically composed of a mix
of low-power nodes underlying the conventional homogeneous
macrocell network. By deploying additional small cells within
the local-area range and bringing the network closer to the users,
overall network capacities can be significantly boosted through
a better reuse of the spatial resources.
Inspired by the attractive features and the potential advantages
of the HetNets, their development has gained much momentum
in the wireless industry and research communities in recent
years.
The distributed smart city scenario is also an environment
where the SMDs can choose and exploit different available
RATs for transmitting the application data and for offloading
toward the cloud. However, the application can be computed lo-
Fig. 1. Reference scenario with multiple SMDs in a multi-RAT environment. cally by the requesting SMD if this operation is not convenient.
The proposed model focuses on direct connections between a
RAT and multiple SMDs, considering the availability of direct
Regarding the use of biased, or skewed, randomization, the interaction in a pervasive environment (note that multihop links
SimoRouting Generalized Clarke and Wrights Savings with are not taken into account in this paper). Furthermore, we con-
Cache and Splitting algorithm is proposed in [15] for solving sider energy savings from the user-perspective, differently from
the capacitated vehicle routing problem. It combines a biased other application-focused works, which include the energy con-
randomization process with a base heuristic. A geometric dis- sumed by the cloud when sending the resultant data to the SMDs
tribution is used to randomize the constructive process while [34]. In the following, the entities involved in this scenario are
keeping the logic behind the heuristic. Similarly, in [31], the introduced.
authors developed the RandSHARP algorithm for solving the
arc routing problem. This algorithm combines a savings-based A. Cloud
heuristic for the arc routing problem with a biased randomiza- The cloud infrastructure Ccc refers to the presence of a remote
tion process also guided by a geometric distribution. Likewise, cloud computing infrastructure with a huge amount of storage
in [32], the authors propose the Iterated Local Search Efficient, space and computing power. It allows the citizens to interact
Simple, and Parameter-Free (ILS-ESP) algorithm for solving remotely, e.g., for accessing open data delivered by the public
the permutation flow-Shop problem. The ILS-ESP uses an iter- administration. The cloud infrastructure is often used for deliv-
ated local search framework and combines the Nawaz, Enscore, ering the computing processes to remote clusters, leading to a
and Ham heuristic [33] with a biased randomization process higher computing power, and/or for storing big amounts of data.
guided by a descending triangular distribution. The centralized cloud allows to reduce the computational time
by exploiting powerful processing units, but it could suffer from
III. PROBLEM DESCRIPTION AND OPTIMIZATION MODEL the distribution latency, due to the data transfer from the users
In this section, the optimization problem is introduced. The to the cloud and vice versa, the congestion, due to the multiple
focus is on an urban area with a pervasive wireless coverage, users exploitation, and the resiliency, due to the presence of a
where several SMDs require cloud services. Each SMD can single performing infrastructure leading to the single-point-of-
connect to the cloud after selecting one of the available HetNet failure issue.
radio access technologies (RATs), as shown in Fig. 1. The cloud entity Ccc is characterized by its own computa-
The SMDs aim at delegating storage and computing tasks tional speed, fcc . The storage availability of Ccc can be con-
to overcome their resource constraints and to decrease com- sidered infinite; thus, it does not represent a constraint in the
putational times. Beyond data storage, that is one of the most interaction with the SMDs. Hence, it is possible to write that the
common activities transferred to remote cloud infrastructures, Ccc is a function of fcc ,
it is now possible to distribute the whole computational load, or Ccc = Ccc (fcc ). (1)
just a fraction, to a remote unit.
Multiple RATs, such as IEEE 802.11 WLANs, mobile
B. HetNet nodes
WiMAX, HSPA+, and LTE, are being integrated to create a
heterogeneous wireless network. Different strategies can be The wireless HetNet infrastructure is composed of several
carried out in order to increase the network capacity, such as RAT nodes, characterized by different features.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

4 IEEE SYSTEMS JOURNAL

(1) BW : the nominal bandwidth of a certain communication Among different parameters that can affect the system perfor-
technology that is available for the SMDs. mance, we will focus on the energy consumption, which is very
(2) n: the number of devices connected to the RAT node. important when considering a scenario with mobile devices. The
(3) posRAT (cx , cy ): the spatial position of the RAT node, energy consumed by the jth SMD is composed by the energy
where cx and cy represent the coordinates. spent in transmitting the application data, the energy consumed
Thus, the generic RAT is a function depending on the above in idlewhile the application is computed on the cloud server
features and the energy consumed for the local computation.
Considering an overall number of N devices and a total num-
RAT = RAT(BW, n, posRAT (cx , cy )). (2) ber of M RAT nodes, the energy spent by the jth SMD for
offloading through the ith RAT is
C. Applications
Ptr j D Pid j O
Ei,j = + , i = 1, 2, . . . , M ; j = 1, 2, . . . , N.
Even if all the SMDs can simultaneously request the com- Si,j fcc
putation offloading of one or more applications, we focus on (7)
a scenario where each SMD request is restricted to a single In the case of a local computation at the jth SMD, its con-
application that is characterized by the number of operations sumed energy is
to be executed, O, and the amount of data to be exchanged, Pl j O
D. We neglect other parameters usually considered in more E0,j = , j = 1, 2, . . . , N (8)
fm d j
application-focused studies (e.g., Mobibyte [23]) because our
study is mainly based on the communication links optimization, where the index 0 stands for the local computation.
aside from the application partitioning. This simplification does In classical approaches, the SMD computing an application
not prevent the generalization of the system, since the general App has to evaluate the computation offloading benefit and se-
case in which a user requests several applications at a given lect the most convenient RAT aiming at minimizing the energy
time can be seen as aggregation of several single-application consumption of the user requesting the service. This decision is
requests. Hence, it is possible to express a generic application selfish or greedy in the sense that it does not take into account
App as a function of O and D, the current and future needs of other users.
In contrast to this, we propose to optimize the global energy
App = App(O, D). (3) consumption, assuming an environment where the devices of
the set N = {1, 2, . . . , j, . . . , N } are requesting to offload an
D. SMDs application using the nodes of set M = {0, 1, 2, . . . , i, . . . , M },
where the zeroth element represents the fictitious node related
Each SMD is characterized by the following features which
to the local computation. By defining xi,j , as a binary variable
influence the performance of the computation offloading.
that stands for the presence of a communication link between
(1) Ptr : the power consumed by the SMD to transmit data to
the SMDj and the RATi , and yj , as a binary variable accounting
the RAT node.
for the local computation performed by SMDj , it is possible to
(2) Pl : the power consumed by the SMD for local computa-
write the energy consumed by SMDj when offloading through
tion.
the RATi in (7), by exploiting (6), as
(3) Pid : the power consumed by the SMD in waiting mode  
while the computation is performed in the cloud. Ptr j D Pid j O
(4) fm d : the speed to perform the computation locally. Ei,j = xi,j BWi k d i , j }
+
n i log2 {1 + SNRj e
fcc
(5) SNR: the signal-to-noise ratio at the receiver side of the  
link between the SMD and the selected RAT node. Pl j O
+yj (9)
(6) posm d (cx , cy ): the spatial position of the SMD, where cx fm d j
and cy represent the coordinates.
where
Thus, for a generic SMD, we can write
1, if SMDj is assigned to RATi
xi,j =
SMD = SMD(Ptr , Pl , Pid , fm d , SNR, posm d (cx , cy )). (4) 0, otherwise

Given these entities, with the respective features, and after hav- 1, if SMDj computes App locally
yj =
ing defined as di,j the distance between the ith RAT and the 0, otherwise
jth SMD, we can resort to the Shannon formula for evaluating and ni is the number of SMDs connected to RATi , defined as
the throughput Si,j of the link between the ith RAT and the jth

SMD, where SNRj is a reference signal-to-noise ratio (SNR) ni = xi,l .

value at the receiver side and k is an attenuation factor due to lN
the signal propagation Rearranging the constant terms and the variable terms, it is
  possible to write
di,j = posRAT,i (cx , cy ) posm d,j (cx , cy ) (5)  

BWi   tr
Ei,j = xi,j Ki,j id
xi,l + Kj + yj E0,j (10)
Si,j = log2 1 + SNRj ek d i , j (6)
ni lN
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

MAZZA et al.: SUPPORTING MOBILE CLOUD COMPUTING IN SMART CITIES VIA RANDOMIZED ALGORITHMS 5

Algorithm 1 Greedy Link Selection Heuristic

Inputs:
M = {0, 1, 2, . . . , M } // M number of RAT nodes
N = {1, 2, . . . , N } // N number of SMDs
RATi = RAT(BWi , ni , posRAT,i (cx , cy ))i M \ {0}
App = App(O, D)
Output:
X = (xi,j )
Y = (yj )
Initialization :
RATi .n 0i M \ {0}
xi,j 0i Mj N
yj 0j N
Fig. 2. Scheme of the biased-randomization approach. for j = 1 to N do
SMDj = SMD(Ptr,j , Pl,j , Pid,j , fm d,j , SNRj , )
posm d,j (cx , cy ))
where the two constants are RATi .n RATi .n + 1i M \ {0}
tr Pt r j D calculateEi,j i M
Ki,j = BWi log 2{1+SNRj e k d i , j } choose xi,j yj s.t. Ei,j = min{El,j ; l M}
RATl .n RATl .n 1l M s.t. xl,j = 1
and
X xi,j
Pid j O Y yj
Kjid = .
fcc end for
Therefore, the optimization problem can be expressed as
 

min Z = xi , j Kitr, j xi , l + Kji d + yj E0 , j (11a) Algorithm 2 Biased-Randomization Algorithm

x,y
i M l N j N
j N Inputs:

M = {0, 1, 2, . . . , M } // M number of RAT nodes
s.t. xi , j + yj = 1, jN (11b)
N = {1, 2, . . . , N } // N number of SMDs
i M
RATi = RAT(BWi , ni , posRAT,i (cx , cy ))i M \ {0}
xi , j , yj {0, 1}, iM j N. (11c) App = App(O, D)
The objective function (11a) minimizes the overall energy Output:
spent, while ensuring that all devices are either assigned to a X = (xi,j )
RAT or leaving the computations to be done locally (11b). Y = (yj )
Initialization :
IV. OPTIMAL AND HEURISTIC SOLUTIONS d0,j d0
RATi .n ni i M \ {0}
In this section, we present different methods to find an optimal xi,j 0i Mj N
or near-optimal solution to the previously described optimiza- yj 0j N
tion problem. First, we use an off-the-shelf commercial solver for j = 1 to N do
able to provide optimal solutions. However, for real large-size SMDj = SMD(Ptr,j , Pl,j , Pid,j , fm d,j , SNRj ,
instances, it requires unsuitable computation times. Moreover, posm d,j (cx , cy ))
since we take into account mobile devices whose position can calculate di,j ; Ei,j i M
change, the objective function has to be solved repeatedly. Given sort di,j Ei,j ascending i M
that the selection of the RAT by the SMDs has to be done in a choose xi,j yj s.t. Ea,j = geoinv(Ei,j ; i M)
very short time, we propose to use a fast probabilistic procedure. X xi,j
Our algorithm makes use of biased (oriented) randomization, Y yj
and it allows us to obtain near-optimal solutions in real-time, end for
which represent a significant improvement over the previously
proposed greedy algorithm [16] summarized below.
In order to evaluate the effectiveness of our approach, the
performance is compared with the optimal solution. Note that approach assumes that all users requests are processed at the
the global optimum can only be found ex-post, i.e., once all same timeor, alternatively, that the allocation decision is post-
the SMDs requests have been set. Thus, since the global opti- poned until all requests have been receivedwhich in practice
mization needs to know the position and the requests of all the might require customers to wait a few seconds for their requests
SMDs, it cannot be used for online allocation of resources. This to be served.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

6 IEEE SYSTEMS JOURNAL

Fig. 3. Scenarios in case of 500 and 5000 SMDs, respectively, three Wi-Fi APs, and one LTE eNodeB.

In contrast, the heuristic method makes use of a greedy sorts each possible access node based on a self-calculated ob-
behavior, consisting on the selection of the most promising jective function. Since the requests of the SMDs appear in time
next step from a list of potential constructive movements [16]. sequence, the cost function is evaluated on the basis of the cur-
This allows us to find a good solution in real time, while each rent situation, i.e., considering only the previous requests. If the
agent tends to maximize his/her individual utility function. The offloading cost is lower than the cost for the local computation,
biased randomized method takes into account a solution from a the SMD will connect to the node which minimizes the cost
collective or social point of view by simultaneously considering function; otherwise, it will compute the application locally.
a set of users demands altogether instead of one users demand The selection of the ith node for connecting the jth SMD
at a time. modifies the values of the throughput Si,k and, consequently,
the energy Ei,k for the SMDs already connected with the same
A. Optimal Solution
node, i.e., for k = 1, 2, . . . , j 1. Thus, this strategy, reported
The mathematical model (11) can be seen as a particular case in Algorithm 1, does not take into account any forecast of future
of the quadratic semiassignment problem (QSAP) which is NP- connections, leading to a feasible but suboptimal solution.
hard [35]. A first approach to find the solution is to use the
global optimizer Branch-And-Reduce Optimization Navigator C. Biased-Randomized Algorithm
(BARON) [36]. This solver combines constraint propagation,
By exploiting biased-randomization techniques [15], the
interval analysis, and duality with advanced branch-and-bound
proposed probabilistic algorithm is able to find near-optimal
optimization concepts.
solutions in real time a few seconds or even milliseconds
Note that the QSAP solution not only needs to know the
if parallel strategies are employed. Thus, this approach out-
position of all the SMDs requests before assigning them to an
performs by far the solution provided by the greedy approach,
antenna, but it also takes longer time than available to quickly
while approaching the optimal solution in significant lower time
serve SMD requests, as shown in Section V. The computational
than the optimal method. The main idea behind the proposed
difficulty for solving the QSAP comes from the fact that the
approach is to introduce a slight modification in the greedy con-
objective function is quadratic, which justifies the use of fast
structive behavior. Hereby the heuristic logic is maintained but a
heuristic methods. Therefore, the optimal solver is only taken
certain degree of randomness is introduced. This random effect
as a reference to compare heuristic procedures (which can do the
is generated by the use of a skewed probability distribution: at
assignment dynamically or following a wait-and-go approach)
each step of the constructive process, a selection probability is
with an optimal solution from the performance point of view.
assigned to each potential movement, whereby the probability
is higher for the more promising movements.
B. Greedy Heuristic The use of biased-randomization significantly improves the
A heuristic algorithm can be used to solve the link selec- quality of the solutions generated by the original heuristic
tion problem following a greedy behavior [16]. If the offloading in different dimensions when considering social or collective
operation is advantageous with respect to the local computa- performance, reaching a solution closer to the optimal without
tion, the link selection scheme allows us to select the most needing to know all SMD positions.
promising next node from the list of the available ones. As Moreover, it is important to note that if a uniform probabil-
explained in [16], this list is completed for each SMD, which ity distribution would be used instead of a skewed one, this
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

MAZZA et al.: SUPPORTING MOBILE CLOUD COMPUTING IN SMART CITIES VIA RANDOMIZED ALGORITHMS 7

TABLE I
THROUGHPUT [MBPS]

N SMDs 500 1000 2000 5000

Avg. Dev. [%] Avg. Dev. [%] Avg. Dev. [%] Avg. Dev. [%]

Nearest Node 1.5693 50.26 0.7847 50.33 0.3926 50.36 0.1570 50.46
Greedy 3.1551 ref. 1.5799 ref. 0.7909 ref. 0.3169 ref.
Biased Rand. 3.1699 +0.47 1.5840 +0.26 0.8004 +1.20 0.3240 +2.24
Baron 3.1707 +0.50 1.5844 +0.29 0.7999 +1.14 0.3240 +2.22

TABLE II
ENERGY [W*S]

N SMDs 500 1000 2000 5000

Avg. Dev. [%] Avg. Dev. [%] Avg. Dev. [%] Avg. Dev. [%]

Local 22 500.0000 +445.26 22 500.0000 +173.19 22 500.0000 +36.80 22 500.0000 2.17

Nearest Node 4609.3141 +11.70 9819.6194 +19.23 18 832.6714 +14.50 46 060.3241 +100.26
Greedy 4126.5039 ref. 8236.0653 ref. 16 447.8885 ref. 22 999.8480 ref.
Biased Rand. 4106.4181 0.49 8239.5161 +0.04 14 848.2169 9.73 19 711.6979 14.30
Baron 4104.4986 0.53 8210.9327 -0.31 14 716.6647 10.53 19347.2418 15.88

TABLE III
TIME FOR APPLICATION COMPUTATION [S]

N SMDs 500 1000 2000 5000

Avg. Dev. [%] Avg. Dev. [%] Avg. Dev. [%] Avg. Dev. [%]

Local 25 000.0000 +685.69 25 000.0000 +294.13 25 000.0000 +97.47 25 000.0000 +19.01

Nearest Node 3553.3186 +11.67 7561.2457 +19.20 14 494.3626 +14.49 35 438.7108 +68.70
Greedy 3181.9261 ref. 6343.1271 ref. 12 659.9142 ref. 21 007.3416 ref.
Biased Rand. 3166.4755 0.49 6345.7816 +0.04 14 003.7438 +10.61 20 918.6292 0.42
Baron 3164.9989 0.53 6323.7944 0.30 13 745.0152 +8.57 20 432.3306 2.74

TABLE IV in the sorted list. That is, elements at the top of the list have a
TIME FOR COMPLETING THE ALGORITHMS [S]
higher probability of being selected than those at the bottom,
whereas all elements are potentially eligible. This allows not
N SMDs 500 1000 2000 5000 only to avoid the problem of selecting the proper size of the
Greedy 0.41 0.85 1.65 4.07 restricted list, but we can also guarantee that the selection prob-
Biased Rand. 3.34 6.64 11.74 28.44 abilities are proportional to the position of each element in the
Baron stopped after 1000 s list.
Each time the randomized algorithm is executed, a new prob-
abilistic solution is obtained (see Fig. 2). Some of these solutions
improvement would rarely occur since the logic behind the con- allow us to improve those obtained through the greedy heuristic.
structive heuristic would be destroyed and the process would be Moreover, the proposed approach allows us to achieve different
random but not correctly oriented. solutions depending on some of the inputs of the model (i.e.,
To avoid losing the logic behind the heuristic, GRASP meta- throughput, energy, latency, etc.).
heuristics [37] propose the consideration of a restricted list Focusing on the selected problem, the biased-randomization
of candidatesi.e., a sublist including just some of the most algorithm for the link selection consists in a skewed criteria
promising movements, corresponding to those at the top of the based on the objective function defined in (11), which has been
listand then apply a uniform randomization to the order in also used for the greedy heuristic.
which the elements of that restricted list are selected. In this When there is a new SMD request, the cost function Ei,j
way, a deterministic procedure is transformed into a random- is evaluated for every possible RAT node (also considering the
ized algorithmwhich can be encapsulated into a multistart fictitious node 0 related to the local computation). The obtained
processwhile most of the logic or common sense behind the values are sorted in such a way that the probabilities of being
original heuristic is still respected. selected depend not only on the cost function Ei,j , but also on
The proposed biased-randomization approach goes one step the distances di,j of the device from the RAT nodes and on the
ahead. Instead of restricting the list of candidates, different se- number of devices ni that potentially can be connected to each
lection probabilities are assigned to each potential movement RAT node.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

8 IEEE SYSTEMS JOURNAL

Fig. 4. Comparison of the performance in terms of throughput for different SMDs number. (a) Throughput for 1000 SMDs. (b) Throughput for 2000 SMDs. (c)
Throughput for 5000 SMDs.

To improve the link selection, the criteria for sorting the SMDs characteristics) are also taken into account for compar-
choice list takes into account the product di,j Ei,j , with i M, ison purposes. However, we choose to compare all the values
instead of Ei,j . The value of d0,j is adjusted empirically and with those obtained with the greedy algorithm as reference, due
depends on the overall number of SMDs which are expected to its strict relationship with the proposed biased-randomized
to interact in the system. This strategy is reported in Algo- approach. The throughput, the energy consumed by the SMDs,
rithm 2. Note that the modifications included in Algorithm 2 and the time employed for accomplishing the application are
do not increase the computational complexity with respect to evaluated and observed for each configuration. Also, different
Algorithm 1. scenarios are considered, each of them involving an increasing
number of SMDs: 500, 1000, 2000, and 5000.
V. NUMERICAL RESULTS Tables IIII show, respectively, the results related to the
In this section, a realistic test case is evaluated in order to com- throughput, the energy, and the time. The ex-post solu-
pare the performance of the different algorithms. A deployment tion has been computed with Baron under the Neos server
area equal to 1000 1000 m2 is considered, where one LTE (http://www.neos-server.org). The model was implemented with
eNodeB with a bandwidth of 100 MHz is positioned at point the mathematical modeling language GAMS. The other meth-
(500,500) and three Wi-Fi access points (AP) with a bandwidth ods were implemented in Matlab.
equal to 22 MHz are positioned at point (0,0), (500,999) and The results show that the biased-randomized algorithm
(1000,0). An attenuation coefficient k for the propagation equal clearly outperforms the link selection configuration obtained
to 103 has been considered. Furthermore, similarly to [16], with the greedy heuristic. Improvement of 2.24% in throughput,
the capacity constraints on the antennas is not considered, as- and reductions of 14.30% in energy consumption, and 0.42%
suming there is no limitation on the number of SMD per each in computational time can be observed. Moreover, the average
eNodeB/AP. energy and time values provided by the randomized algorithm
The location coordinates of each SMD in the deployment area are similar to the optimal/near-optimal solutions provided by
are sampled from a uniform distribution. In particular, we have Baron, which has been stopped after 1000 s of computational
chosen a controlled random number generation of the Mersenne time. This demonstrates that our probabilistic algorithm is able
Twister type and a seed equal to 1 (this pseudorandom number to provide near-optimal solutionssimilar to the ones provided
generator is implemented in MATLAB). Fig. 3 represents the by the exact solverin much lower computation times. The
region assuming 500 and 5000 SMDs, respectively. higher improvement in terms of energy is expected, since the
The parameters fm d , Pid , Ptr , Pl , and SN R depend on the objective function in (11a) is designed for minimizing the energy
selected mobile device. We considered a HP iPAQ PDA with a consumption.
400-MHz Intel XScale processor (fm d equal to 400 MHz) and Table IV shows indeed the time needed by the algorithms
the following values: Pl equal to 0.9 W, Pid equal to 0.3, Ptr for achieving a stable solution of the link selection problem;
equal to 1.3 W, and SNR equal to 30 dB. For the cloud server, such time is not directly related to the time needed for the
we suppose a computation speed fcs equal to 106 MHz [20]. application computation in Table III, while it is more related to
We have chosen an application which is accomplished through the ability of the algorithms to solve the problem in a fast way
a number of operation O equal to 107 and, if offloaded, needs a and, hence, its exploitation in a real-life dynamic scenario. On
data transfer D equal to 104 bits. the one hand, it has to be noted that the exact solver requires
The algorithms are compared in terms of throughput, en- computation times that cannot be employed in real-life practice.
ergy, and time needed for accomplishing the application. The On the other hand, the computing time employed by the biased-
no-connection configuration (i.e., all the SMDs compute lo- randomized algorithm is below a few seconds, which is still a
cally) and the nearest-node configuration (i.e., each SMD is little higher than the one employed by the greedy algorithm. This
connected to the closest RAT independently from the RATs and is due to the fact that the biased-randomized algorithm is run
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

MAZZA et al.: SUPPORTING MOBILE CLOUD COMPUTING IN SMART CITIES VIA RANDOMIZED ALGORITHMS 9

Fig. 5. Performance in terms of throughput, energy consumption, and latency for a variable number of devices. (a) Throughput. (b) Energy. (c) Latency.

Fig. 6. Link selection in case of 500 SMDs. Each gray indicates that the SMD is connected to the RAT marked with the same gray. (a) Nearest node configuration.
(b) Greedy heuristic configuration. (c) Biased-randomized configuration. (d) Ex-post optimal configuration.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

10 IEEE SYSTEMS JOURNAL

Fig. 7. Link selection in case of 5000 SMDs. Each gray indicates that the SMD is connected to the RAT marked with the same gray, whereas the black SMDs are
computing the application locally. (a) Nearest node configuration. (b) Greedy heuristic configuration. (c) Biased-randomized configuration. (d) Ex-post optimal
configuration.

several times in a multistart process. Nevertheless, it is still an Considering the average energy consumed by one SMD
acceptable computing time. Moreover, if necessary the biased- to perform the application, Fig. 5(b) shows that the biased-
randomized algorithm could be easily parallelizedby simply randomized algorithm clearly outperforms the greedy heuristic.
running each iteration of the multistart process in a different The same observation can be inferred from Fig. 5(c) for the
core or computerso its computing time would be the same as average time that a device needs to accomplish the application.
the one employed by the greedy algorithmi.e., in the order of Both the average energy and time values are very close to the
milliseconds. respective results of Baron. Thus, the biased-randomized al-
The results are visually compared in Figs. 4 and 5(a), where gorithm can be exploited by the SMDs to reach near-optimal
we can see that the performance of the biased-randomized configurations in real time, while providing solutions that tend
algorithm is very similar in terms of average throughput, com- to minimize the overall energy consumption. This is useful for
pared to Barons ex-post configuration. The greedy heuristic offloading applications which need a fast link selection decision,
is clearly outperformed. This means that the capacity of the for example, when the users are moving and the configuration
RAT nodes is well exploited in the biased-randomized algo- must be updated very often.
rithm. In Fig. 4, the average throughput values are shown Figs. 6 and 7 report the different configurations (of the de-
in a different scale to make possible the comparison of the vices placed in the observed area) for an overall number of
three algorithms. SMDs equal to 500 and 5000, respectively. Comparing these
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

MAZZA et al.: SUPPORTING MOBILE CLOUD COMPUTING IN SMART CITIES VIA RANDOMIZED ALGORITHMS 11

configurations, we can note that the case referring to the biased- [11] L. Lei, Z. Zhong, K. Zheng, J. Chen, and H. Meng, Challenges on wire-
randomized algorithm is very similar to the ex-post optimal less heterogeneous networks for mobile cloud computing, IEEE Wirel.
Commun., vol. 20, no. 3, pp. 3444, Jun. 2013.
solution, where the SMDs connected to the same RAT are clus- [12] A. U. R. Khan, M. Othman, S. A. Madani, and S. U. Khan, A survey of
tered. In particular, in these figures, the nodes represent SMDs mobile cloud computing application models, IEEE Commun. Surv. Tuts.,
connected to each eNodeB/AP of the same gray or, alternatively, vol. 16, no. 1, pp. 393413, 2014.
[13] S. Barbarossa, S. Sardellitti, and P. Di Lorenzo, Communicating while
performing local computation, represented as black dots. It is computing: Distributed mobile cloud computing over 5G heterogeneous
possible also to highlight that the biased-randomized approach networks, IEEE Signal Process. Mag., vol. 31, no. 6, pp. 4555,
allows to adapt the link selection based on the node density, Nov. 2014.
[14] X. Chen, Decentralized computation offloading game for mobile cloud
while the greedy approaches only take into account the nodes computing, IEEE Trans. Parallel Distrib. Syst., vol. 26, no. 4, pp. 974
density in a limited way. This confirms the socially aware 983, Apr. 2015.
optimization behavior of the proposed biased-randomized [15] A. A. Juan, J. Faulin, J. Jorba, D. Riera, D. Masip, and B. Barrios, On the
use of Monte Carlo simulation, cache and splitting techniques to improve
approach. the Clarke and Wright savings heuristics, J. Oper. Res. Soc., vol. 62,
no. 6, pp. 10851097, Jun. 2011.
[16] D. Mazza, D. Tarchi, and G. E. Corazza, A user-satisfaction based of-
VI. CONCLUSION floading technique for smart city applications, in Proc. IEEE Global
Commun. Conf., Austin, TX, USA, Dec. 2014, pp. 27832788.
In this paper, we have presented different strategies for an [17] W. Zhang, Y. Wen, and H.-H. Chen, Toward transcoding as a service:
energy-efficient offloading policy for green mobile cloud, IEEE Netw.,
efficient link selection in pervasive wireless environments. In vol. 28, no. 6, pp. 6773, Nov. 2014.
particular, we have proposed a novel randomized algorithm that [18] B. Han, P. Hui, V. S. A. Kumar, M. V. Marathe, J. Shao, and A. Srinivasan,
allows to consider a more global point of viewinstead of just Mobile data offloading through opportunistic communications and social
participation, IEEE Trans. Mobile Comput., vol. 11, no. 5, pp. 821834,
a greedy or individual oneduring the real-time resource allo- May. 2012.
cation. Thus, the mobile communication system is managed in [19] F. Liu et al., Gearing resource-poor mobile devices with powerful clouds:
a cooperative way, considering the QoS of the entire population Architectures, challenges, and applications, IEEE Wireless Commun.,
vol. 20, no. 3, pp. 1422, Jun. 2013.
of mobile users. This vision could be adopted for the provi- [20] K. Kumar and Y.-H. Lu, Cloud computing for mobile users: Can of-
sion of services in a smart city, improving performance and floading computation save energy? Computer, vol. 43, no. 4, pp. 5156,
well-being of the community through a social use of HetNets Apr. 2010.
[21] E. Cuervo et al., MAUI: Making smartphones last longer with code
and MCC. Furthermore, it can be seen as an enhancing method offload, in Proc. 8th Int. Conf. Mobile Syst., Appl. Serv., San Francisco,
from an ecofriendly perspective, since it aims to save SMDs en- CA, USA, Jun. 2010, pp. 4962.
ergy. Our approach is based on the use of biased-randomization [22] B.-G. Chun, S. Ihm, P. Maniatis, M. Naik, and A. Patti, CloneCloud:
Elastic execution between mobile device and cloud, in Proc. 6th Conf.
techniques, which have been used in the past to solve similar Comput. Syst., Salzburg, Austria, Apr. 2011, pp. 301314.
combinatorial optimization problems in the fields of logistics, [23] A. Ur et al., MobiByte: An application development model for mobile
transportation, and production. This work extends their use to cloud computing, J. Grid Comput., vol. 13, no. 4, pp. 605628, 2015.
[24] D. Huang, P. Wang, and D. Niyato, A dynamic offloading algorithm
the field of smart cities and mobile telecommunications. Some for mobile computing, IEEE Trans. Wireless Commun., vol. 11, no. 6,
numerical results contribute to illustrate the potential of the pp. 19911995, Jun. 2012.
proposed approach. [25] M. L. Pinedo, Scheduling: Theory, Algorithms, and Systems, 3rd ed.
New York, NY, USA: Springer, 2008.
[26] J. Caceres-Cruz, P. Arias, D. Guimarans, D. Riera, and A. A. Juan, Rich
vehicle routing problem: Survey, ACM Comput. Surv., vol. 47, no. 2,
REFERENCES pp. 132, Jan. 2015.
[27] E. M. Loiola, N. M. M. de Abreu, P. O. Boaventura-Netto, P. Hahn, and
[1] World Urbanization Prospects, Population Division, United Nations. T. Querido, A survey for the quadratic assignment problem, Eur. J. Oper.
(2014). [Online]. Available: http://esa.un.org/unpd/wup/ Res., vol. 176, no. 2, pp. 657690, Jan. 2007.
[2] Cloud and Mobile Network Traffic Forecast, VNI, Cisco. (Jun. [28] N. Mladenovic, J. Brimberg, P. Hansen, and J. A. Moreno-Perez, The
2016). [Online]. Available: http://www.cisco.com/c/en/us/solutions/ p-median problem: A survey of metaheuristic approaches, Eur. J. Oper.
service-provider/visual-networking-index-vni/index.html Res., vol. 179, no. 3, pp. 927939, Jun. 2007.
[3] C. Manville et al., Mapping smart cities in the EU, European Par- [29] A. A. Chaves and L. A. N. Lorena, Clustering search algorithm for the
liament, Tech. Rep., Jan. 2014. [Online]. Available: http://www.Euro capacitated centered clustering problem, Comput. Oper. Res., vol. 37, no.
parl.europa.eu/RegData/etudes/etudes/join/2014/507480/IPOL-ITRE_E 3, pp. 552558, Mar. 2010.
T(2014)507480_EN.pdf [30] P. Festa and M. G. C. Resende, An annotated bibliography of GRASP
[4] Scientific American, Ed., Designing the Urban Future: Smart Cities. New Part II: Applications, Int. Trans. Oper. Res., vol. 16, no. 2, pp. 131172,
York, NY, USA: Scientific Amer., 2014. Mar. 2009.
[5] C. Zhu, V. Leung, L. Shu, and E. Ngai, Green Internet of Things for [31] S. Gonzalez-Martn, A. A. Juan, D. Riera, Q. Castella, R. Munoz, and
smart world, IEEE Access, vol. 3, pp. 21512162, 2015. A. Perez, Development and assessment of the SHARP and RandSHARP
[6] M. Dohler, C. Ratti, J. Paraszczak, and G. Falconer, Smart cities, IEEE algorithms for the arc routing problem, AI Commun., vol. 25, no. 2,
Commun. Mag., vol. 51, no. 6, pp. 7071, Jun. 2013. pp. 173189, Apr. 2012.
[7] H. T. Dinh, C. Lee, D. Niyato, and P. Wang, A survey of mobile cloud [32] A. A. Juan, H. R. Lourenco, M. Mateo, R. Luo, and Q. Castella, Us-
computing: Architecture, applications, and approaches, Wireless Com- ing iterated local search for solving the flow-shop problem: Paralleliza-
mun. Mobile Comput., vol. 13, no. 18, pp. 15871611, Dec. 2013. tion, parametrization, and randomization issues, Int. Trans. Oper. Res.,
[8] B. Batty, The New Science of Cities. Cambridge, MA, USA: MIT Press, vol. 21, no. 1, pp. 103126, Jan. 2014.
2013. [33] M. Nawaz, E. E. Enscore Jr., and I. Ham, A heuristic algorithm for the
[9] R. Hu, Y. Qian, S. Kota, and G. Giambene, HetnetsA new paradigm m-machine, n-job flow-shop sequencing problem, Omega, vol. 11, no. 1,
for increasing cellular capacity and coverage, IEEE Wirel. Commun., vol. pp. 9195, 1983.
18, no. 3, pp. 89, Jun. 2011. [34] A. Ur, R. Khan, and A. N. Khan, Context-aware mobile cloud comput-
[10] L. Zhou et al., QoS-aware energy-efficient resource allocation in OFDM- ing and its challenges, IEEE Cloud Comput., vol. 2, no. 3, pp. 4249,
based heterogenous cellular networks, Int. J. Commun. Syst., 2015, May./Jun. 2015.
doi:10.1002/dac.2931
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

12 IEEE SYSTEMS JOURNAL

[35] R. E. Burkard, Quadratic assignment problems, in Handbook of Com- Angel Alejandro Juan received the M.Sc. degree in
binatorial Optimization, P. M. Pardalos, D.-Z. Du, and R. L. Graham, Eds. applied mathematics and the Ph.D. degree in indus-
New York, NY, USA: Springer, 2013, pp. 27412814. trial engineering.
[36] M. Tawarmalani and N. V. Sahinidis, A polyhedral branch-and-cut He completed a predoctoral internship at Harvard
approach to global optimization, Math. Program., vol. 103, no. 2, University, Cambridge, MA, USA, and a postdoc-
pp. 225249, Jun. 2005. toral internship at the MIT Center for Transportation
[37] T. A. Feo and M. G. C. Resende, Greedy randomized adaptive search and Logistics, Cambridge. He is currently an Asso-
procedures, J. Glob. Optim., vol. 6, no. 2, pp. 109133, Mar. 1995. ciate Professor of optimization and simulation with
the Department of Computer Science, Open Univer-
sity of Catalonia, Barcelona, Spain. He was an In-
vited Researcher with the University of Southampton
Daniela Mazza received the masters degrees in elec- (U.K.), at LAAS-CNRS (France), at the University of Natural Resources and
tronic engineering and communication science and Life Sciences (Austria), and at the University of Portsmouth (U.K.). His research
the Ph.D. degree in electronics, telecommunications, interests include applications of randomized algorithms and simheuristics in lo-
and information technologies engineering from the gistics and transportation, production, and Internet computing. He has authored
University of Bologna, Bologna, Italy. or co-authored more than 150 peer-reviewed papers in these fields.
She is an expert on system optimization in pub-
lic administration organizational contexts at Emilia
Romagna Local Government, Italy. Her research in-
terests include the services to the citizens, including
solution for distributed multimedia system and ser-
Giovanni Emanuele Corazza (M92SM07) is cur-
vices management in a context of smart city. She pos- rently a Full Professor with the Alma Mater Stu-
sesses more than 15 years of experience in the local government organizations
diorum, University of Bologna, Bologna, Italy, a
and has worked as a Project Manager for technological packaging industries
Member of the Alma Mater Board of Directors,
since 1991.
the Founder of the Marconi Institute for Creativity
(2011), a Member of the Marconi Society Board of
Directors, a Member of the Board of the 5G Infras-
tructure Association, the Vice-Chairman of the Net-
Adela Pages-Bernaus received the Ph.D. degree in World2020 European Technology Platform, and the
statistics and operations research from the Polytech- founder of the Mavigex S.r.l. spin-off company. He
nic University of Catalonia (UPC), Barcelona, Spain. was Head of the Department of Electronics, Com-
She is currently a Postdoctoral Researcher at the puter Science and Systems (DEIS) in 20092012, the Chairman of the School
Department of Computer Science, Open University of for Telecommunications in 20002003, the Chairman of the Advanced Satellite
Catalonia, Barcelona, Spain. She was a Postdoctoral Mobile Systems Task Force (ASMS TF), the Founder and Chairman of the
Researcher at UPC and Norwegian University of Sci- Integral Satcom Initiative, a European technology platform devoted to satellite
ence and Technology, Trondheim, Norway. Her re- communications. He has authored or co-authored more than 260 papers. His
search interests include operations research applied research interests include wireless and satellite communications, mobile radio
to the fields of logistics and transportation, Internet channel characterization, Internet of Things, navigation and positioning, esti-
computing, and energy systems. mation and synchronization, spread spectrum and multicarrier transmission, and
scientific creative thinking.
Prof. Corazza served as an Editor for Communication Theory and Spread
Spectrum for the IEEE TRANSACTIONS ON COMMUNICATIONS in 19972012. He
was the recipient of the Marconi International Fellowship Young Scientist Award
Daniele Tarchi (S99M05SM12) received the in 1995, the IEEE 2009 Satellite Communications Distinguished Service Award,
M.Sc. degree in telecommunications engineering and the 2013 Newcom# Best Paper Award, the 2002 IEEE Vehicular Technology
the Ph.D. degree in informatics and telecommunica- Society Best System Paper Award, the Best Paper Award of the IEEE Interna-
tions engineering from the University of Florence, tional Symposium on Spread Spectrum Techniques and Application (ISSSTA)
Florence, Italy, in 2000 and 2004, respectively. 1998, at the IEEE International Conference on Telecommunication 2001, and at
He is currently an Assistant professor at the Uni- the 2nd International Symposium on. Wireless Communication Systems 2005.
versity of Bologna, Bologna, Italy. His research inter- He was the General Chairman of the IEEE ISSSTA 2008, ASMS 20042012
ests include the telecommunication area, with partic- Conferences, and MIC Conference 2013.
ular interests to resource allocation and link adapta-
tion algorithms in wireless and satellite networks. He
has been involved in several national projects as well
as European projects and has been active in several industry funded projects.
Dr. Tarchi has been an Editorial Board Member for the Wiley Wireless Com-
munication and Mobile Computing, Hindawi Journal of Engineering, and the
Scientific World Journal since 2012 and has served as an Associate Editor for the
IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS from 2008 to 2013. He
was the Symposium Chair at IEEE Wireless Communications and Networking
Conference 2011 and at IEEE Globecom 2014, and the Workshop Co-Chair at
the IEEE International Conference on Communications 2015.