=Paper=
{{Paper
|id=Vol-1867/w19
|storemode=property
|title= SMARTSAN: A P2P Social Agent Network for Generating Recommendations in a Smart City Environment
|pdfUrl=https://ceur-ws.org/Vol-1867/w19.pdf
|volume=Vol-1867
|authors=Antonello Comi,Domenico Rosaci
|dblpUrl=https://dblp.org/rec/conf/woa/ComiR17
}}
== SMARTSAN: A P2P Social Agent Network for Generating Recommendations in a Smart City Environment==
https://ceur-ws.org/Vol-1867/w19.pdf
108
1
SMARTSAN: A P2P Social Agent Network for
Generating Recommendations in a Smart City
Environment
Antonello Comi†
†
DIIES, University Mediterranea of Reggio Calabria, Via Graziella, Località Feo di Vito, 89122 Reggio Calabria,
Italy, antonello.comi@unirc.it
Domenico Rosaci†
†
DIIES, University Mediterranea of Reggio Calabria, Via Graziella, Località Feo di Vito, 89122 Reggio Calabria,
Italy, domenico.rosaci@unirc.it
Abstract—The rapid advances of the Internet of Things (IoT) with high levels of effectiveness and efficiency [3]–[5]. In
has implied that new kind of social networks are becoming particular, if the agents are provided with social capabilities,
pervasive, and the use of the multi-agent technology has been making them able of establishing interactions between each
recognized to have a key role in designing distributed software
architecture capable of implementing autonomous, adaptive and others, there is the possibility of making social networks
proactive services for the citizens, with high levels of effectiveness of agents, encoding in them some useful information about
and efficiency.In this scenario, we can imagine the possibility to the actors of the scenario, and their existing relationships.
form groups of agents associated with users and objects that are In this scenario, we can imagine the possibility to form
geographically close, and use these groups to provide the users groups of agents associated with users and objects that are
with recommendations about services of potential interest.In this
paper, we propose a novel recommender system for smart city geographically close, and use these groups to provide the users
environments, based on a P2P social agent network, designed to with recommendations about services of potential interest. We
faced these highlighted problems. The P2P network topology is recognize that two key issues arise in this context, namely
continuously adapted to the changes of desires and necessities of (i) the necessity of making highly dynamic the formation of
the users, using an algorithm that match the different exigences groups, to address the changes of the desires expressed by the
for forming groups of agents geographically close with each other.
Moreover, in order to consider also the issue (ii) above, the local users during their movements in the city and (ii) the possibility
groups are integrated in a global social network, that can be used of considering also the possibility that objects not belonging
to discover services of interest for a given user also managed by to the group of a given user can be of interest for that user,
agents that do not belong to the same local group of the user. in the case no other suitable alternatives are present in the
Index Terms—Recommendation, Online Communities, Trust, local group. However, although several proposals have been
Group. presented in the literature, that use middleware architecture to
support the development of applications in the scope of Smart
Cities [6]–[10], and also considering that some approaches
I. I NTRODUCTION
have been introduced for constructing social networks of
Nowadays, the recent developments of communication tech- agents [11]–[13], any of these proposals, at the best of our
nology and intelligent systems generated an increasing ten- knowledge, addresses the issues above. In this paper, we pro-
dency to improve human life in its social aspects, including pose a novel recommender system for smart city environments,
entertainment, commerce, socialization, education, transporta- called SMART City Social Agent Network (SMARTSAN),
tion, etc. These advances should make social contexts, in par- based on a P2P social agent network, designed to faced
ticular the city, to become a better place to work and socialize, these highlighted problems. The P2P network topology is
implying that communication should be effective and low cost, continuously adapted to the changes of desires and necessities
and the urban organization should be intelligent enough to of the users, using an algorithm that match the different
enable ubiquitous computing environment for delivering smart exigences for forming groups of agents geographically close
services to a much wider audience [1]. The rapid advances with each other. Moreover, in order to consider also the issue
of the Internet of Things (IoT) [2] has implied that new (ii) above, the local groups are integrated in a global social
kind of social networks are becoming pervasive, connecting network, that can be used to discover services of interest for
not only networked devices like PCs and smartphones, but a given user also managed by agents that do not belong to the
also un-networked things as sensors, actuators, refrigerators, same local group of the user.
TVs, vehicles, clothes, food, medicines, books, luggage and, The paper is organized as follows: in Section II we deal
obviously, people. In this context, the use of the multi-agent with some related work. Section III provides technical details
technology has been recognized to have a key role in design- about the Recommender System architecture and the Social
ing distributed software architecture capable of implementing Network of Agents, while Section IV describes the algorithm
autonomous, adaptive and proactive services for the citizens, for matching the different agent goals in local groups. Finally,
109
in Section V we draw our conclusions and illustrate some owners. In such an approach a Users-to-Group matching algo-
possible future works. rithm allows the agents to dynamically manage the evolution
of the social network organization. Finally, in [13], the authors
present a framework that exploits homophily to establish an
II. R ELATED W ORK
integrated network linking a user to interested services and
In [6], the authors developed a multi-agent based Smart connecting different users with common interests, upon which
Mobile Virtual Community Management System (SMVCMS) both friendship and interests could be efficiently propagated.
capable of providing a decentralized and open management The proposed friendship-interest propagation (FIP) framework
of virtual communities, based on the agent-oriented platform devises a factor-based random walk model to explain friend-
JaCaMo and its Android client based platform JaCa-Android. ship connections, and simultaneously it uses a coupled latent
In such a system, a participant in virtual communities is factor model to uncover interest interactions.
supported by a Jason agent that encapsulates the logic and
the control of the participation in a virtual community (such III. T HE M ULTI -L AYER R ECOMMENDER S YSTEM
as publishing posts, notifying members, making recommenda- A RCHITECTURE
tions for the user, etc.). The authors exploited SMVCMS in the
context of Smart Cities, showing that the system fulfills decen- The SMARTSAN recommendender system generates rec-
tralization of community management, personalized automatic ommendations for each logged user, computing them by means
management and discovery of communities, so that any agent of a set of software agents, and exploiting a number of inert
can create its own community. In [7], the authors presented objects with IoT capabilities. We suppose that both users
a scalable agent architecture with emergent properties in and objects can provide services (e.g. information about city
the context of Smart Cities. These agents form on-demand attractions, museums, restaurants, products to buy or trip
control loops within the urban system, by considering both to organize etc.). The system architecture is composed by
the protection and the comfort of its inhabitants, at varying components organized in a stack of four layers, as graphically
degrees of intelligence and abstraction of tasks. A case study represented in Figure 1. The layers are described below:
is analysed regarding a real-time creation of a control loop for IoT. It is at the bottom of the layer stack, and
an underground railway intersection system. In [9], a system contains all the objects in the Smart City that are reg-
has been designed to provide an easy and ubiquitous access istered in the system, and that have IoT capabilities.
to the desired information about tourist attractions, and to All these objects can communicate with the device
generate proactive recommendations of attractions by means agents of the users (see the P2P communication
of a hybrid recommendation system that considers elements layers) and also with the server agents of the Social
such as the user profile and preferences, the location of Agent Network to which they transmit information
the tourist and the activities, and the opinions of previous about their position and the deployed services.
tourists. This system is capable to adapt itself to changes in P2P Communication. This layer is placed on the
the activities and incorporate new information transparently top of the IoT, and it is composed by a set of
at execution time. In [10] a system architecture and design device agents, where each of them, denoted by
methods are proposed to support the delivery of location-based du , is associated with a fixed user of the system
recommendation services to create personalized tour planning, and lives in a device exploited by the user (e.g. a
based on tourists current location and time, as well as personal smartphone, a tablet). All the device are organized in
preferences and needs. The system is capable to efficiently a P2P Network, and communicate with each others
provide various recommendations regarding sightseeing spots, by means of a P2P protocol. Moreover, the device
hotels, restaurants, and packaged tour plans. Regarding the agent of a user u builds and updates a profile Pu
formation of social networks of agents, in [11] the authors of u, containing some information about the u’s
deal with the dynamics behind group formation and evolution preferences and past behaviours. Such a profile is
in social networks of agents, analysing in particular the notion periodically sent to the DF agent belonging to the
of compactness of a social group and arguing that the mutual Smart City Recommendation Layer.
trustworthiness between the group members should be consid- Social Agent Network. It contains a set of social
ered as an important factor in defining such a notion. They agents, each of them associated with a different user
propose a quantitative measure of group compactness that u or with an IoT object o of the system. This layer
takes into account both the similarity and the trustworthiness maintains a representation a social network, where
among users, and introduce an algorithm to optimize such each node represents a social agent and each arc
a measure. In [12], after defining a notion of compactness between two nodes represents a trust relationship
for a group, that integrates similarity and mutual trust, the between the two respective social agents. Formally,
authors propose to provide each user with a software agent the social network is denoted by SN = ⟨SA, G⟩,
associated with each topic of interest for the user, and that where SA is the set of social agents and G is the
represents a user’s avatar in the corresponding dimension. This set of groups contained in SN . We also assume that
allows the user to delegate to his/her agent the management of each group g is managed by an administrator agent
group joining requests regarding a given topic, selecting only ag . The formation of a group is a process based on
those interlocutors which appear the most appropriate for their two main events: a user asks for joining with a group
2
110
Fig. 2. The Social Agent Network, as integration of local networks
asymmetric, in the sense that we do not automatically expect
that v trusts u at the same level.
Generally, in traditional approaches as in [14], a trust
measure combines two components ru,v and Ru , where ru,v is
Fig. 1. The SMARTSAN Multi-Layer Architecture called reliability of u, and represents the trustworthiness that v
has in u based on the past interactions between u and v, while
Ru is called reputation of u, representing the trustworthiness
and the administrator of the group accepts or refuses that the whole social network assigns to u.
the request. We assume that the group formation As for the reliability, we denote it by the mapping ru,v ,
follows the algorithm described in Section IV. ∪
assuming values ranging in the domain [0 · · · 1] N U LL,
Smart City Recommendation. This layer is at
while ru,v = NULL means that v did not have past interactions
the top of the stack, and it contains the following
with u and thus it is not able to evaluate u’s trustworthiness.
components:
Regarding the reputation of u, we denote it by Ru in the
• An Agent Management System (AMS), that interval [0 · · · 1]ϵR. The reputation is computed as follows
manages the registration of each user and each
∑
IoT object in the system. Ru = fρ (1)
• A Directory Facilitator (DF) that provides a ρ∈F EDu
service of yellow pages, storing for each user
u his profile Pu . where |F EDu | is the set of the services provided by the
• A set of recommender agents, where each of
agent u and f is the feedback, representing the level of
them, denoted by ru , is capable to generate satisfaction of the other agents for those services.
some useful recommendations that are sent to The two trust components reliability and reputation are
the device agent d : u. integrated in a unique value to compute the mapping trust
tu,v of u about v, producing a input ranging in [0 · · · 1] as
follows:
A. Trust Measures
tu,v = ω · relu,v + (1 − ω) · repv (2)
The trust measure tu,v between two social agents u and v
represents the degree of trustworthiness that u assigns to v: where ω is a real number, ranging in [0...1], which is set
tu,v = 0 (resp. tu,v = 1) means that u assigns the minimum by u to weight the relevance he/she assigns to the reliability
(resp. maximum) trustworthiness to v. The trust measure is with respect to the reputation.
3
111
B. Similarity
∑ s
∑ s
The similarity measures how much close the profiles of two v∈gu ratev + / u tu,v · ratev
v ∈g
agents are. The information stored in agent profiles strictly ru∗s = ψ·s(u, k)+(1−ψ)· ∑ ∗
v∈|Rs | tu,v
depends on the application domain. Generally, it can be for- (4)
malized as a tuple of pair, ← (f1 , v1 ), (f2 , v2 ), ..(fn , vn ) →, where gu is the group to which the agent u belongs and
where f1 ,..,fn are the features composing the profile, i.e. some |Rs | is the set of the agents who have evaluated the service
agent characteristics representing interests, preferences, tech- s. It is calculated as the combination of two contribution: the
nological parameters etc., and v1 ,..,vn are the corresponding average rating of the agents that belong to the group of the
values, such that each value vi ∈ D(fi ), where D(fi ) is the agent u and the score that the other groups give to the product
domain of fi . The similarity between two agents u and v is a multiplied by the trust that u assign to its agents.
mapping su,v yielding values ranging in the real interval [0.1], This way, the recommendation of services depends on the
representing the degree of closeness between the profiles Pu topological structure of the social agent network, that can
and Pv : su,v = 0 (resp. su,v = 1) means that Pu completely viewed as composed by the groups, that are a sort of local
differs (resp. perfectly coincides). Differently from the trust, social networks, whose formation follows the algorithm that
the similarity measure is symmetric. Many proposals have will be described in the next section
been presented in the literature for modelling similarity (e.g.,
those described in [11], [15]). The approach presented in this IV. T HE G ROUP F ORMATION A LGORITHM
paper is dully orthogonal to the particular definition chosen
for the similarity measure. The algorithm for forming the groups periodically repeat a
fixed behaviour, allowing the groups’ composition to dynam-
ically change with the evolution of the social agent network.
C. Agent recommendation Let T be the time between two consecutive executions of the
The user receives, at the current step, some recommenda- algorithm. On the single agent side this procedure is executed
tions about the services present in the system. In other words, to join with a set of groups focused on the same topic (or set
rus is the recommendation that the agent u receives about a of topics) for taking benefits from joining with more than one
service s provided by an agent k. It is calculated as follows: group. We also assume that agent can query to a distributed
database, named GR (Group Repository), on which the list of
∑ s groups of the social network is stored.
v∈ρ,v̸=u tu,v · ratev
rus = ψ · s(u, k) + (1 − ψ) · ∑ (3) The behaviour performed by the social agents is as follows.
v∈ρ,v̸=u tu,v Let Xi be the set of the groups which the agent ai is affiliated
In other words, is composed by two components, the first to, and NM AX the maximum number of groups which a trader
depnding on the similarity between the agent u and the agents agent can analyze at time t. It is supposed that NM AX ≥ |Xi |.
k that provides the service s, and the second depending on Furthermore, we suppose that ai stores into a cache the group
the opinions expressed by the whole community of the agents profile of each group contacted in the past and the timestamp
about s, taking into account the trust measures. Each of the d of the execution of the procedure for that group. Finally, let
two component are weighted, using a weight ψ, appositely ξi a timestamp threshold and χi ∈ [0, 1] be a threshold fixed
chosen depending on the application domain, ranging from by the agent ai . The ratio behind the procedure executed by
[0, 1]. the social agent is represented by the attempt, of the social
As for the second component, ratesv is the opinion of the agent, to improve the advantage in joining with a group. For
the agent v about the service s (a number between 0 and 1), this aim, first of all, the values of advantage are recalculated
weighed by the trust of v. This means that his/her opinion if they are older than the fixed threshold ξi . Then, candidate
about a service is taken into account if his/her trustworthiness groups are sorted in a decreasing order with respect to their
is high. The weighted average allows us to identify an average trust value. If some groups in the set Lgood are not in the set
value in which the starting numerical values have their own Xi , then agent ai can potentially improve convenience of the
importance, specified by its weight. In particular, we can user ui , if they accept the user itself to join with. The only
identify the centre of gravity of the rate. In this way, we give constraint of the algorithm is the maximum number of groups
more importance to the rate from users that the agent u trusts. the agent want/can to join with.
On the group side, the algorithm works as follows. Let
Kj be the set of the agents affiliated to the group gj , where
D. Groups ||Kj || ≤ KM AX , being KM AX the maximum number of
At this point, we introduce the group’s concept in the social users allowed to be within the group gj . Suppose that the
network. In this context, we define trust t∗u,v in two different administrator agent Aj stores into its cache the profile Pi of
ways. We suppose that the trust perceived by an agent u with each agent ai ∈ Kj and the timestamp di of its acquisition.
respect to the component of his/her group is equal to 1 (i.e, Moreover, let ωj a timestamp and πj ∈ [0, 1] ∈ R be a
t∗u,v =1), instead the agent u considers the trust that the agent threshold fixed by the agent Aj . The procedure performed
has in the whole community (see Equation 2). In this way, we by the group agent Aj is triggered whenever a join request by
define rg∗w that is the recommendation that the user u receives the social agent ai (along with its profile Pi ) is received by
about the service s in presence of the groups: Aj . First of all, parameter KM AX represents the maximum
4
112
number of users that can join with a given group. In fact, [2] L. Atzori, A. Iera, and G. Morabito, “Understanding the internet
if the group has reached this maximum, no more users can of things: definition, potentials, and societal role of a fast evolving
paradigm,” Ad Hoc Networks, vol. 56, pp. 122–140, 2017.
be accepted to join with the group. The group agent asks the [3] M. Roscia, M. Longo, and G. C. Lazaroiu, “Smart city by multi-agent
updated profile of the components of the group itself, therefore systems,” in Renewable Energy Research and Applications (ICRERA),
the trust tgj ,ai is computed 2013 International Conference on. IEEE, 2013, pp. 371–376.
∪ for all these agents and a new, [4] M. Longo, M. Roscia, and G. C. Lazaroiu, “Innovating multi-agent
sorted set Kgood ⊂ {Kj ai } is built. Then, the group agent systems applied to smart city,” Research Journal of Applied Sciences,
will send a leave message to all the social agents ai showing Engineering and Technology, vol. 7, no. 20, pp. 4269–4302, 2014.
[5] C. Maciel, P. C. de Souza, J. Viterbo, F. F. Mendes, and A. E. F.
a trust tgj ,ai . Finally, if ai ∈ Kgood , its request is accepted. Seghrouchni, “A multi-agent architecture to support ubiquitous applica-
tions in smart environments,” in Agent Technology for Intelligent Mobile
Services and Smart Societies. Springer, 2015, pp. 106–116.
V. C ONCLUSION [6] M. Fahad, O. Boissier, P. Maret, N. Moalla, and C. Gravier, “Smart
places: Multi-agent based smart mobile virtual community management
In this paper, we described a novel recommender system system,” Applied intelligence, vol. 41, no. 4, pp. 1024–1042, 2014.
for smart city applications, called SMART City Social Agent [7] M. Patrascu, M. Dragoicea, and A. Ion, “Emergent intelligence in agents:
A scalable architecture for smart cities,” in System Theory, Control and
Network (SMARTSAN), organized in a multi-layer architec- Computing (ICSTCC), 2014 18th International Conference. IEEE,
ture comprehending both a P2P platform and a social agent 2014, pp. 181–186.
network, built on top of an IoT infrastructure. The social [8] A. Sassi and F. Zambonelli, “Towards an agent coordination framework
for smart mobility services,” in 8th International workshop on agents in
agent network allows the presence of groups of social agents, traffic and transportation (May 2014), 2014.
formed on the basis of trust measures, and the overall topology [9] M. Batet, A. Moreno, D. Sánchez, D. Isern, and A. Valls, “Turist@:
derives by the integration of these groups, that are dynamically Agent-based personalised recommendation of tourist activities,” Expert
Systems with Applications, vol. 39, no. 8, pp. 7319–7329, 2012.
adapted in time to the changes of desires and necessities of the [10] C.-C. Yu and H.-P. Chang, “Personalized location-based recommenda-
users, using an algorithm that match social agent desires and tion services for tour planning in mobile tourism applications,” in In-
existing groups’ exigences. The use of local groups is used ternational Conference on Electronic Commerce and Web Technologies.
Springer, 2009, pp. 38–49.
to discover services of interest for a given user, integrating
the recommendations coming from the agents of the local [11] P. De Meo, E. Ferrara, D. Rosaci, and G. M. Sarné, “Trust and com-
pactness in social network groups,” IEEE transactions on cybernetics,
groups the user belongs which and those of the agents that vol. 45, no. 2, pp. 205–216, 2015.
do not belong to the same local groups of the user. Our [12] P. De Meo, F. Messina, D. Rosaci, and G. M. Sarné, “Improving the
ongoing research is currently devoted to apply the approach compactness in social network thematic groups by exploiting a multi-
dimensional user-to-group matching algorithm,” in Intelligent Network-
to real smart city applications, in which the advantages and ing and Collaborative Systems (INCoS), 2014 International Conference
limitations introduced by our proposal can be quantitatively on. IEEE, 2014, pp. 57–64.
and effectively evaluated. [13] S.-H. Yang, B. Long, A. Smola, N. Sadagopan, Z. Zheng, and H. Zha,
“Like like alike: joint friendship and interest propagation in social
networks,” in Proceedings of the 20th international conference on World
wide web. ACM, 2011, pp. 537–546.
R EFERENCES [14] D. Rosaci, G. M. Sarné, and S. Garruzzo, “Integrating trust measures
[1] A. J. Meijer, J. R. Gil-Garcia, and M. P. R. Bolı́var, “Smart city in multiagent systems,” International Journal of Intelligent Systems,
research: Contextual conditions, governance models, and public value vol. 27, no. 1, pp. 1–15, 2012.
assessment,” Social Science Computer Review, vol. 34, no. 6, pp. 647– [15] D. Rosaci, “Trust measures for competitive agents,” Knowledge-Based
656, 2016. Systems, vol. 28, pp. 38–46, 2012.
5