Utility-based Decision Making in Collective Adaptive Systems

Vasilios Andrikopoulos

, Marina Bitsaki

, Santiago G

omez S

aez

, Dimka Karastoyanova

Christos Nikolaou

and Alina Psycharaki

Institute of Architecture of Application Systems, University of Stuttgart, Stuttgart, Germany

Transformation Services Laboratory, University of Crete, Heraklion, Greece

Keywords:

Collective Adaptive Systems, Utility, Decision Making, Choreography.

Abstract:

Large-scale systems comprising of multiple heterogeneous entities are directly inﬂuenced by the interactions

of their participating entities. Such entities, both physical and virtual, attempt to satisfy their objectives by

dynamically collaborating with each other, and thus forming collective adaptive systems. These systems are

subject to the dynamicity of the entities’ objectives, and to changes to the environment. In this work we focus

on the latter, i.e. on providing the means for entities in such systems to model, monitor and evaluate their

perceived utility by participating in the system. This allows for them to make informed decisions about their

interactions with other entities in the system. For this purpose we propose a utility-based approach for decision

making, as well as an architecture that allows for the support of this approach.

1 INTRODUCTION

Collective systems comprise heterogeneous entities

collaborating towards the achievement of their own

objectives, and the overall objective of the collec-

tive (Andrikopoulos et al., 2013). Such systems are

usually large scale, typically consisting of both phys-

ical and virtual entities distributed both organization-

ally and geographically. Entities in these systems ac-

tively inﬂuence the operation of the system by inter-

acting with other entities. Their behavior is guided

by their individual objectives and the decisions they

make in order to satisfy them. In this respect, sup-

port for decision making on the level of entities is an

integral part of the realization of such systems.

For this purpose, in the EU ALLOW Ensembles

project

, we use the concept of Collective Adaptive

System (CAS) (Kernbach et al., 2011), and deﬁne the

underpinning concepts for modeling, execution and

adaptation of CAS entities and their interactions. In

particular, we propose to model and manage entities

as collections of cells encapsulating their functional-

ities. Entities collaborate with each other to achieve

their objectives in the context of ensembles describing

the interactions among them. As a way to measure

entity objectives’ achievement, we use the concept of

utility as a function over its preferences, context, state

ALLOW Ensembles: http://www.allow-ensembles.eu

and interactions with other entities.

The contributions of this work can be summarized

as follows:

1. We propose a utility-based approach for the pur-

pose of decision making in CAS systems based on

the conceptual model introduced in the project.

2. We extend an existing architecture in order to

enable the modeling and execution of enabling

mechanisms, and discuss their implementation

based on well-established technologies.

The remaining of the paper is structured as fol-

lows: starting from a motivating scenario in Section 2,

we then introduce our proposal on how to deal with

decision making in CAS systems (Section 3). Conse-

quently, in Section 4 we extend the architecture intro-

duced in (Andrikopoulos et al., 2013) in order to en-

able the modeling and execution of the mechanisms

supporting our proposal as distributed, large scale,

pervasive systems; we also discuss the implementa-

tion of these concepts based on well-established tech-

nologies. The paper closes with a summary of related

work (Section 5), and concludes with an outline of

research challenges and future work (Section 6).

308

Andrikopoulos V., Bitsaki M., Goméz Sáez S., Karastoyanova D., Nikolaou C. and Psycharaki A..

Utility-based Decision Making in Collective Adaptive Systems.

DOI: 10.5220/0004937403080314

In Proceedings of the 4th International Conference on Cloud Computing and Services Science (CLOSER-2014), pages 308-314

ISBN: 978-989-758-019-2

 2014 SCITEPRESS (Science and Technology Publications, Lda.)

2 MOTIVATION

Supporting citizens mobility within the urban envi-

ronment is a priority for municipalities worldwide. As

part of the effort to offer “smart services” to citizens,

in the ALLOW Ensembles project the case of an ur-

ban mobility scenario is used as a CAS for demonstra-

tion and evaluation purposes (Andrikopoulos et al.,

2013). The FlexiBus service is a special type of bus in

this scenario that operates a ﬂexible route set by pas-

senger needs, allowing the advance booking of pickup

points along a (dynamically) calculated route to a set

of predeﬁned destinations (i.e., city center).

Using the FlexiBus system, each Passenger can

request a trip to one of the predeﬁned destinations in

the system by communicating with the Route Plan-

ner, asking to start at a certain time and from a pre-

ferred pickup point. Each passenger can pay their

trip directly in the bus (cash, with a credit card or a

monthly pass) or through the FlexiBus company web

site. Furthermore, during the route execution, each

passenger waiting for a bus can be notiﬁed for prob-

lems on a selected route (e.g. bus delays, accidents,

etc.). Each Bus Driver is assigned by the Route Man-

ager to a precise but dynamically calculated by the

Route Planner route to execute, including the list of

passengers assigned to it, and a unique ﬁnal desti-

nation. Bus drivers communicate with an assigned

Route Manager which monitors and ensures the cor-

rect execution of the route, in order to ask for the next

pick-up point, and to communicate information like

passengers’ check-in. During the route realization, a

bus on route can also accept passengers that have not

previously booked if there are available seats and the

passenger pick-up point is along the route.

The Route Planner attempts to deﬁne routes in a

manner that reduces the wasteful overlap that occurs

with many individual passengers traveling to the same

destination. This means carrying more people in one

route to radically reduce the number of FlexiBuses

on the road, fuel usage, CO

emissions, trafﬁc con-

gestion, etc. The passengers on the other hand have

concrete requirements on e.g. their arrival time, and

they attempt to optimize across changing dimensions

e.g. fare cost or time spent on the bus. The FlexiBus

system is therefore required to support the interac-

tion of different entities (software or physical) striving

to achieve potentially conﬂicting and changing over

time goals. In the following sections we present our

proposal for using utility as the means for allowing

these entities to co-exist and reconciliate their goals

in the system.

3 APPROACH

3.1 The ALLOW Ensembles Model

In large-scale collective adaptive systems, entities

which are heterogeneous actors of both human and

software actors, interact in various ways to form a dis-

tributed and dynamic CAS system that evolves con-

tinuously to achieve a goal in a given context. For this

purpose, in ALLOW Ensembles we model both types

of actors as entities aggregating different functionali-

ties, e.g. the check-in on a bus route, or the payment

of the fare, as reusable cells. Cells encapsulate some

functionality that the entity offers to, or requires from

the system. Cells from different entities interact with

each other as part of their predeﬁned functionality.

The outcome of these interactions are ensembles

that are created to fulﬁll speciﬁc goals initiated by the

entities. Even though each entity has its own self-

ish interests, the cooperation with others entails an

increase in each one’s satisfaction expressed in util-

ity terms. For this purpose, in the scope of the project

we propose to model the economic perspective of sys-

tems of ensembles. In particular, we consider that en-

tities take part in games according to various criteria

in order to achieve a set of objectives. Entities have

potentially conﬂicting interests within the system and

strive to achieve individual objectives, thus formu-

lating non-cooperative games (Osborne and Rubin-

stein, 1994). On the other hand, entities cooperate

to achieve a common task and fulﬁll a set of group

objectives, a situation that can modeled as coopera-

tive games (Wiese, 2010). We incorporate both types

of games to model entities’ behavior according to the

goals needed to be achieved.

3.2 Utility & Strategic Ensembles

Our position for dealing with such CAS systems as

the ones discussed above is to analyze economic mod-

els that assign utility (Norstad, 1999) to entities ac-

cording to their properties and impose constraints ac-

cording to the desired goals. The utility of an entity is

a measure of satisfaction experienced by the entity for

using a service or consuming a good. Entities make

choices so as to maximize their utility. A utility func-

tion is a way to quantify utility, assigning a number to

every possible choice of the entity such that more pre-

ferred choices get assigned larger numbers than less

preferred ones (Varian, 2010). We consider that the

cells of an entity that are involved in the selected en-

semble operate in order to satisfy the goal that is ful-

ﬁlled by the ensemble and contribute to the improve-

ment of the entity’s utility. If the entity participates

Utility-basedDecisionMakinginCollectiveAdaptiveSystems

309

in another ensemble (this is determined by the inter-

actions that the entity decides to get involved to), the

same cells operate in a new context that attributes a

new level of utility to the entity. Game theory can be

used to model the selection from a set of candidate

ensembles based on the criterion of utility maximiza-

tion (Varian, 2010): each entity chooses the ensemble

that maximizes its utility taking into account the ac-

tions taken by the other entities too. Once an ensem-

ble is being executed, it can be evaluated and com-

pared to competitors or former ensembles through the

collective utility that aggregates the individual utilities

of all entities involved.

In order to model the economic behavior of our

system, we deﬁne the concept of meta-cells that rep-

resent the economic characteristics of functional cells.

We distinguish different activities that are performed

by meta-cells according to the goals assigned to the

respective functional cells. More speciﬁcally, they

calculate the utility of an entity when participating in

a given ensemble, as well as the collective utility of an

ensemble derived by the various entities that interact

to form the ensemble, they communicate with other

cells and take decisions/compute strategies of entities,

they collect data and perform measurements (resource

consumption, satisfaction, costs, delays, prices, etc.),

and they run optimization algorithms to improve the

performance of ensembles.

Interactions of functional cells initiated by the

entities result in the creation of functional ensem-

bles (or simply, ensembles) in order to fulﬁll speciﬁc

goals. Simultaneously, strategic utility-based interac-

tions trigger the initialization of meta-cells that are

managed by a new structure called the strategic en-

semble in order to handle decision making on the level

of interaction between entities. The objectives of a

strategic ensemble include the following:

• to impose constraints according to entity goals in

order to reduce the various choices of entities,

• to evaluate ensembles from the perspective of one

entity according to the entity’s preferences dy-

namically,

• to assign utility to each entity when participat-

ing in an ensemble in order to make the optimal

choice and manage the negotiation between two

entities.

Figure 1 illustrates one of the strategic ensembles

in the FlexiBus scenario discussed in the previous sec-

tions. The meta-cells of the entities in the FlexiBus

scenario are described as follows:

1. Route Evaluation (Route Planner): evaluates each

route according to passenger preferences, passen-

ger’s previous choices and route’s reliability based

Figure 1: A Strategic Ensemble in the FlexiBus Scenario.

on historical data.

2. Routes Calculation (Route Planner): makes a list

of routes according to passenger goals, examines

whether the requirements of current passengers

are violated (estimate new travel times) for all

routes in the list and removes the routes that do

not satisfy passengers’ goals or negotiates with

passengers, evaluates the routes according to pas-

sengers’ preferences and FlexiBus’ beneﬁts.

3. Driver Allocation (Route Manager): negotiates

with the FlexiBus drivers to ﬁnd an appropriate

Bus Driver entity to execute a route.

4. Utility Evaluation (Bus Driver): calculates the

utility of the driver for the requested route accord-

ing to his preferences and accepts or rejects the

offer of the Route Manager to execute a route.

5. Utility Evaluation (Passenger): calculates a pas-

senger’s utility for each route in the list and an-

nounces the optimal choice. In Algorithm 1 we

provide an example of passenger utility as a func-

tion of travel time t, trip cost C and a set of pref-

erences such as payment method, smoking/non-

smoking trips, seat preferences, delay tolerance

and so on.

The structure of our methodology implies that in-

formation ﬂows among the various components of

the system. Meta-cells extract information from en-

tities. Goals may impose constraints on meta-cells

that solve optimization problems. Entities communi-

cate through meta-cells and use them in order to select

appropriate cells to form ensembles or reconﬁgure en-

sembles for increasing system’s utility.

3.3 Strategic Ensembles Life-cycle

The life-cycle of the strategic ensemble is closely re-

lated to that of the execution ensembles. It is created

before the execution ensemble, since the meta-cells

CLOSER2014-4thInternationalConferenceonCloudComputingandServicesScience

310

Algorithm 1 : Utility Evaluation Algorithm for Pas-

senger Entities.

for all routes do calculate U:

U =

∑

i=1

+ w

k+1

−at−bC

, where











(m+1)− j

(m+1)

if choice j of a total of m

choices is satisﬁed

0 if preference i is not satisﬁed,

, i = 1, ..., k are weights for each of k prefer-

ences and w

k+1

the weight for the component that

accounts for travel time and cost (

∑

k+1

i=1

= 1), and

a and b are the impact factors for travel time t

and cost C respectively.

end for

return max{U}

Figure 2: The life-cycle of the Strategic Ensemble.

are responsible for making decisions about the selec-

tion of the most beneﬁcial ensembles. Moreover, it

runs in parallel to the execution ensembles, since it af-

fects the operations of the execution ensembles until

their termination. We consider that each entity has its

own set of meta-cells, which interact with other cells

of this entity and with cells and meta-cells of other

entities as well. Meta-cells of an entity join or leave

a strategic ensemble and as a result the related execu-

tion ensembles, according to the decisions the entity

makes based on her calculated utility.

In particular about the FlexiBus scenario, we con-

sider two phases in the life-cycle of a route (execution

ensemble): the pre-booking phase, where a route is

going to be executed if a certain number of requests

is reached until a certain deadline, and the execution

phase, where the route is bound to start or it has al-

ready started. The timeline for the strategic ensemble

that runs through this execution ensemble is shown in

Figure 2. According to the scenario, the execution cell

of the Passenger invokes her meta-cell by sending the

trip request. Then, a series of meta-cell interactions

start with the aim to make the optimal decision for

each entity. The Utility Evaluation meta-cell of the

Passenger invokes the Routes Calculation meta-cell

of the Route Planner by sending the trip request and

waits to receive a list of routes, in order to evaluate

them. The Routes Calculation meta-cell interacts with

the Route Evaluation meta-cell of the Route Planner,

as well as with the active Route Managers in order

to identify whether the Passenger can join an existing

route. If this is possible, the Route Planner negoti-

ates with the Passengers already in the route, since

they may need to re-evaluate their utility from being

on this route, if e.g. the arrival time changes due to the

additional passenger. The list of routes returned to the

Passenger are evaluated by the Passenger using Algo-

rithm 1 and ranked accordingly. The Passenger then

joins in an (execution) ensemble with the Route Man-

ager, Bus Driver and other Passengers in the route that

she evaluated as optimal.

4 ARCHITECTURE &

IMPLEMENTATION

An architecture design capable of supporting the

phases of the CAS lifecycle is presented in (An-

drikopoulos et al., 2013). However, such architecture

does not support the utility-based decision making ca-

pabilities proposed in the previous section; for this

purpose in this work we enhance it with the necessary

components as depicted in Figure 3.

The architecture for the modeling and the execu-

tion of CAS systems comprises two major compo-

nent groups: a Modeling Tool and a Runtime Environ-

ment. The Modeling Tool consists of three main com-

ponents: a Choreography Modeler to create chore-

ography models which specify the interactions be-

tween cells of functional and strategic ensembles, a

Transformer to generate process skeletons that can

be completed to executable processes, which can be

subsequently modiﬁed in the Process Modeler The

Modeling Tool is developed as a pair of interopera-

ble Eclipse Graphical Editors that support the spec-

iﬁcation of choreographies in the BPEL4Chor lan-

guage (Decker et al., 2008). Both functional cells

and meta-cells are speciﬁed using the WS-BPEL lan-

guage, and therefore the process is developed as an

extension of the BPEL Eclipse designer described

in (Sonntag et al., 2012). The editor is also equipped

with monitoring capabilities that provide the user

real-time information related to the ﬂow execution.

The Runtime Environment comprises the neces-

sary components to enable the execution of both cells

and meta-cells. The cell and meta-cell models are de-

ployed on one or more Execution Engines and can be

instantiated at any time. Cell models may contain ab-

stract activities, and therefore the Execution Engine

must be able to start the execution of incomplete pro-

Utility-basedDecisionMakinginCollectiveAdaptiveSystems

311

Figure 3: Architecture overview

cesses, allowing the dynamic injection of additional

activities retrieved from the Adaptation Manager or

triggered through the Modeling Tool. The realized

Execution Engine is an extended Apache ODE En-

gine

, an open source implementation of BPEL, and

presented in (Sonntag et al., 2012). The Entity Man-

agement System (EMS) deals with all aspects of entity

management. When a new entity is created, its avail-

able functional and meta-cells are deployed in the Ex-

ecution Engine, and its corresponding properties to

its context model. The EMS allows the Adaptation

Manager to access the cell models, instances, context

properties, and the necessary data for utility evalua-

tion, e.g. user preferences, needed for the utility based

planning and adaptation operations.

The interaction of the Utility Module with one

or multiple meta-cells participating in a strategic en-

semble provides the necessary functionality for utility

evaluation and decision making support. The Deci-

sion Module and the Utility Evaluator deal with such

operations in a generic and domain independent man-

ner. Therefore, meta-cells are responsible for retriev-

ing the necessary data, e.g. context information, user

preferences, etc., and orchestrate the multiple services

offered in the Utility Module. Monitoring information

retrieval, e.g. KPIs or WS-BPEL Events (Kopp et al.,

2011) functionalities must be wrapped in the Moni-

toring component.

All components should be provided as services

and communicate through an Enterprise Service Bus

(ESB) solution to facilitate their integration. As mul-

tiple organizational levels can be present in such sys-

tems, we use the ESB

multi-tenant aware ESB so-

lution, as presented in (Strauch et al., 2013), for ensur-

ing communication isolation between multiple orga-

nizations and its users. ESB

enhances the Apache

Apache ODE: http://ode.apache.org/

ServiceMix solution

with multi-tenant communica-

tion support within service endpoints deployed in the

ESB, and multi-tenant aware dynamic endpoint de-

ployment and management capabilities.

5 RELATED WORK

Several studies related to the collective or adaptive

aspects of complex systems have been driven in ar-

eas such as Swarm Intelligence (C. Pinciroli et al.,

2011; Levi and Kernbach, 2010), Autonomic comput-

ing (Bruni et al., 2012; Lewis et al., 2012), and Multi-

agent based systems (Cabri et al., 2011; Lavinal et al.,

2006). These converge in the need for deﬁning in-

teraction rules between participants towards adapting

the system under ordinary and extraordinary events.

However, there exists a gap towards a general solu-

tion deﬁnition for all aspects in CAS. In this work

we aim to target this problem by proposing the usage

and extension of existing modeling and execution ap-

proaches towards providing a generic and evolution-

ary utility-based adaptation in CAS.

Interactions between multiple participants in

a choreography can be modeled following the

interaction or the interconnection modeling ap-

proaches (Barker et al., 2009). In the former, com-

munication between participants is modeled using

atomic interaction activities. The WS-CDL (Ka-

vantzas et al., 2005) and the Savara

project are ap-

proaches that support the explicit speciﬁcation of the

interaction activities. On the other hand, the inter-

connection modeling approach consists on intercon-

necting the communication activities of each partic-

Apache ServiceMix: http://servicemix.apache.org

http://www.jboss.org/savara

CLOSER2014-4thInternationalConferenceonCloudComputingandServicesScience

312

ipant of the choreography. This approach is sup-

ported in the CHOReOS Integrated Development and

Runtime Environment

, in the Open Knowledge Eu-

ropean project

, and in BPEL4Chor (Decker et al.,

2007). As BPEL4Chor enables the choreography

speciﬁcation atop of WS-BPEL and decouples the

choreography behavior speciﬁcation from the techni-

cal communication details, it is considered as an ap-

propriate extensible point in CAS modeling and spec-

iﬁcation.

Value is intimately linked to service systems: the

latter exist only as long as they can create value for

both the providers and the consumers; otherwise they

dissolve and disappear. In (Bloch and Jackson, 2007),

value is deﬁned as “beneﬁts of an agent accrued by

his participation in the network minus any costs in-

volved in setting up the network links directly or in-

directly”. A method is proposed in (Caswell et al.,

2008), for computing values by taking into account

service system partners’ satisfaction and additional

value accrued by the relationship levels of the vari-

ous partners. In (Gordijn et al., 2001), the authors

combine IT systems analysis with business modeling

to build an e-business model that speciﬁes e-business

scenarios rather than deﬁning values.

While various approaches have been proposed to

measure service network performance, little exper-

imental testing or theoretical investigation on com-

peting networks has been performed. (Biem and

Caswell, 2008) describe building block elements of

a value network model and design a network-based

strategy for a prescriptive value network analysis.

(Allee, 2008) provides a systematic but qualitative

way for approaching the dynamics of intangible value

realization, inter-convertibility, and creation. Both

of the above approaches use qualitative methods

to describe value in a service network in contrast

to (Caswell et al., 2008) that calculates value in a

quantiﬁable manner. A service system’s performance

is studied in (Voskakis et al., 2011) by solving value

optimization problems with respect to service prices.

(Van Dinther et al., 2009) studies the strategic be-

havior of service providers within service value net-

works and proposes an auction-based mechanism to

efﬁciently match service offers and requests and de-

termine prices, showing that incentive compatibility

holds under certain conditions. (Katsamakas, 2005)

analyzes industry structure in the presence of value

networks and has developed a model dealing with a

number of design aspects, such as the supplier num-

ber and the partner investments importance. It is

CHOReOS: Large Scale Choreographies for the Future

Internet: http://www.choreos.eu/

Open Knowledge: http://www.openk.org/

shown that industry structure is more likely to shift

to competition between value networks in which IT

plays an important role. Systems encountering com-

petition between service networks, analyzed with re-

spect to their evolution through well-deﬁned strate-

gies, have received little attention. Initial work is

provided in (Bitsaki et al., 2012) on competing ser-

vice networks by deﬁning and simulating strategies

in games of incomplete information. The results show

that Nash Equilibria exist when one player represents

each service system.

6 CONCLUSIONS AND FUTURE

WORK

Collective Adaptive Systems (CAS) are formed by

multiple heterogeneous virtual and physical enti-

ties that collaborate with each other towards ful-

ﬁlling their individual objectives (Kernbach et al.,

2011). However, such collaboration can change over

time, as entities’ objectives can dynamically adapt

or change due to external or internal environmental

events. Therefore, on the one hand entities should

be provided with the necessary mechanisms that will

guarantee a successful fulﬁllment of their objectives,

while on the other hand such mechanisms must min-

imize the impact of entities’ inﬂuences on the collec-

tive. For this purpose, the concepts of utility and deci-

sion making are brought into the cells and ensembles

paradigm of CAS in the ALLOW Ensembles project.

In this work we focus on providing the necessary

support to entities to facilitate the modeling, monitor-

ing, and evaluation of their perceived utility by par-

ticipating in the system. Therefore, we describe in a

ﬁrst step a conceptual view on meta-cells and strategic

ensembles, and propose the strategic ensemble life-

cycle as an emerging negotiation mechanism. A util-

ity evaluation algorithm is then proposed for a con-

crete entity of the FlexiBus scenario. We plan to ex-

tend the approach to model the various interactions

among entities by means of games in order to make

decisions on which ensembles to choose to partici-

pate or which entities to add in an existing ensemble,

taking into account future passenger demands or other

environmental events.

In a second step, an architectural and a work in

progress implementation approach presented in (An-

drikopoulos et al., 2013) is enhanced towards sup-

porting the multiple required phases for utility-based

decision making in CAS. Future work also focuses

on analyzing and utilizing multiple provisioning and

management techniques for supporting entities join-

ing and leaving the system, as well as investigating

Utility-basedDecisionMakinginCollectiveAdaptiveSystems

313

on different distribution and deployment options of

the runtime environment. For these purposes, the pre-

viously described components must be extended and

integrated in the prototype implementation.

ACKNOWLEDGEMENTS

The research leading to these results has received

funding from the EU’s Seventh Framework Pro-

gramme (FP7/2007-2013) project ALLOW Ensem-

bles (grant agreement no. 600792), and from the Ger-

man Research Foundation (DFG) within the Cluster

of Excellence (EXC310) in Simulation Technology.

REFERENCES

Allee, V. (2008). Value network analysis and value con-

version of tangible and intangible assets. Journal of

Intellectual Capital, 9(1):5–24.

Andrikopoulos, V., Bucchiarone, A., S

aez, S. G., Karastoy-

anova, D., and Mezzina, C. A. (2013). Towards Mod-

eling and Execution of Collective Adaptive Systems.

In Proceedings of WESOA’13, pages 1–12. Springer.

Barker, A., Walton, C. D., and Robertson, D. (2009). Chore-

ographing web services. IEEE Transactions on Ser-

vices Computing, 2(2):152–166.

Biem, A. and Caswell, N. (2008). A value network model

for strategic analysis. In Hawaii International Con-

ference on System Sciences, Proceedings of the 41st

Annual, pages 361–361. IEEE.

Bitsaki, M., Nikolaou, C., Voskakis, M., van den Heuvel,

W.-J., and Tsikrikas, K. (2012). Performance analysis

and strategic interactions in service networks. Interna-

tional Journal On Advances in Networks and Services,

4(3 and 4):375–385.

Bloch, F. and Jackson, M. O. (2007). The formation of net-

works with transfers among players. Journal of Eco-

nomic Theory, 133(1):83–110.

Bruni, R., Corradini, A., Gadducci, F., Lluch-Lafuente, A.,

and Vandin, A. (2012). A conceptual framework for

adaptation. In FASE, pages 240–254.

C. Pinciroli et al. (2011). Argos: A modular, multi-engine

simulator for heterogeneous swarm robotics. In IROS,

pages 5027–5034.

Cabri, G., Puviani, M., and Zambonelli, F. (2011). Towards

a taxonomy of adaptive agent-based collaboration pat-

terns for autonomic service ensembles. In CTS, pages

508–515.

Caswell, N. S., Nikolaou, C., Sairamesh, J., Bitsaki, M.,

Koutras, G., and Iacovidis, G. (2008). Estimating

value in service systems: a case study of a repair ser-

vice system. IBM Systems Journal, 47(1):87–100.

Decker, G., Kopp, O., Leymann, F., Pﬁtzner, K., and Weske,

M. (2008). Modeling Service Choreographies Using

BPMN and BPEL4Chor. In Proceedings of CAiSE’08,

pages 79–93. Springer.

Decker, G., Kopp, O., Leymann, F., and Weske, M. (2007).

BPEL4Chor: Extending BPEL for Modeling Chore-

ographies. In Proceedings of ICWS’07, pages 296–

303, Salt Lake City. IEEE Computer Society.

Gordijn, J., Akkermans, H., and Van Vliet, J. (2001). De-

signing and evaluating e-business models. IEEE intel-

ligent Systems, 16(4):11–17.

Katsamakas, E. (2005). Competing value networks, incom-

plete contracts and it. In Proceedings of HICSS’05,

pages 212b–212b. IEEE.

Kavantzas, N., Burdett, D., Ritzinger, G., Fletcher, T., La-

fon, Y., and Barreto, C. (2005). Web services chore-

ography description language version 1.0.

Kernbach, S., Schmickl, T., and Timmis, J. (2011). Collec-

tive adaptive systems: Challenges beyond evolvabil-

ity. ACM Computing Research Repository (CoRR).

Kopp, O., Henke, S., Karastoyanova, D., Khalaf, R., Ley-

mann, F., Sonntag, M., Steinmetz, T., Unger, T., and

Wetzstein, B. (2011). An Event Model for WS-BPEL

2.0. Technischer Bericht Informatik 2011/07, Uni-

versit

at Stuttgart, Fakult

at Informatik, Elektrotechnik

und Informationstechnik, Germany.

Lavinal, E., Desprats, T., and Raynaud, Y. (2006). A

generic multi-agent conceptual framework towards

self-management. In NOMS, pages 394–403.

Levi, P. and Kernbach, S. (2010). Symbiotic Multi-Robot

Organisms: Reliability, Adaptability, Evolution, vol-

ume 7. Springer Verlag.

Lewis, P., Platzner, M., and Yao, X. (2012). An outlook

for self-awareness in computing systems. Awareness

Magazine.

Norstad, J. (1999). An introduction to utility theory. Un-

published manuscript at http://homepage. mac. com/j.

norstad.

Osborne, M. J. and Rubinstein, A. (1994). A Course in

Game Theory. MIT Press.

Sonntag, M., Hahn, M., and Karastoyanova, D. (2012).

Mayﬂower - Explorative Modeling of Scientiﬁc

Workﬂows with BPEL. In Proceedings of BPM’12,

pages 1–5. CEUR Workshop Proceedings.

Strauch, S., Andrikopoulos, V., S

aez, S. G., and Leymann,

F. (2013). ESB

: A Multi-tenant Aware Enter-

prise Service Bus. International Journal of Next-

Generation Computing, 4(3):230–249.

Van Dinther, C., Blau, B., and Conte, T. (2009). Strategic

behavior in service networks under price and service

level competition. In Wirtschaftsinformatik (1), pages

599–608. Citeseer.

Varian, H. (2010). Intermediate microeconomics: a modern

approach. Norton & Company.

Voskakis, M., Nikolaou, C., van den Heuvel, W.-J., and Bit-

saki, M. (2011). Service network modeling and per-

formance analysis. In Proceedings of ICIW’11, pages

58–63.

Wiese, H. (2010). Applied cooperative game theory.

http://www.uni-leipzig.de/ micro.

CLOSER2014-4thInternationalConferenceonCloudComputingandServicesScience

314