TOWARDS A FRAMEWORK FOR AUTOMATED

E-NEGOTIATIONS

Giannis Koumoutsos and Kleanthis Thramboulidis

Electrical & Computer Engineering, University of Patras, 26500 Patras, Greece

Keywords: E-negotiations, Ontologies, Process Specification, Protocol Specification, Semantic Web Services.

Abstract: In this paper a framework to support fully automate e-negotiations is presented. A novel approach that

exploits ontologies, Semantic Web Services and software agent platforms to define an architecture that

favours flexible, dynamically created and adapted e-negotiations, is described. Negotiating agents can enter

a negotiation area, acquire the negotiation protocol or even suggest their own in a widely understandable

notation and participate. Rule-based protocol specification along with process specification for describing

interactions during e-negotiations is presented.

1 INTRODUCTION

E-negotiations try to virtualize the real life

negotiation procedures in the future internet-based

world-wide market. Most of the negotiations used

today are manual e-negotiations involving only

human participants. However, the trend is to create

systems able to participate in fully automated

negotiation procedures. This means that the involved

parties should behave as autonomous software

agents. In the context of this paper only trade e-

negotiations are considered, which are according to

(Bichler et. al., 2002) “negotiation processes in

electronic markets for the exchange of goods and

services based on bargaining, bidding, or dispute

resolution.”

Except from the fixed-price sales through

uncountable e-shops which involve no real

negotiations, auctions are currently the most widely

used form of e-negotiations. Due to the low cost

server-based implementation, auctions have rapidly

proliferated on the internet with multiple alternative

schemes (Wurman et. al., 2001). E-negotiations can

also take a more complex form called bargaining

which involves proposals and counter-proposals

until an agreement is reached. Despite the growth of

e-auctions, complex negotiation systems have not

been extensively studied and implemented and

mature tools for such an effort are not available. The

main objective of this work is to define a framework

for an effective analysis, design and automated

implementation of complex negotiations.

Negotiation protocols and negotiation strategies

are the basic concepts of e-negotiation. A

negotiation protocol is a set of rules which govern

the interaction, whereas a Negotiation Strategy is a

decision making model that participants should

employ in order to achieve their goal. The

negotiation strategy is built upon the selected

negotiation protocol and is private for each

participant. The focus of this paper is on the

negotiation protocol specification.

The remainder of this paper is organized as

follows: In section 2 we discuss the state of the art

and the related background work. In section 3 we

present the proposed layered architecture and in

section 4 we describe the negotiation process based

on this architecture. In section 5 we present our

approach on modelling the negotiation domain and

processes specification, as well as the automated

implementation with the use of Semantic Web

Services (SWS). A simple negotiation example

using the proposed notation is presented in section 6

and finally we conclude the paper in section 7.

2 STATE OF THE ART

Various research groups are working towards

automated e-negotiations with most of them

focussing on auctions. For example, in (Kersten, et

al., 2004) a configurable e-negotiation server is

proposed to support bargaining, splitting negotiation

process into well-defined phases. A set of rules that

govern the information processing, the decision-

making, and communication acts is presented.

More recent research goes beyond auctions and

considers more complicated negotiation schemes. In

168

Koumoutsos G. and Thramboulidis K. (2007).

TOWARDS A FRAMEWORK FOR AUTOMATED E-NEGOTIATIONS.

In Proceedings of the Second International Conference on e-Business, pages 168-173

DOI: 10.5220/0002114001680173

 SciTePress

(Kim and Segev, 2005) a state-chart description of

one e-negotiation protocol and the corresponding

BPEL4WS process is provided. (Chiu et al., 2005)

presents an approach for developing e-negotiation

plans providing meta-models for e-contract

templates and e-negotiation processes. A complete

framework to the same direction is described in

(Benyoucef and Rinderle, 2006). A service oriented

environment is defined using state-charts for model-

driven development as well as automatic generation

of orchestration code for existing web services.

Advanced tools and technologies such as rules

and ontologies are utilized to automate the

negotiation process. Bartolini proposed in (Bartolini,

2002) a simple interaction protocol to support any

mechanism that can be described with a taxonomy

of predefined declarative rules A similar work based

on Bartilini’s approach uses agent technologies and

tools to provide an initial implementation in (Badica

et. al, 2006). Tamma in (Tamma et. al., 2005) also

uses this taxonomy of rules utilizing ontologies to

make the representation of the rules of encounter

explicit, machine readable and sharable; agents

willing to participate to a negotiation session commit

to the shared ontology, which represents the

mechanisms governing the negotiation.

A very interesting approach for a general

description of processes and protocols is OWL-P

presented in (

Desai et. al., 2004). Ontology Web

Language (OWL) and Semantic Web Rule

Language (SWRL) are used to specify interactions

as rule based commitment protocols. The authors

separate public protocols from private policies thus

allowing protocols to be easily reused and extended.

3 ARCHITECTURE OF THE

PROPOSED FRAMEWORK

The layered architecture shown in figure 1 was

defined to favour flexible, customizable and

automated e-negotiations.

The bottom layer is the basic interaction protocol

that provides the basic required functionality for e-

negotiations. This protocol based on existing Web

Service protocols encapsulates well defined in agent

standards communication acts. The description of

SWS, which will implement the proposed

negotiation scheme is given in OWL-S (language for

describing Web Services) (

http://www.daml.org/

services/owl-s/). Messages exchanged among

negotiating participants are a subset of the Fipa

agent interaction patterns (http://www.fipa.org/

repository/index.html) and are formulated using an

ontology.

The basic negotiation protocol, in the second

layer, is a rule based protocol specification. Rules

are used to define and constraint the context of the

interaction for the participants. These rules are

expressed with the use of concepts from the well

defined negotiation domain. Constraints can be

broadly classified as either enabling or limiting.

They provide all the appropriate freedom within

predefined rules for the participants to behave

according to their strategy and achieve their goals.

Figure 1: The proposed Architecture.

The third layer uses the same concepts to define

a specific negotiation scheme in consistency with

layer two constraints. Rules are also used but a way

to describe interaction processes is a key issue for

the negotiation scheme description.

In the fourth layer, i.e. the negotiation strategy

layer, private strategies are defined for each

participant. These strategies are based on the

specific negotiation scheme that was adopted for the

interaction. The description of strategies will again

use the same tools and vocabulary forming a flexible

and easy to understand negotiation architecture.

The specification of the basic negotiation

protocol is a difficult task since it has to provide the

infrastructure required by all possible general

interactions that any negotiation may use. Such a

specification may have the form of transition

diagrams (stds) to fully describe the states of the

protocol. This is the approach used in (Su et al.

2000) where an FSM was used to model bilateral

bargaining negotiations. Another approach adopts a

rule-based framework to define the appropriate

behaviour according to the protocol that these rules

specify. Rules in (Bartolini et. al. 2002) are used for

enforcing the negotiation mechanism and are

organized into a taxonomy. Rules compared to STDs

have the advantage of giving simple, easy to

understand and modify, description of the protocol

specification. The same protocol needs a large

number of states and very complicated description if

expressed using STDs. This is why rules and

constraints are our preposition for the basic

TOWARDS A FRAMEWORK FOR AUTOMATED E-NEGOTIATIONS

169

negotiation protocol that is used as basis for the

above layers to further specialise the negotiation

process.

4 THE NEGOTIATION PROCESS

According to our architecture the involved

negotiating parties, i.e., a consumer and a merchant

should follow a well defined interaction process,

depicted in figure 2. According to this the Consumer

may use a predefined negotiation scheme or define

his own. Well defined knowledge bases are

exploited to select a predefined negotiation scheme

(1A), while model- driven development techniques

are used to aid the development process of a new

scheme (1B). Basic protocols and named negotiation

schemes, such as the English and Dutch auctions,

are already defined. After selecting the scheme an

appropriate negotiation strategy can be adapted or a

new one constructed (2). Fore the case a new

scheme is created, this will be published to the

knowledge base (3).

At this stage the consumer is ready to negotiate

with the selected scheme the location of which

(URI) he sends to the merchant (4). The merchant

acquires the OWL description of the negotiation

scheme (5) and designs a strategy for it (6). After

that both parties automatically generate a SWS

(Semantic Web Service) described in OWL-S based

on OWL description of the negotiation scheme (7)

and the negotiation between SWS begins (8).

Merchant can publish the negotiation service in a

semantic annotated UDDI with references to all

available negotiation schemes (9).

In the context of the proposed architecture: a)

semantics offer machine understandable concepts,

and b) XSLT-based OWL to OWL-S transformation

offers the ability to automate the whole process after

the negotiation scheme and strategy selection.

5 MODELING E-NEGOTIATIONS

5.1 Modeling the Negotiation Domain

In the e-negotiation context there is need for a basic

vocabulary that can be understood and used by all

participating agents. Such a vocabulary will be part

of the modelling of the e-negotiation domain which

is a very complicated task. Several attempts were

made to this direction each with a different point of

view of the domain. In (Wurman et al., 2001) the

focus is on auctions, whereas in (Lomuscio et al.,

2002), they stresses on automation aspects.

We adopted the Montreal Taxonomy (Strobel

and Weinhardt, 2002)

which can be considered an

abstract collection of the most important views

which provide a framework that can be used for

descriptive and prescriptive purposes. It covers

aspects such as the negotiating participant roles,

processes (offer specification, offer submission,

offer analysis, offer matching, offer allocation, offer

acceptance), information revelation, and business

model implementation of e-negotiations.

Figure 2: The negotiation process.

Based on the Montreal Taxonomy a negotiation

ontology was developed to exploit the reasoning

capability that is provided through a generic

reasoning engine, a great advantage of an ontology

compared to a simple taxonomy. Moreover,

inference can be used as an extension to the

proposed framework to draw conclusions which will

drive decisions in strategy and negotiation scheme

selection.

The Negotiation ontology captures, except from

the important concepts of the domain, the well-

known named negotiation schemes like English

auction, Dutch auction and many others. For the

description of complicated negotiation schemes,

such as bargaining, other ontologies are also

required, as for example the time ontology or the

ontology describing the item to be negotiated.

Moreover, a message ontology was defined

exploiting the Fipa ACL (Agent Communication

Languages) messages, to represent the messages that

can be exchanged between participants.

Among the classification criteria in the

negotiation area modelling we have adopted, we

discriminate: the number of participants, the bid

privacy, the number of items to be negotiated, the

attributes of negotiated items, the number of

suppliers, and the use of a broker or mediator.

Participants have to execute, during a

negotiation, certain well-defined actions. For

example every participant in any kind of negotiation

has to process the received proposal and decide his

ICE-B 2007 - International Conference on e-Business

170

next action based on the negotiation protocol and the

adopted strategy. We have collected the most

important actions and depicted them as concepts in

our ontology in an attempt to describe internal

behavior in each phase of the negotiation for each

participant. Among the concepts we discriminate:

Prepare Preference, Send, Receive, Prepare

Proposal, Make Choice, Matchmaking-Critic etc.

Negotiations depend a lot on the modeling of the

item that is under negotiation, since different

approaches can be adopted depending on the

negotiated-item. For example, negotiation schemes

for multi-attribute items can become very

complicated. An agreement on the negotiated object

properties that should not be altered, such as

minimum and maximum values of properties and

flexible or “don’t care” properties, has to be

established in an early step. Participants may express

these constrains upon the concepts of the negotiated

item ontology.

Every negotiation can be modeled as a peer-to-

peer interaction. Based on that, we decided to model

the negotiation process as peer-to-peer interaction in

flexible way that can be extended.

5.2 Modeling Interactions

The above defined representation of the negotiation

domain captures the static view and not the dynamic

view that is very important in interactions. The FSM

ontology was defined to address this requirement.

The most important concepts of the UML FSM

notation were used to describe the negotiation

process as an FSM diagram. Such an attempt that led

to an FSM ontology based on UML FSM notation is

also presented in Dolog (2004).

Negotiating participants go through a process

that can be described as a sequence of states. Each

state describes the exact phase of the negotiation.

During a state one or more activities are performed

by each negotiating party. Usually one participant

performs some activity while the other waits for an

event to occur. The description of the activity will

come from the negotiation ontology. Entry actions

of states can be used to perform the setup needed

within a state, as for example the check for validity

in an incoming message. Exit actions can be used

for the required clean up before exiting the state.

Transitions can also have actions which usually

produce a message from the message ontology for

the waiting participant. The transition guards contain

SWRL rules in order to create Boolean Expressions

for the firing of transitions. It should be noted that

the Completion Transition has no explicit trigger

event but it is fired by the completion of the

activities of the current state.

FSM notation has already been used for

modeling the negotiation process domain (Kumar,

1998 - Su, 2000). In (Kumar, 1998) a very simple

representation is presented trying to catch only basic

states in an English auction negotiation mechanism.

In (Su, 2000) again the possible phases of the

negotiation are represented as FSM states and the

transitions between states are fired from “send” and

“receive” proposals. The diagram produced is more

complex and uses numerous states trying to catch

the internal behaviour for each participant for the

bilateral bargaining negotiation scheme. Our

approach uses: a) a rule-based basic negotiation

protocol to give the desired generality, and b) the

FSM notation for specific scheme description able to

catch all important details using well defined

semantics.

Benyoucef and Rinderle stating the FSM

limitations propose Statecharts due to their

advantage in executability, popularity and

completeness. The Process Specification Language

(PSL-

http://www. mel.nist.gov/psl/) provides another

alternative for modeling interactions. PSL was used

for this purpose in (Tamma et al., 2005) although the

applicability is not proven.

5.3 Automated Generation of Stateful

Web Services

One important issue towards automation is the

translation of the OWL document describing the

overall negotiation process, to an OWL-S document

that will give the semantic description of the SWS

created to handle the negotiation for each party.

What is needed is to translate and map between

XML-based (OWL and OWL-S) documents. For

this, XSLT will be used to grasp all appropriate

information from the OWL document and generate

the OWL-S SWS description. In this way our engine

will only have to be in position of using a simple

XSLT engine and understanding and using the well

known ontology concepts.

Web services, by their nature, typically do not

maintain state information during their interactions.

However their interfaces must frequently allow for

the manipulation of state, that is, data values that

persist across and evolve as a result of SWS

interactions. Especially in our case a way of

remembering previous proposals must be

implemented. We could leave this job to the private

negotiating engine but selecting stateful resources is

another way of taking burden from the internal

proprietary engine of each party and injecting it to

TOWARDS A FRAMEWORK FOR AUTOMATED E-NEGOTIATIONS

171

the automated web service interaction. Stateful

resources will be mapped to state variables used in

FSM process ontology.

6 AN EXAMPLE NEGOTIATION

SCHEME

An example negotiation scheme for acquiring an

Service Level Agreement (SLA), for an internet

access service, from an ISP is used to demonstrate

the applicability of our approach. The example is a

“small” auction that allows service providers to

Figure 3: Example scheme interaction.

improve their proposal only ones before the final

decision is made (figure 3). Before the actual

propose-counter_propose process begins, the client-

initiator sends his preference which is the ability of

the provider to guarantee the end-to-end Quality of

Service (QoS). The SLA and QoS ontologies that

were imported in our framework during the design

of the example negotiation scheme are utilized along

with SWRL for describing constraints upon

concepts. After the confirmation the client sends the

complete preferred SLA as an SLA-ontology

instance asking for the providers to make their

suggestions. Involved ISPs answer with their

preposition and the client sends back the best choice

after comparing. An improved preposition is allowed

to be send by ISPs before the small auction is

terminated.

Figure 4 presents the example negotiation

scheme using an FSM diagram with concepts from

fsm, negotiation and SLA ontologies. It is the

representation of the OWL document describing the

whole interaction. There we can see how each state,

transition, event, action etc bears the appropriate

semantic annotation in order to be easily recognized

and translated by the engine of the participants. In

each state a fundamental activity or activities from

the negotiation ontology is given that must be

performed in this phase of the negotiation by

participants. For this simple example in each state

only one participant is performing an activity while

the other is in the wait state. Each transition has also

actions which is usually the Send action of a

message between participants.

For example the first state is annotated with

State: Initial from the FSM ontology and Role:

initiator from the negotiation ontology meaning that

it is the starting state and the initiator performs an

activity during it while Negotiation: Role:

Participant is at wait state. The following transition

bares an Action: Send from the Initiator which is a

Start negotiation message to participant. In the

second simple state the Initiator executes one

Activity: Prepare_ Preference for the Service: End-

to-End QoS concept from the Service ontology. In

order to enter the second state our Guard: Condition

must be true which is an Event: Receive: Start

message from participant. In the second transition a

Query_if is sent by the Initiator-client about the

discussed concept and we enter the third state when

this preference is received by the Participant-ISP.

There the Participant performs a Matchmaking

Activity for the particular concept and answers with

a Confirm or Disconfirm message in the following

Transition. After another Prepare_Preference for a

complete SLA instance which is send to Participant

with a Call_for_Proposal message, we enter the

states where they exchange Proposals and Counter-

Proposals in Composite states where more than one

activity is performed. At the end the End state with

all appropriate messages follows the

Accept_Proposal message. In the particular figure

only some basic concepts are depicted from the

selected ontologies in order to keep it simple and

readable.

7 CONCLUSIONS

In this paper an approach to automate negotiations

between machines acting on behalf of their users has

been proposed.

An FSM and a negotiation ontology were defined

and utilized to construct negotiation schemes that

can guide the interaction of negotiating parties that

has no previous knowledge of the negotiation

scheme. This entire infrastructure can be used along

with semantic web services to automate the

generation process of negotiating interface of each

participant.

ICE-B 2007 - International Conference on e-Business

172

ACKNOWLEDGEMENTS

This work is funded by the Greek General

Secretariat for Research and Technology in the

context of PENED 2003 03ED723 project, (75%

EC, 25% Greek Republic, according to 8.3, 3

Framework programme).

REFERENCES

Badica C., Ganzha M., Paprzycki, 2006. Rule-Based

Automated Price Negotiation: Overview and

Experiment. ICAISC: 1050-1059.

Bartolini C., Preist C., Jennings N.R., 2002. Architecting

for Reuse: A Software Framework for Automated

Negotiation, in Giunchiglia, F., Odell, J., Weiss G.

(Eds.): Agent-Oriented Software Engineering III,

Springer-Verlag LNCS 2585/2003.

Bichler, M., G. Kersten, and Strecker S., 2002.

Engineering of Negotiations. Group Decision and

Negotiation, Submission for this same issue.

Benyoucef M., Rinderle S., 2006. Modeling e-Negotiation

Processes for a Service Oriented Architecture. Group

Decision and Negotiation 15: 449–467.

Chiu, D.K.W., S.C. Cheung, P.C.K. Hung, S.Y.Y. Chiu,

and A.K.K Chung, 2005. Developing e-Negotiation

Support with a Meta-Modeling Approach in aWeb

Services Environment. International Journal on

Decision Support Systems. Special Issue on Web

Services and Process Management 40(1), 51–69.

Dolog P., 2004 Model-Driven Navigation Design for

Semantic Web Applications with the UML-Guide. In

Maristella Matera and Sara Comai (eds.), Engineering

Advanced Web Applications

Desai nj., Ashok U. Mallya, Amit K. Chopra, Munindar P.

Singh. 2005. OWL-P: A Methodology for Business

Process Modeling and Enactment. In Proceedings of

the AAMAS 2005 Workshop on Agent Oriented

Information Systems.

Kersten, G., K. P. Law, and S. Strecker. 2004. A Software

Platform for Multi-Protocol E-Negotiations. Available

at http://interneg.org.

Kim, J. B. and A. Segev. 2005. A Web Services-Enabled

Marketplace Architecture for Negotiation Process

Management. Decision Support Systems. Special Issue

on Web Services and Process Management 40(1), 71–

87.

Kumar, M. and S. I. Feldman. 1998. Business negotiations

on the Internet. Technical Report, IBM Research

Division, New York.

Lomuscio, A.R., Wooldridge, M., Jennings, N.R. 2002. A

classification scheme for negotiation in electronic

commerce. Agent Mediated Electronic Commerce: The

European AgentLink Perspective, LNCS 1991,

Springer Verlag 19–33.

Strobel M., Weinhardt C., 2002. The Montreal Taxonomy

for Electronic Negotiations.

Su S., C. Huang, and J. Hammer., 2000 A replicable web-

based negotiation server for ecommerce. In Proc. of

33rd Intl. Conf. on System Sciences, Hawaii.

V. Tamma, S. Phelps, I. Dickinson, and M. Wooldridge.

2005. Ontologies for supporting negotiation in e-

commerce. Engineering applications of artificial

intelligence, 18(2):223-236.

Wurman, P., M.Wellman, and W.Walsh. A

Parameterization of the Auction Design Space. Games

and Economic Behaviour, 35:304–338, 2001.

Figure 4: The example negotiation scheme.

TOWARDS A FRAMEWORK FOR AUTOMATED E-NEGOTIATIONS

173