A Service-Oriented Negotiation Model between
Autonomous Agents
Carles Sierra
, Peyman Faratin, Nick R. Jennings
Dept. Electronic Engineering, Queen Mary and Westfield College,
University of London, London E1 4NS, UK.
{C.A.Sierra,P.Faratin,N.R.Jennings}@qmw.ac.uk
Abstract. We present a formal model of negotiation between au-
tonomous agents. The purpose of the negotiation is to reach an agreement
about the provision of a service by one agent for another. The model de-
fines a range of strategies and tactics that agents can employ to generate
initial offers, evaluate proposals and offer counter proposals. The model
is based on computationally tractable assumptions and is demonstrated
in the domain of business process management. Initial proofs about the
convergence of negotiation are also presented.
1 Introduction
Autonomous agents are being increasingly used in a wide range of industrial and
commercial domains [2]. These agents have a high degree of self determination –
they decide for themselves what, when and under what conditions their actions
should be performed. In most cases, such agents need to interact with other
autonomous agents to achieve their objectives (either because they do not have
sufficient capabilities or resources to complete their problem solving alone or
because there are interdependencies between the agents). The objectives of these
interactions are to make other agents undertake a particular course of action
(e.g. perform a particular service), modify a planned course of action (e.g. delay
or bring forward a particular action so that there is no longer a conflict), or
come to an agreement on a common course of action. Since the agents have
no direct control over one another, they must persuade their acquaintances to
act in particular ways (they cannot simply instruct them). The paradigm case
of persuasion is negotiation – a process by which a joint decision is made by
two or more parties. The parties first verbalise contradictory demands and then
move towards agreement by a process of concession making or search for new
alternatives, (cf. [3]).
Given its pervasive nature, negotiation comes in many shapes and forms.
However in this work we are interested in a particular class of negotiation –
On sabbatical leave from Artificial Intelligence Research Institute –IIIA, Span-
ish Council for Scientific Research –CSIC. 08193 Bellaterra, Barcelona, Spain.
sierra@iiia.csic.es. With the support of the Spanish Ministry of Education grant
PR95-313.
namely service-oriented negotiation. In this context, one agent (the client) re-
quires a service to be performed on its behalf by some other agent (the server)
2
.
Negotiation involves determining a contract under certain terms and conditions.
The negotiation may be iterative in that several rounds of offers and counter
offers will occur before an agreement is reached or the negotiation is terminated.
When building an autonomous agent which is capable of flexible and sophisti-
cated negotiation, three broad areas need to be considered [7] – what negotiation
protocol will be used?, what are the issues over which negotiation takes place?,
and what reasoning model will the agents employ? This paper concentrates pre-
dominantly on the final point although the protocol and negotiation object are
briefly defined. A comprehensive reasoning model for service-oriented negoti-
ation should determine: which potential servers should be contacted, whether
negotiation should proceed in parallel with all servers or whether it should run
sequentially, what initial offers should be sent out, what is the range of acceptable
agreements, what counter offers should be generated, when negotiation should
be abandoned, and when an agreement is reached.
This paper presents a formal account of a negotiating agent’s reasoning com-
ponent –in particular it concentrates on the processes of generating an initial
offer, of evaluating incoming proposals, and of generating counter proposals.
The model specifies the key structures and processes involved in this endeavour
and defines their inter-relationships. The model was shaped by practical consid-
erations and insights emanating from the development of a system of negotiating
agents for business process management (see [5] and Section 2 for more details).
The main contributions of this work are: (i) it allows rich and flexible negotia-
tion schemes to be defined; (ii) it is based on assumptions which are realistic for
autonomous computational agents (see Section 3.2 for the set of requirements
and Section 7 for a discussion of related approaches); and (iii) it presents some
initial results on the convergence of negotiation (see Section 6).
In this paper we concentrate on many-parties, many-issues, single-encounter
negotiations with an environment of limited resources (time among them). Sec-
tion 2 gives details of the type of applications and scenarios we are interested in.
Sections 3 to 5 present the proposed model. Finally, some results on negotiation
convergence and future avenues of work are outlined.
2 Service-Oriented Negotiation
This section characterises a context in which service oriented negotiation takes
place. The scenario is motivated by work in the ADEPT project [5] which has de-
veloped negotiating agents for business process management applications. How-
ever, we believe that the characteristics emerging from this domain have a wide
variety of application. To provide a detailed context for this work, a multi-agent
2
A service is a problem solving activity which has clearly defined start and end points.
Examples include diagnosing a fault, buying a group of shares in the stock market,
or allocating bandwidth to transmit a video-conference.
Vet_Customer
Cost_&_Design_Customer_Network
Legal_adviceSurvey_Customer_Site
Provide_Customer_Quote
Customer Service
Division Agent
(CSD)
Design
Department
Agent
(DD)
Legal
Department
Agent
(LD)
Surveyor
Department
Agent
(SD)
Sales Agent
Vet
Customer
Agents
(VC)
system for managing a British Telecom (BT) business process is presented (sec-
tion 2.1). This scenario is then analysed in terms of its key characteristics and
assumptions as they relate to the process of negotiation (section 2.2).
2.1 BT’s Provide Customer Quote Business Process
This scenario is based on BT’s business process of providing a quotation for
designing a network to provide particular services to a customer (figure 1)
3
. The
overall process receives a customer service request as its input and generates as
its output a quote specifying how much it would cost to build a network to realise
that service. It involves up to six agent types: the sales department agent, the
customer service division agent, the legal department agent, the design division
agent, the surveyor department agent, and the various agents who provide the
out-sourced service of vetting customers.
Fig. 1. Agent system for BT’s provide customer quote business process.
3
The negotiations between the agents are denoted by arrows (arrow head toward
client) and the service involved in the negotiation is juxtaposed to the respective
arrow.
The process is initiated by the sales agent which negotiates with the CSD
agent (mainly over time, but also over the number of invocations and the form in
which the final result should be delivered) for the service of providing a customer
quote. The first stages of the Provide
Customer Quote service involve the CSD
agent capturing the customer’s details and vetting the customer in terms of their
credit worthiness. The latter sub-service is actually performed by one of the VC
agents. Negotiation is used to determine which VC agent should be selected – the
main attributes which are negotiated over are the price of the service, the penalty
for contract violation, the desired quality of the service and the time by which the
service should be performed. If the customer fails the vetting procedure, then the
quote process terminates. Assuming the customer is satisfactory, the CSD agent
maps their requirements against a service portfolio. If the requirements can be
met by a standard off-the-shelf portfolio item then an immediate quote can be
offered based on previous examples. In the case of bespoke services, however, the
process is more complex. The CSD agent negotiates with the DD agent (over time
and quality) for the service of designing and costing the desired network service.
In order for the DD agent to provide this service it must negotiate with the LD
agent (over time) and perhaps with the SD agent. The LD agent checks the design
to ensure the legality of the proposed service (e.g. it is illegal to send unauthorised
encrypted messages across France). If the desired service is illegal, then the entire
quote process terminates and the customer is informed. If the requested service
is legal then the design phase can start. To prepare a network design it is usually
necessary to have a detailed plan of the existing equipment at the customer’s
premises. Sometimes such plans might not exist and sometimes they may be out
of date. In either case, the DD agent determines whether the customer site(s)
should be surveyed. If such a survey is warranted, the DD agent negotiates with
the SD agent (over price and time) for the Survey
Customer Site service. On
completion of the network design and costing, the DD agent informs the CSD
agent which informs the customer of the service quote. The business process
then terminates.
The structure of the negotiation object is based almost directly on the legal
contracts used to regulate agreements in the current manual approach to busi-
ness process management. This structure is fairly rich and covers both service
and meta-service attributes. In more detail, it contains: (i) the service name; (ii)
a unique agreement identifier (covering the case where there are multiple agree-
ments for the same service); (iii) the agents involved in the agreement (client
and server); (iv) the type of agreement (one off agreement for a single service
invocation versus on-going agreements for multiple invocations of the same ser-
vice); (v) timing information (duration represents the maximum time the server
can take to finish the service, and start time and end time represent the time
during which the agreement is valid); (vi) the volume of invocations permissible
between the start and end times (for on-going agreements only); (vii) the price
paid per invocation; (viii) the penalty the server incurs for every violation of the
agreement; (ix) the information the client must provide to the server on service
invocation; and (x) the policy used for disseminating the service’s intermediate
and final results to the client.
2.2 Characteristics and Assumptions
The following negotiation characteristics can be noted from the ADEPT business
process scenario. Moreover, it is believed that these characteristics are common
to a wide range of service oriented negotiations between autonomous agents.
– A given service can be provided by more than one agent (e.g. multiple agents
can provide the vet customer service to the CSD agent). The available ser-
vices may be identical in their characteristics or they may vary along several
dimensions (e.g. quality, price, availability, etc.).
– Individual agents can be both clients and servers for different services in
different negotiation contexts.
– Negotiations can range over a number of quantitative (e.g. price, duration,
and cost) and qualitative (e.g. type of reporting policy, and nature of the
contract) issues. Each successful negotiation requires a range of such issues
to be resolved to the satisfaction of both parties. Agents may be required to
make trade-offs between issues (e.g. faster completion time for lower quality)
in order to come to an agreement.
– The social context and inter-relationships of the participants influences the
way agents negotiate. Some negotiations involve entities within the same
organisation (e.g. between the CSD and DD agents) and hence are gener-
ally cooperative in nature. Other negotiations are inter-organisational and
purely competitive – involving self interested, utility maximising agents (e.g.
between the VC agents and the CSD agent). Some groups of agents often ne-
gotiate with one another for the same service (e.g. the CSD and DD agents),
whereas other negotiations are more open in nature (for example, the set
of VC agents changes frequently and hence the CSD agent often negotiates
with unknown agents).
– As the agents are autonomous, the factors which influence their negotia-
tion stance and behaviour are private and not available to their opponents
(especially in inter-organisational settings). Thus agents do not know what
utilities their opponents place on various outcomes, they do not know what
reasoning models they employ, they do not know their opponent’s constraints
and they do not know whether an agreement is even possible at the outset
(i.e. the participants may have non-intersecting ranges of acceptability).
– Since negotiation takes place within a highly intertwined web of activity
(the business process) time is a critical factor. Timings are important on
two distinct levels: (i) the time it takes to reach an agreement must be rea-
sonable; and (ii) the time by which the negotiated service must be executed
is important in most cases and crucial in others. The former means that
the agents should not become involved in unnecessarily complex and time
consuming negotiations – the time spent negotiating should be reasonable
with respect to the value of the service agreement. The latter means that the
agents sometimes have hard deadlines by which agreements must be in place
