BT FIPA Report

Application of Intelligent Agents

to Telecommunications

Author(s) : Paul O'Brien, Nader Azarmi, Richard Nicol.

 

Intelligent Systems Research

Advanced Applications and Technology

0. Summary

This proposal describes four applications of intelligent agent technology to telecommunications. Each application area has an associated reference model to identify the key interfaces and provide a framework for standardisation in FIPA. These reference models take into account existing standardisation initiatives.

1.0 Personal Agent Services

1.1 Scenario

Personal agent systems model the interests of users and provide services as and when required. For example, the Meeting Scheduler personal agent organises meetings on behalf of a user. It behaves like an automated personal assistant. Its functionality includes :

1.2 Reference Model

A reference model can be identified to delineate the key interfaces for personal agent systems. Figure 1 illustrates these interfaces and the main system components of a personal agent system.

Persistent information relating to a users personal preferences would be stored in a User Profile (UP). When a user wishes a task to be performed (e.g. set-up a meeting with a number of participants), the relevant task (e.g. meeting scheduler) accesses the users UP. This composite object of task + UP represents a personal agent. The personal agent then interacts with other agents to organise the meeting.

Four interfaces can be identified. Interface A supports interaction between UPs and tasks. Interface B supports the interaction between personal agents. Interface C supports user interaction with tasks. In the case of an internet service this would be via an internet browser, in which case it would have to conform to the hypertext transfer protocol (HTTP). Interface D supports inter-operability between tasks in support of the service offered by a personal agent.

One important issue associated with standardising these interfaces is security. Personal agents represent the interests of individuals and will have access to personal information. Ensuring the confidentiality and integrity of this information is paramount if user confidence in personal agent technology is to be achieved.


Figure 1, Personal Agent Reference Model

The value of standardising these interfaces would be :

2.0 Network Provisioning

2.1 Scenario

A global telecommunications system is a large, massively distributed system. Components of that system are owned by different organisations which inter-operate to ensure a global end-to-end service can be provided. Network provisioning consists of those activities required to resource an end-to-end network connection.

An agent-based network provisioning system would consist of autonomous agents representing the interests of telecommunication providers. The agents buy and sell network capacity in support of end-to-end communication links. The agents would negotiate over parameters such as quality, time and cost, as well as how such links are to be managed. When exceptions occur, the agents co-operate in order to resolve the exception.

There are two areas of agent functionality in this application area :


2.2 Reference Model

Agreed standards already exist covering communications management. The ITU M3010 provides a reference model for telecommunications management which is supported by the telecommunications industry. Network provisioning agents can be incorporated into the existing ITU reference architecture by adapting existing ITU interfaces and by the addition of other interfaces.

Figure 2, provides a possible reference model for agents in communications management. Interfaces Q3* and X* represent enhanced Q3 and X ITU interfaces (cf. ITU M3010 "Principles for Telecommunications Management Network") with existing non-agent management systems.

Interfaces Z and Z* represent new interfaces. Z represents the interface between agents within the same management domain, whereas Z* represents the interface between agents from different management domains. A key difference between these two interfaces would be the inclusion of security safeguards in Z* because it represents an inter-organisational interface.


Figure 2, Reference Model for Agents and Communications Management

3.0 Multi-Media Service Management

3.1 Scenario

Telecommunications services comprise many types of media, (i.e. voice, text, pictures, video, raw data) which can be accessed through a variety of terminal devices, (i.e. telephone, computer, TV etc.) all of which are not totally inter-operable. A Multi-Media Service Management (MMSM) system is capable of delivering a variety of media in the most appropriate form for the recipient. For example, an urgent email being sent to a mobile user could be translated from text to speech and transmitted to the users mobile phone.

Such a system requires :

3.2 Reference Model

Defining a reference model for this application area is difficult due to the variety of telecommunication service platforms an agent based MMSM system would inter-operate with.

Figure 3 outlines a reference model which could be adopted for MMSM. Interfaces B and C are the same as those described in section 1. Service agents manage value-added services such as text-to-speech translation. Network agents manage the access to and use of network resources. Interfaces E and F support access to these services and resources by personal agents.


Figure 3, Reference Model for MMSM

The modules outlined in figure 3 would not exist in isolation but would interact with existing telecommunications service platforms. There are three main bodies which define such platforms which should be related into this reference model, they are :

4.0 Agent-based Business Process Management (APMS)

4.1 Scenario

Organisations are naturally distributed into business units or departments which need to collaborate in order to manage an end-to-end business process. Agent-based business process management systems (APMS) use autonomous agents to represent the interests of business units which negotiate about how to provision and execute business processes.

  1. Reference Model

Figure 4, The APMS Reference Model

The APMS reference model has four key agent interfaces. W supports interoperability between agent based process management systems, X* enables the agent to manage a process task, Y* supports the maintaining and monitoring of agent activities and V* supports interoperability with non-agent based process management systems.

The only standardisation body in the area of business process management is the Workflow Management Coalition (WMC) which has defined its own interfaces. Y*, X* and V* represent enhanced WMC interfaces. FIPA could represent a pressure group pushing for the extension of WMC interfaces to support APMS.

5.0 Conclusions

This document describes key agents application areas in telecommunications. It is proposed that FIPA follows these key steps to standardising both these agent application areas and others :

  1. Select agent application areas.
  2. Adopt and enhance application specifications.
  3. Agree reference models for each application area identifying the key inter-operable interfaces for standardisation.
  4. Define and reach consensus on the interfaces identified.
  5. Once agreement on reference models and interfaces has been reached within each application area, seek consensus where necessary across application domains to support interoperability between application areas.

4.0 Further Information

ITU (International Telecommunications Union)

http://www.itu.ch/

http://www.itu.ch/TMN/

WMC (Workflow Management Coalition)

http://www.aiai.ed.ac.uk:80/WfMC/

TINA : Telecommunications Information Networking Architecture

http://tinac.com.4070/root.html