HP Laboratories Position Statement

FIPA Second Meeting, Yorktown, NY, June 1996

Background

Our project is a relative newcomer to the field of agent research. Though HP was involved in very early agent technology [], this work has not been actively continued within HP Labs. We have recently begun an investigation into the role of agents in mediated communications systems. Along with many others, we hypothesise the continued rapid growth and integration of mobile communications and computational devices, with radical changes in the form factor, pervasiveness and functionality of such devices. Our lab is investigating the new services and capabilities that will be enabled by such technologies. Our view is that the agent metaphor will provide a powerful approach to solving problems of: service description, service delivery, human to human computer-mediated communications, robust network construction and maintenance and adaptability to changes in the environment.

Current activities

Our department has a number of mediated communications activities underway. Amongst these are two agent related activities:

• an investigation of negotiation protocols for task-distributing agents
• development of an agent-based communications platform for mobile communications

Areas of interest

The following agent capabilities are of specific interest to us:

• large communities of heterogeneous agents, interacting with each other, underlying network resources and hardware, and human users
• a spectrum of cognitive capabilities, from simple, reactive agents closely related to underlying hardware, to complex abstract (possibly themselves multi-agent) entities representing advanced services and products
• specialisation of generic agents and generic system behaviours to individual user preferences
• agents that are robust to incomplete or erroneous knowledge of time-varying worlds
• dynamic extensibility of complex, distributed services

Views on standardisation

A number of activities are underway in various groups that are developing component-based architectures for general agent platforms. Given the commonality of this approach, standardising the semantics of components, and of the operators for composing and specialising them, would encourage researchers to develop and distribute components and results. Commercial development of standard components would also follow in time.

Given our interest in heterogeneous communities, we would be interested in the standardisation of agent interaction protocols, languages, and tasks. However, we also recognise the number of basic open research questions that remain, and hence the need to retain flexibility and openness to new approaches. Striking the balance between expressive freedom, and sufficient specificity to be useful will be an ongoing challenge.

Author

Ian J. Dickinson
Hewlett-Packard Laboratories
Filton Rd
Stoke Gifford
Bristol BS12 6QZ
UK
Tel: +44 117 922 8796
Email: ijd@hplb.hpl.hp.com