The April Agent Platform provides a FIPA-compliant agent platform using the April programming language and InterAgent Communications Model (ICM). It features FIPA parsers and specifications as well as platform services like the Agent Management System and Directory Facilitator. The platform is organized with modules for different protocols that communicate through ICM. It is being used in several projects including information services for cities.
The role of citizen science in environment management
The April Agent Platform
1. The April Agent Platform
Features and Summary
Dr Luigi Ceccaroni
Dr Jonathan Dale
Network Agent Research Group
Fujitsu Laboratories of America
John Knottenbelt
Department of Computing
Imperial College of Science, Technology and Medicine
February, 2002
AAP Group composition: jono, tofu, collaborative work with jak, from Imperial (IC)
NAR: research group working in smart services: services on the Web: Agentcities (AC), use of formal representations in contracts and agreements (network of trusts) above the AC services. Agent technology: our own platform: AAP: a research agent-deployment environment, used for example for the SF Agentcity.
AAP is conformant to the FIPA-2000 standard, that is all about interoperability among agents. It includes white pages and yellow pages, for example. Experimentation with the AAP can result in something not always applicable to all environments. Suitable for long term agent deployment. ICM is for agents what email is for humans, a message transport system.
April was developed by FLA and IC. designed for multiple execution environments small core plus basic macros; high flexibility in language design (higher order in the functional sense) Prolog without backtracking and with restrictions applicable to pattern matching no shared memory, no concurrent things going on ICM is the core of message passing. it takes the data structure and encodes terms in a suitable format for transmission over the Internet ICM servers are needed on each machine; they work asynchronously: they fire and forget; there is no order in message transmission, but messages can then be selected-filtered thru pattern matching; in this way the system is not slowed down by slow agents.
In pink are things taken from FIPA apart from the ICM transport model. In light blue, are our additions used for projects like AC. On the top row: parsers Mid row: services: AMS = agent management system MTS = message transport system Bottom row: PQA: allow to query any remote AMS and DF
Bottom part (from client libraries): internal structure Top part: what is presented to the external world In blue is the FIPA speaking part The basic idea is of services wrapped with FIPA. This allows to integrate legacy systems into an agent environment without changing the old system We have this wrapping for FIPA, but we could have it for any other web-service environment. It is easy to retarget to different environments.
How the AAP communicates with the outside world The ICM is designed to be modular On the bottom are the agents. In blue are the agents which speak FIPA. In the upper part there are the modules that the ICM needs to communicate with other transport systems: XML protocol is “extended SOAP” The general idea is that the ICM abstract the transport system from the agents.
The event organizer is an example of service composition, like hotel-transportation-restaurant-transportation-cinema-transportation-hotel. IC is using the AAP for one of the London Agentcities. Pervasive computing to retrieve information that is then used for personal profiling in the Agentcities environment.