Extending Loosely Coupled Federated Information Systems Using Agent Technology - Distributed Artificial Intelligence, Agent Technology, and Collaborative Applications - page 126

o Instantiation of a new IC agent that is unlike any other exiting IC agent

• Adjustments to existing IC agents to interact with a new data source without having to create a

new IC agent

• Provision of agent creation and instantiation mechanisms (downloadable .jar files, .zip files, or

.xml files defining the agent-creating criteria) and code templates. If it is determined that there
exists no IC agent that can interact with the new data source, EMA will initiate a dialog with the
user informing the need to create a new IC template programmatically. The EMA can assist the
user in this process by providing the skeletal templates needed to create the new IC agent.

It is the EDF that informs the EMA whether existing IC agents can be modified or a new IC agent
has to be created. The EDF uses inference mechanisms by exploiting the ontology information from
the OWL-S document to make this decision. If the EDF determines that an existing IC agent can be
adjusted, it communicates all necessary reconfiguration information to the EMA, which in turn ex-
ecutes the necessary changes. The EDF will use the OWL-S document to infer the threshold at which
dynamic reconfiguration of an existing IC agent will no longer be able to cater to the changes made to
the data source (Figure 3).
Since the location of the templates and the source-code archives will also be stored in the OWL-S
document managed by the EDF, all necessary files can be made available to the EMA using Java Web-
start technology or compressed and archived file formats (such as .gz, .tar, etc.) downloadable from a
secure URI location.
Utility in the Running Example: In the FIS environment where the loan officer is currently operat-
ing, two IC agents are already active. Let IC edb represent the agent interacting with the employment
data source, and IC eb be the agent for the bank. The IC agents can query the local data sources using
their corresponding domain ontologies. For example, financial institutions can use any of the open
Web financial standards (OFX, XBRL, FIXML, SwiftML, FpML, etc.) for semantic markup that can
be interpreted by the IC agent. For a stock investment firm that newly joined the federation, the EDF
uses inference drawn from the firm's service-definition URI to determine if an existing IC agent can
accomplish the task or a clone has to be instantiated to interface with the data source. The service defi-
nition may show that the financial institution uses standard financial markup tags for data description
and therefore informs the EMA to create an additional agent, IC finvst , similar to IC eb . In cases where
the EDF finds no IC agents with matching roles and no templates are available to invoke the creation
of a new IC agent instance, the loan officer will be informed of the exception condition. It is evident
that all new IC agents with roles similar to an existing agent, IC f , can be created with minimal human
intervention. This is shown in Table 1 (Row 5).

Figure 3.

create new IC agent

make changes to
existing IC agent

threshold