IDA (INTELLIGENT DECISION ADVISOR)


<- PREVIOUS Chapter

4. Project Tasks ( and the functioning of the IDA kernel)

The suggested allocation of the tasks to the project partners , has been done according to the main IDA components presented on the Fig. 3.

4.1 General assumptions related to the architecture of Multi-agente

1. Multi-agente is an intelligent shell of the personoid with role dependent: preferences, knowledge and with its own information about its domain of activity..

2. Role of a Multi-agente can be changed during the system set-up session.

3. Basic elements of Multi-agente is a monad;

4. Project objective includes realization of Multi-agente as a simple personoid , therefore

5. Multi-agente is composed of three monads:

6. The Consigliere Diretto (CD) is connected with:

- the general system interface: Interfaccia; for CD interactions with the real-world components and with the IDA end-user, responsible of decisions.

- Simulators system: Simulatori; which serves for the prediction of possible situation on the base of the current date about the emergency domain.

4.2 Main Tasks Lists

Task Nr
Activity
Responsibility,
Main executor
 Cooperating  partners (links)
Comments
T1. Design and implementation of Interfaccia ENEA IRST
- no agents,
- visual interf.
T2. Design and implementation of Simulatori ENEA SAPENZA, IRST
- no agents
- rather qualitative simulations,
T3. Design and implementation of Consigliere Diretto ENEA SAPIENZA,IRST - monad-based architecture
T4.
Design and implementation of
Gestore Preferenze
SAPIENZA ENEA - rather monad-based architecture framework is prefered
T5. Design and implementation of Pianificatore IRST ENEA
- rather monad-based
architecture framework is prefered
T6. Elaboration of a simple Test-case, and set-up ENEA IRST, SAPENZA
- based on the one of the MUSTER managerial roles
- incremental application
T7. Integration and Testing of Multi-agente ENEA SAPIENZA, IRST -
T8. Project Management ENEA - -

T1. Design and implementation of Interfaccia

The Interfaccia module may be developed in Visual Basic or in IlogView , supported by MS Access.

For the prototype, only its selected functions must work, others can be simulated in fthe orm of adequate windows.

The elaboration of the demo of Interfaccia, requires:

ST1.1 Elaboration a generic scenario of the interactions between the end-user and Multi-agente.

ST1.2 ...

. . . .

- Interfaccia must have access to the DB, PB, MKB, and OKB bases which will be specifie in task T3, ST3.2.

T2. Design and implementation of Simulatori

This task should develope some simplest What-if model-based reasoning.

In the future version of IDA, this module can include more sofisticated consequience estimating reasoning engines, as well numeric simulations of specific events.

In general, MDB of Simulatori should be more complex than this one which is used by Multi-agente.

In this project, MDB and SDB from the task T3. will be sufficient.


T3. Design and implementation of Consigliere Diretto, CD

This ENEA tasks includes unique monad which interacts with the real world, the next two support its functioning. Consigliere Diretto is the main element of the Multi-agente.

  • Functional behavior of CD, an elementary loop
  • - The CD is activated by a signals or information from its external real domain (RD) of activity.

    - This information modifies the current state of the Situation Data Base, SDB (in the abstract Domain System, DS) of the monade. - An example of the O-based conceptualization of current state is on the Fig.5.

    - This new state activates the preference system (PS) of the monad.

    - The preference system, using its preference rule base, confronts possible states of the Real Domain, and , if it is possible, it indicates a state of the domain in SDB, which has maximal preference.

    This state is called intervention-goal.

    - If any preference-rule is not found then it doesn't react on the new state of the domain

    - The intervention goal activates Knowledge System (KS), the next component of the monad.

    - The Knowledge system searchs in its Operational Knowledge Base (OKB) an procedure/rule represented an action available (for the predefined role of the monad) , which is able to transform the current state of the Situation Data Base in the state defined by intervention-goal.

    The procedures/rules in OKB which represent possible actions are called methods.

    - If the adequate method is found then, for its activation, the conditions, requested from SDB, by the method are mached with the current state.

    - If the method satisfied the goal and current state conditions, then KS sends this information to the DS. The state of SDB is changing and the system user is informed which intervention is suggested.

    The structure of a monad is described in the document "IDA: General Fuctional and Structural Requirements".
    The above presented flow is very simple and not yet realistic.
     Fig.5.
  • Internal situations and events
  • The monade should take under considerations the following situations and internal events don't present in the elementary loop:
    - new external data is not identified by DS
    - new local state can be a cause of another local state, in the current situation of the domain, which is more important for the emergency-management goal.
    - more than one intervention-goal has been found.
    - not all attributes/attribute-values requested by the chosen method is available in SDB.
    - more than one method satisfy current state and intervention-goal.
    All this problem should be resolved by one or more DS engines.

    The functions of the next two monads: Gestore Preference and Pianificatore are to modify CD's Preference Base, and OKB, subsequently if Consigliere Diretto is not able either to select or to achive intervention goal or to find a proper method.

  • Subtasks
  •  
    The following subtasks are expected.

    ST 3.4 Incremental testing of the CD functioning starting from a simplified versions of the test-case.

    Go to -> T4. T5. . . Tasks