Transcript
Page 1: AAMAS-2006 TANDEM Design Method (poster format)

The TANDEM Design MethodPaper #112: Steve Goschnick, Sandrine Balbo, Leon Sterling and Christine Sun

Task 1 – Movie-oriented search• Find all recently released movies• Find all movies within a genre• Find which movies feature chosen Star• Find all movies by a chosen Director

Task 2 – Cinema-oriented search

• Show movies playing in a cinema• Show cinemas playing a movie• What type of theatre is movie in

Task 3 – Location-oriented search• What cinemas are within a radius• What distance is a cinema• Show closest cinema with a movie

Task 4 - Seat selection & allocation

• Find a seat’s coordinates in theatre plan• Access availability of a seat• Book a selected seat • Get the cost for a ticket type• Pay for a booking of seat/s

Example Application: The Cinema-Movie Problem

The client – a movie cinema owner – wants a system accessible by movie-going customers via the Internet, one that will allow users of smart hand-held devices, to search for specific movie details, for all oftheir cinema complexes, ultimately leading to the booking of cinema tickets.

A Task Analysis

Data Normalisation

An enhancedAware Ego Agentvia Role C agent

R ole ispartic ipatingat a lesser

priority thanother activities

R ecus ive levelsof Shadowboard

sub-agents

Social World 1 - facilitated by ShadowPlaces

R ole D

R o le CR o le B

R o le A

R oles o f userpa rtic ipa ting

in o ther S ocia lW orlds

Four ro lesw ith in

th is Socia lW orld

Individual Knowledge Management

Impromptu Intelligent InteractionImpromptu Web Service Orchestration

Respect for our most under-valued asset - our attention

The DigitalFriend - - the Human-Centred MASAppropriate levels of filtering of messages.

In this case, differentiatedby the user Role that the

messaging sub-agent is serving.

Source of Figure: Goschnick, S.B & Graham, C. (2005) Journal of UniversalAccess for the Information Society, 4(3).

The user’s Role Hierarchy, expressed

using the ShadowBoardMethodology, a reflection

of the social worlds that theindividual participates in.

Role hierarchy has functionalityin the implemented software,

as well as, in the methodology.

Alerts and notifications ofnewly arrived information,or newly computed data,

can attract the user’s attention, or not,

with a range of visual andaudio cues via ~600 differententities displayed at the one

time (no limit on total #).

The ShadowBoard Agent Architecture is being extended,

into a multiple Human/Agent model, where each one

has the sophisticatedShadowBoard agent model,

em/powering it. An accompanying

methodology is also being developed.

Multiple, individual Web services,together with other sub-agents/agent-types,

including SQL/DB Lookup Agents, canbe combined into new user-envisaged

functionality with CoLoG. All without the need forservice agreements and long development cycles,

and built by the end-users themselves.

Integrated Knowledge Tree, based on an individual’s own ontology.

Web Services configuredand tested within user-friendly

GUI interfaces.Support for multiple Social Worldsthat each individual engages in,and moves through, during their

daily life.

Speech-Acts in the formof an agent communication

language (ACL), enable networksof work-flow, between agents

and the human user/s,that share the Knowledge Tree,

as a semantic storagearea of data, information and

other resources.

Identify LogicFlow Lines

e.g. All entities touched by Flow Line #2

Then, in order to divide the ER diagram into appropriate sub-areas, which will in

turn be represented by sub-agents (often sourced via external web services), we

identify logic flow lines. These lines are chartered by taking the 4 high level tasks identified in the Task Analysis, and then

joining each of the entities in the ER diagram which are involved in each

task - without breaking the line.

We now set about designing 4 sub-agents which will

encompasses the 4 groupsof sub-tasks, hereafter called:

Movie AgentCinema AgentLocation Agent

Ticket Agent

Creating well-formed Logic Rules. We include one logic rule for each of the 4 sub-agents. For each rule, we include all the terms necessary to answer all of our sub-tasks as identified in the Task Analysis, within the 4 appropriate head-of-the-rule predicates:cinemaAssistant(CinemaName, Location, TheatreType, MovieTitle, GenreName, Rating, Date, StartTime)movieAssistant(MovieTitle, Rating, GenreName, RunningTime, DvdRelease, RoleName, FirstName, Surname)localityAssistant(Distance, Radius, Location, TownSuburb, Postcode, CinemaName)seatSelection(CinemaName, Location, TheatreID, ImageMap, MovieTitle, Date, StartTime, RowNumber, SeatNumber, CoordX,

CoordY, allocated)

A customer Query like, “What movies feature Sharon Stone?” can be answered via a logic language such as thosebuilt into several existing multi-agent system (MAS), using the movieAssistant predicate:movieAssistant(MovieTitle, Rating, GenreName, RunTime, DvdRelease, RoleName, FirstName, Surname)As follows:movieAssistant(_,_,_,_,_,star, sharon, stone);Beyond Patterns and ‘experienced logic programming’ there are surprisingly few methodologies fordevising good logic programs. That is why we devised TANDEM – to devised well-formed logic rules

The correct predicates in the rule body are:movie(MovieID, MovieTitle, Rating, RunningTime, Website, ReleaseDate, Trailer, DvdRelease, Synopsys) ANDmoviePeople(MoviePersonID, Firstname, Secondname, Surname, Nationality) ANDmovieGenre(MovieID, GenreName) ANDgenre(GenreName, GenreDescription) ANDcastCrew(MoviePersonID, MovieID, RoleName) ANDrole(RoleName, Description); which are taken from a reading of the entity/class diagramabove, choosing those touched directly by the logic flow line.

Implemented using the DigitalFriend MAS…

Top Related