Agent-Oriented Modelling Prof Kuldar Taveter, Tallinn University of Technology, Estonia

Transcription

1 Agent-Oriented Modelling Prof Kuldar Taveter, Tallinn University of Technology, Estonia

2 Overview Introduction Motivating example Conceptual space Analysis level models Design level models Viewpoint Modelling Framework Summary

3 The book

4 Problems in airports Capacities are not used in full Connectivity is jeopardized Air and ground operations cannot be handled economically People are exposed to high workload

5 En-route delays vs. airport delays

6 Major stakeholders Air traffic control (ATC) Airports Airlines

7 Islands of automation Departure and Arrival Traffic Managers (ATC) Resource Allocation Manager (airport) Environment Manager (airport) Information Manager (ATC,( airport, airlines) Internal IS (airlines)

8 Global optimization Super system over existing systems? (client-server solution) To have the right knowledge by the right agent at the right time + negotiations (P2P solution) To act based on the knowledge

9 Advanced Collaborative Decision Making (A-CDM) IMAN control flow data flow messages Global data ATC Airport Airline / Ground Handler ATC Agent Airport Agent Airline/GH Agent User AMAN DMAN User User EMAN RMAN User Airline/GH IS User Flight data Env. and operational data Resource data Internal data

10 Simulation environment ATC simulation GUI control flow data flow messages ATC Agent Environment simulator Airline / GH Agent Airline / GH simulation GUI Historical event data Airport Agent Airport simulation GUI

11 The conceptual space Motivation layer System design layer IMPLEMENTATION DESIGN Deployment layer

12 Motivation layer

13 System design layer

14 Deployment layer

15 Relationships between the concepts

16 Types of models at different layers Motivation layer: goal models, role and organisation models,, domain models System design layer: agent and acquaintance models, interaction models, knowledge models, scenarios, agent behaviour models,, service models Deployment layer: agent a interface and interaction specifications, data models, agent behaviour specifications, service specifications

17 The overall Goal Model Maximal safety Maximal airport efficiency Passenger Allocate resources Minimal environmental nuisance Allocate ATC resources Allocate airport resources Allocate airline / GH resources Local ATC Airport Airline / GH

18 Notation for business process modelling External Object Type Agent Type Internal Object Type sends Message Type receives Agent Type triggering event Rule mental state condition action Activity Type start event perceives Non-Communicative Action Event Type does action Rule receives Message Type sends outgoing message mental effect Internal Object Type perceives Non-Action Event Type Predicate

19 Organization and knowledge model Airline / Ground Handler FlightPlan 1..* Outbound 1 CallSign ADES TakeOffRunway SID SOBT EOBT /OffBlockTime = inbound.inblocktime + + aircraft.estimatedturnroundtime /TakeOffTime = inbound.landingtime + + aircraft.minimumturnroundtime + + TaxiOutTime TaxiOutTime ASRT TSAT <<represents> 1 Airport 1 Stand number isavailable (FlightPlan) * Pier number isavailable (FlightPlan) FlightPlan isscheduled isinitial isairborne isfir isfinal islanded Gate 1 1..* number isavailable (FlightPlan) * isinblock isboarding isready isoffblock isdeparted 0..* Runway Inbound Aircraft AircrafRegistration AircraftIdentification AircraftType HandlingAgent BoardingGate ParkingStand MinimumTurnRoundTime EstimatedTurnRoundTime = = MinimumTurnRoundTime Local ATC CallSign ADEP LandingRunway TimeSlot EET starttime OffBlockTime endtime TakeOffTime sector LandingTime runway SIBT /InBlockTime = LandingTime + isoptimal + TaxiInTime (FlightPlan) TaxiInTime AGHT = InBlockTime 2..* <<represents> * * PlaneType AircraftType Arrival Slot FlightPlan Departure Slot 1

20 Business process related to an inbound flight

21 Detailed business process model

22 Simulation

23 Viewpoint Framework Viewpoint models Viewpoint aspect Abstraction layer Interaction Information Behavior Conceptual domain modeling Platform-independent computational design Platform-specific design and implementation Role models and organisation models Agent models and acquaintance models, interaction models Agent interface and interaction specifications Domain models Goal models and motivational scenarios Knowledge models Scenarios and behavior models Data models and service models Agent behavior specifications

24 AOSE methodologies

25 Summary Agent-oriented modelling very useful for understanding and designing socio- technical systems AOM provides piecess for agreements modelling AOSE essential for multi-agent systems to be widely deployed in industry No one right approach - several of the methodologies are complementary