Behavioural Rules In Multi Agent Systems Max
Download as PPT, PDF1 like522 views
The document describes a multi-agent system called "City-Live" that was developed to model and simulate urban mobility. It discusses: 1) The goals of developing the City-Live model to analyze the effects of transportation/urban planning decisions on traffic, public transport usage, and quality of life. 2) How the model represents the city as an environment for agents to interact in, and the agents which represent residents and commuters. 3) How rules governing agent behavior were extracted from a population survey using knowledge discovery techniques.
![The behavioural rules in Multi Agent Systems: a “not a toy” approach Alessandra LAPUCCI M assimiliano PETRI Diana POLETTI L.I.S.T.A. – Laboratorio di Ingegneria dei Sistemi Territoriali e Ambientali University of Pisa, Department of Civil Engineering [m.petri, alessandra.lapucci, diana.poletti]@ing.unipi.it](https://image.slidesharecdn.com/behavioural-rules-in-multi-agent-systemsmax-1215618462390243-9/85/Behavioural-Rules-In-Multi-Agent-Systems-Max-1-320.jpg)
![L.I.S.T.A. – Laboratorio di Ingegneria dei Sistemi Territoriali e Ambientali THE END Alessandra Lapucci [email_address] M assimiliano Petri [email_address] Diana Poletti [email_address] University of Pisa Department of Civil Engineering Thank you !!](https://image.slidesharecdn.com/behavioural-rules-in-multi-agent-systemsmax-1215618462390243-9/85/Behavioural-Rules-In-Multi-Agent-Systems-Max-23-320.jpg)
Behavioural Rules In Multi Agent Systems Max
- 1. The behavioural rules in Multi Agent Systems: a “not a toy” approach Alessandra LAPUCCI M assimiliano PETRI Diana POLETTI L.I.S.T.A. – Laboratorio di Ingegneria dei Sistemi Territoriali e Ambientali University of Pisa, Department of Civil Engineering [m.petri, alessandra.lapucci, diana.poletti]@ing.unipi.it
- 2. L.I.S.T.A. – Laboratorio di Ingegneria dei Sistemi Territoriali e Ambientali The starting point A first award
- 3. L.I.S.T.A. – Laboratorio di Ingegneria dei Sistemi Territoriali e Ambientali Topics Knowledge Need A particular MAS: Activity-Based Model “ Citylive” Structure and Case Study Application The Environment The Agents The Rules : knowledge extraction from data The model implementation
- 4. L.I.S.T.A. – Laboratorio di Ingegneria dei Sistemi Territoriali e Ambientali “ City-Live” model answers Knowledge Need Interventions on: Road Conditions Traffic Regulations Public Transport Road works Activities (re)localization Activities opening/closing times Limited Access Areas ………… . “ City Live” model Simulations Effects on: Traffic and Congestion Public Transport Demand Parkings Demand Travel Time to Work Travel Time to School Travel Time to Various Services …. “ City as Living Organism” Function Assessment = Life Quality
- 5. L.I.S.T.A. – Laboratorio di Ingegneria dei Sistemi Territoriali e Ambientali - A) an Environment - B) a set of Agents, active entities of the system - C) a set of “Rules” regulating agents’s activities The Case Study The Activity-Based Model Population Survey SCHEDULING - WHERE do city users go? ( in which services ) activities localization - HOW do they get there? (by which transport means) traffic and - WHERE do they park? public transport - WHICH family members are involved? family organization - IN WHICH hours do they move? space use and - HOW MUCH time do they spend? time consume - HOW LONG do they stay? …………
- 6. Why sequential Activity-Based model ? The Case Study The Activity-Based Model L.I.S.T.A. – Laboratorio di Ingegneria dei Sistemi Territoriali e Ambientali Morning act. diary pattern Afternooon act. diary pattern Evening act. diary pattern
- 7. L.I.S.T.A. – Laboratorio di Ingegneria dei Sistemi Territoriali e Ambientali Region : Tuscany (Italy) City : Pisa Residents : approximately 82.000 Surface : 7600 hectares A) The Environment in“City-Live” The Study Area
- 8. L.I.S.T.A. – Laboratorio di Ingegneria dei Sistemi Territoriali e Ambientali A Reactive Agent Temporal Geometric Network Geodatabase of activities located in the study area Population data related to the 918 census sections involved The Environment is structured as e real agent The environment is implemented on a G.I.S. platform It allows efficient and dynamic spatial queries A) The Environment in“City-Live” Behaviours /Attributes vary through time through space according to interactions with agents
- 9. City-Live Population Survey Two different City Users Commuters Universe: the commuters working in the activity with more than 20 employees (source: firm direct contact) Sample: based on a spatial accessibility and homogeneity criteria Residents Universe: the total residents in the Census Areas selected (source: Statistical National Agency) Sample: a two-steps sample method Residents Commuters Pisa city centre Pisa city center Arrival points Residents Activities Commuters Activities
- 10. L.I.S.T.A. – Laboratorio di Ingegneria dei Sistemi Territoriali e Ambientali City-Live Population Survey Accessibility index Road Network (with one-way) Census Area centroids
- 11. L.I.S.T.A. – Laboratorio di Ingegneria dei Sistemi Territoriali e Ambientali City-Live Population Survey Accessibility index Gravitational Potential PG a = k j Σ j M j / d aj α where PGa = Gravitational Potential fotr the Census Area a Kj = Census Area j weight Mj = Number of emplyees in the Census Area j daj = Distance between a and j calculated on the Network alfa = distance sensitiveness We use this index to create Census Area Clusters based on homogeneous accessibility criteria
- 12. L.I.S.T.A. – Laboratorio di Ingegneria dei Sistemi Territoriali e Ambientali City-Live Population Survey Accessibility index
- 13. L.I.S.T.A. – Laboratorio di Ingegneria dei Sistemi Territoriali e Ambientali Questionnaire Structure 2 – Daily Activities: Activity Type (14); Start/End Activity Period; Activity Localization; Activity Duration; Transportation Means; Reason for Choosing or not Public Transports ( specifically requested from Pisa Province ) Trip Time; Planning Moment; Accompainment Possibility (number of people). Questionnaire Structure 3 – Individual preferences : Preferred transport means Used transport means Judgements about urban services … Questionnaire Structure City-Live Population Survey Questionnaire Structure 1 – Personal Data: Class (commuter, domiciled or resident); Residence/Arrival area in Pisa; Sex; Age Band; Civil Status; Number of Transfers; Single Component Occupation; Individual Salary Range; Educational Qualification; Number of Family Components; Family Composition; Head of a family Age; Number of Children in the Family; Driving Licence Number in the Family; Car Numerousness in the Family. Individual Data Family Data
- 14. L.I.S.T.A. – Laboratorio di Ingegneria dei Sistemi Territoriali e Ambientali Questionnaire Structure City-Live Population Survey
- 15. L.I.S.T.A. – Laboratorio di Ingegneria dei Sistemi Territoriali e Ambientali City-Live Population Survey The web site
- 16. L.I.S.T.A. – Laboratorio di Ingegneria dei Sistemi Territoriali e Ambientali City-Live Population Survey Personal data survey The web site
- 17. L.I.S.T.A. – Laboratorio di Ingegneria dei Sistemi Territoriali e Ambientali City-Live Population Survey Activity diary data web-GIS For clients with editing not allowed (administrations, firms, etc..) The web site For clients with allowed editing (sample)
- 18. L.I.S.T.A. – Laboratorio di Ingegneria dei Sistemi Territoriali e Ambientali City-Live Population Survey Questionnaire: results Legend Residence Arrival com. Activity Travel by car Travel by bike Travel by bus Travel on foot Activity duration Time axis Ore 12.30-14.00
- 19. L.I.S.T.A. – Laboratorio di Ingegneria dei Sistemi Territoriali e Ambientali Survey Use - 1 Sample Survey : Questionnaires Agents: Residents inserted in their own Familiar Context Singles Commuters Iterative Proportional Fitting Whole Population Reconstruction City-Live B) The Agents
- 20. L.I.S.T.A. – Laboratorio di Ingegneria dei Sistemi Territoriali e Ambientali Examples: Which choice is made first in the model? Which transport means do an individual choose? At what time does the activity start? … . Knowledge Discovery in Databases Knowledge Extraction for Model Building City-Live C) The Rules Sample Survey : Questionnaires Survey Use - 2
- 21. L.I.S.T.A. – Laboratorio di Ingegneria dei Sistemi Territoriali e Ambientali City-Live : C) The Rules Example: Survey & KDD Decision Tree IF .. THEN .. Rules
- 22. L.I.S.T.A. – Laboratorio di Ingegneria dei Sistemi Territoriali e Ambientali Cube environment City-Live : Activity-Based Model - It incorporates most of the Activity-based demand techniques. It allows the input of GIS data and their editing in a ArcGIS-like environment It distributes model run processes across multiple computer processors, cutting model run times It contain a scripting language to insert the KDD rules in the choice processes modules It allows choice aggregation combining the effects of individual choice for such things as travel destination, time of day, cost and parking to provide aggregate representations
- 23. L.I.S.T.A. – Laboratorio di Ingegneria dei Sistemi Territoriali e Ambientali THE END Alessandra Lapucci [email_address] M assimiliano Petri [email_address] Diana Poletti [email_address] University of Pisa Department of Civil Engineering Thank you !!
Editor's Notes
- #2: The research i’m going to present is an attèmpt to build an operational Multi Agent System not a spatial simulation toy