Post on 10-Jan-2017
transcript
Spatial simulation of complex adaptive systems
why “agents” only cannot do the job
Arnaud Banos (International mobility support 2016 from InSHS-CNRS)
Globally addicted tocomplex systems
A large number of localized interacting entities,
operating within an environment
These entities, being human or not, act and
are influenced by the local environment and
interaction network they are situated in
While the behaviour of these entities may be
inspired, guided or limited by upper structures,
they are not directly controlled but operate (at
least partly) on their own, having some “self-
control” over their actions and internal states Non coordinated but interdependant local actions
Emergence of global structures
Non coordinated but interdependant local actions
Emergence of global structures
Eric Daudé
Complex Adaptive Spatial Systems
Party game
AA BB
Strategy 1
YOUYOUAfter Eric Bonabeau https://hbr.org/2002/03/predicting-the-unpredictable
Party game
AA BB
Strategy 2
YOUYOUAfter Eric Bonabeau https://hbr.org/2002/03/predicting-the-unpredictable
Micro vs Macro
Difficult to guess macro from micro in presence of interactions SIMULATION
Difficult to guess micro from macro reconstruction (SIMULATION)
http://www.cartoonstock.com/directory/f/flight_simulator.asp
?
?
ABM can help a lot!
Fiegel, Banos, Bertelle, 2009
Canned meat
Urban ant-hill
Fosset, Banos, et al. 2016
LEZ ⇒ + 8,3 % exposure to PM10 (average emission rate weighted by living population)
Possible impact of LEZ in Grenoble
Fosset, Banos, et al. 2016
http://www.deviantart.com/
« There is no one best way! »
Herbert SimonNobel Prize Economy 1978
Turing Prize 1975
Spatial simulation of complex adaptive systems
why “agents” only cannot do the job
http://quotesgram.com/agent-smith-quotes-about-purpose/
1- a FEW agents may help revealing:
● Intentionality● Preferences● Constraints and adaptation to constraints● Cooperation● Strategies● And so much more !
SMArtAccess
A = Work
B = Universal Service
C = Commercial Service
D = Home
D choosen randomly
A choosen randomly with proba p
Fixed trip chain: D ⇒ A ⇒ B ⇒ C ⇒ D
Chain =
Traffic = Deterministic Single-Regime speed-Density (Underwood):
min(T = TD,A +TA ,B +TB ,C +TC ,Då ), cc V f
Rules
Vi =V f i e-a
n iC i
æ
è ç
ö
ø ÷
Creating cities from scratch
Air pollution
Multi-Agents & Multi-Actors (M2A2S)
M2A2S
Cooperative game: no competition but individual and collective objectives
« Economic » work places, universal
and commercial services
« Citizen » home places
« Public » road network, public transport, air pollution
restricted areas
M2A2S
Objective functions
« Economic »Max #consumers
Max population coverage for US
« Citizen » Min unhappiness
(accessibility, traffic, pollution)
« Public » Min congestion
Max public transportMin air pollution
(concentration and exposure)
« Global »Sustainable city!
Collaborative Game PAMs
Cooperation
2- agents are not always relevant
● Efficiency (computation time)● Scale of the processes
➔Model Coupling
2- agents are not always relevant
● Efficiency (computation time)● Scale of the processes
➔Model Coupling
Road traffic modeling
VANWAGENINGEN-KESSELSF.,VANLINTH.,VUIKK.,SERGEH.,«Genealogy of traffic flow models », EURO Journal on Transportation and Logistics, vol. 4, no° 4, p. 445–473, Springer, 2015
macro
meso
micro
Theoretical developments
Genealogy of trafic models based on the fundamental diagram
Fundamental Diagrams for Uninterrupted Traffic Flow
(Source: Austroads Guide to Traffic Management Part 2: Traffic Theory)
Road traffic modeling
Banos et al., under press
« Ring City »
Micro: NaSch (acceleration/deceleration based on front vehicle)
Meso: Underwood
Macro: LWR (Lighthill, Whitham, Richards)
Flow
Traffic conservation
Vi =V f i e-a
n iC i
æ
è ç
ö
ø ÷
https://www.researchgate.net/publication/258397885_Formation_and_Propagation_of_Local_Traffic_Jam
Road traffic modeling
« Ring City »
Banos et al., under press
Hybrid Micro/Macro Model
Road traffic modeling
Taillandier, Banos, Corson., Coupling macro and micro models to simulate traffic, in progress
2- agents are not always relevant
● Efficiency (computation time)● Scale of the processes
➔Model Coupling
Epidemic spread
http://www.humanosphere.org/global-health/2013/09/unleashing-big-data-against-disease/
Macro VS micro
Banos et al., 2016
Macroscopic approach
SIR macro model
→ Metapopulation model
Node = CityEdge = Flight connection
==> Mobility rate « g »==> Probability of Flowij « mij »
Banos et al., 2016
Model CouplingNode = city ==> SIREdge = flights and passengers (agents)
Banos et al., 2016
Mean Field Approach
If we assume:- Instantaneous trips - Complete network- Constant “g” and “mij” THEN we can calculate :- MaxI- TimeOfMaxI- Duration BOTH MODELS ABLE TO REPRODUCE THESE VALUES
Banos et al., 2016
Containment strategy: quarantine
. Metapopulation model
. MicMac model (100 replications)Banos et al., 2016
Risk culture (not travelling if infected)
. Metapopulation model
. MicMac model (100 replications)Banos et al., 2016
Main advantages
Simple but not too simple models “Einstein's razor” > “Occam's razor”
Deepening understanding → coupling agents and actors (serious games)
Coupling processes in space and time and across scales and levels → coupling models, formalisms and theories
Collaboration with Mr Robert
Data-scarce context → be SMART!
Data-scarce context → be SMART!
Topography + Gravity + Active Particles
Data-scarce context → be SMART!
Topography + Gravity + Active Particles+ Dynamic Lanscape
Data-scarce context → be SMART!
Topography + Gravity + Active Particles+ Dynamic Lanscape + Dynamic Rivers
Data-scarce context → be SMART!
Topography + Gravity + Active Particles+ Dynamic Lanscape + Dynamic Rivers + Dynamic Particles/Rivers interactions
Data-scarce context → be SMART!Flooding scenario in Jakarta (in red, penalized river sections in term of capacity)
Next:● Calibration (sensors, models) ● validation (Remote sensing, sensors)● Data assimilation?● Serious game?● Complexification ?
Incremental complexification of models
● Fiegel J., BANOS A., Bertelle C., 2009, Modeling and simulation of pedestrian behaviors in transport areas: the specific case of platform/train exchanges, ICCSA 2009, 29 june-2 july, Le Havre
● Fosset P., BANOS A., Beck E., Chardonnel S., Lang C., Marilleau M., Piombini A., Leysens T., Conesa A., André-Poyaud I., Thèvenin T., 2016, Exploring intra-urban accessibility and impacts of pollution policies with an agent-based simulation platform: GaMiroD, Systems, 4(1), 5.
● BANOS A., 2015, The city, a complex system? The new challenges of urban modelling, in Lagrée S, Diaz V., A glance at sustainable urban development: methodological crosscutting and operational approaches, pp. 110-119, AFD, Paris
● BANOS A., Corson N., Marilleau N., Taillandier P., Multi-scale Traffic Modelling in NetLogo, in BANOS A., Lang C., Marilleau N. (Dir), Agent-Based Spatial Simulation with NetLogo: Advanced Users, Wiley, London, To be published
● Taillandier P., BANOS A., Corson N., Coupling macro and micro models to simulate traffic, in progress
● BANOS A. Corson N., Gaudou B., Laperriere V., Rey S., 2015, The importance of being hybrid for spatial epidemic models: a multi-scale approach, Systems,3(4), 309-329
References