Download presentation
Presentation is loading. Please wait.
Published byWilfred Lewis Modified over 9 years ago
1
Mo So A. Horni IVT ETH Zürich Juli 2012 Simulation einer Woche mit MATSim http://synonyme.woxikon.de/
2
Gemeinschaftsprojekt KTH, ETH, EPFL, DTU Eingebettet: (Entfernte) Verwandtschaft mit Diss-Thema Zielwahl für (etwas) grösseren Zeithorizont zeitliche Variabilität → in MATSim UTF Ordóñez et al. 2012 Szenario aufgesetzt, grobe Idee für Experimente, noch keine Resultate 2 Project Surprice (sic) - Kontext MATSim week
3
temporal variation Project Surprice (sic) - Problem VOT income individual preferences trip context trip context CC equity MATSim week
4
4 output execution replanning scoring controler analyses input config input: plans (demand) config (parameters) network (supply) plan agent population with day plans Implementation in MATSim: MATSim Principle
5
5 output execution replanning scoring controler analyses input config plan execute plans mobility simulation: event-based queue model modes: motorized individual traffic public transport bike (teleported) walk (teleported) ride (experimental) Implementation in MATSim: MATSim Principle
6
6 output execution replanning scoring controler analyses input config utility function generalized costs: plan evaluate plans Implementation in MATSim: MATSim Principle
7
7 output execution replanning scoring controler analyses input config share (usually 10%) decision dimensions: time choice (local random mutation) route choice (best response) mode choice (random mutation) destination choice (experimental) plan change plans Implementation in MATSim: MATSim Principle
8
8 output execution replanning scoring controler analyses input config statistics counts plans events → post-processing e.g., in visualizer Implementation in MATSim: MATSim Principle exit conditon: „relaxed state“, i.e. equilibrium exit conditon: „relaxed state“, i.e. equilibrium
9
9 output execution replanning scoring controler input Evolutionary algorithm Implementation in MATSim: MATSim Principle
10
More Precisely: A Co-Evolutionary Algorithm Interpretation 10 Agent i Plan 0 Plan 1 Plan 2 Plan 3 Agent j Plan 0 Plan 1 Plan 2 Plan 3 species i Individ. i,0 Plan 0 Individ. i,1 Plan 1 Individ. i,2 Plan 2 Individ. i,3 Plan 3 species j Individ. j,0 Plan 0 Individ. … Plan.. Competition on the infrastructure → Score (fitness) The weakest die → Generations person.removeWorstPlan() Agent … Plan.. Coevolution: „The evolution of two or more interdependent species, each adapting to changes in the other. […]“ The American Heritage Dictionary of the English Language Evolution: "Change in the gene pool of a population from generation to generation by such processes as mutation, natural selection, and genetic drift. […]“ „www.dictionary.com“
11
instantiation microsimulation (model) output input U max (day chains) population situation (e.g. season, weather) situation (e.g. season, weather) choice model generalized costs census travel surveys infrastructure data estimatione.g., network constraints, opening hours e.g., socio- demographcis network load simulation constraints («demand/supply equilibration») fixed point problem solved with co-evolutionary algorithm Implementation in MATSim: MATSim Principle feedback
12
Implementation in MATSim: Approach Q & D MonSunSat shopleisure MATSim integration T. Dubernet Q&D: no opening of MATSim 24h-cycle execution replanning scoring controler execution replanning scoring controler execution replanning scoring controler endogenous: time, route and mode choice exogenous: chain and destination choice lagged variables
13
base: ZH scenario (state WU) population: census 2000; demand: MZ 2000/2005; infrastructure: IVTCH, BZ 2001 chains (Thurgau, 231 respondents, 6 weeks) h-*-h-chains MATSim activity types locations (h, w from census; s, l, e with neighborhood search Balmer) 13 Implementation in MATSim: Scenario
14
14 Implementation in MATSim: Scenario cont’d population: census 2000 demand: MZ 2000/2005 s/l destinations: nb search Balmer net execution replanning scoring controler
15
U t = U act,t + U trav,t person characteristics/preferences U t-1 = U act,t-1 + U trav,t-1 lagged vars mode main,t-1 f income i pref Implementation in MATSim: UTF in More Detail agent memory Estimated UTF: 148 parameters! Reduce for MATSim application:
16
Implementation in MATSim: UTF in More Detail Estimation of MATSim UTF for iterative context? Constrained preferences! De Palma et al. (2006), Discrete choice models with capacity constraints: an empirical analysis of the housing market of the greater Paris region “ex ante and ex post demand”
17
Dependency on spatial configuration and CC scheme (in MATSim: distance, cordon, area) toll road free road net losers with CC higher tt with CC net winners with CC lower tt with CC Road Pricing: Equity Effects …where and how? C = tt + m ~ income; ~ 1/income; m: e.g., toll rich: large, small → C potentially larger with CC if ~ income less strong (trip context, variable prefs) → C more dispers == hypotheses equity in terms of C not ( C), tt? trip context averages out?
18
18 Temporal Variability, Counts 17-18 Uhr
19
aggregate results variability (e.g., link volumes) Var(X 0 +X 1 ) = Var(X 0 ) + Var(X 1 ) + 2 Cov(X 0, X 1 ) Input x + input sets Output output sets chains, destinationtime, route, mode Model x + Model x + input variability (exogenous) model variability (endogenous) total variability o measured variability (spatio-temporal) Temporal Variability and Correlations week chains lagged variables temporal variability equidirectional rhythm of life
20
Modeling of observed (“real”) variability or uncertainty? meaning of ? meaning of measures of dispersion? meaningless → Sampling method with sampling error confidence intervals 20 Variabilität – Interpretation of Results
21
search space extent → curse of dimensionality, combinatorial explosion Discussion: Week vs. Day Optimization Mon Sun planning horizon of decision makers? dependent on choice dimension (e.g., chain vs. time)
22
24h: Zurich scenario WU: 10%, 1 day KTI: 25%, 4 days (more choice dimensions and modes) Herbie: 10%, 5h (more threats) Zurich scenario with destination choice: 1 day 22 Discussion: Week vs. Day Optimization – Computation Times
23
23 Questions
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.