Location Choice Modeling for Shopping and Leisure Activities with MATSim: Utility Function Extension and Validation Results A. Horni IVT ETH Zurich
MATSim: Overview Local search based on time geography Utility function extension and validation results Future research
dynamic, disaggregated Measures: Travel distance distribution Travel time distribution Link loads (?) Winner-loser statistics (WU) ? Catchment area ? Number of visitors of type xy ? … 3 MATSim: Model Purpose Possible level of disaggregation Clustering of population f(person attributes) ? MATSim model purpose: Transport planning simulation Model patterns of people’s activity scheduling and participation behavior at high level of detail. Planning goal: Average working day of Swiss resident population (> 7.5 M) in „reasonable“ time → end of 2009 (KTI project) Method: Coevolutionary, agent-based algorithm
4 Execution: Traffic simulation Replanning Scoring: Utility function f(t) t→ activity participation, travel Share x of agents (usually 10%): Time choice Route choice Location choice Physical layer Strategic layer Agent population Memory Day plans Initial demand: Fixed attributes e.g. (home location) from census data Plan Selection MATSim: Structure Exit conditon: „Relaxed state“, i.e. equilibrium
5 Location Choice in MATSim Relaxed state (i.e. scheduling equilibrium … (not network eqilibrium (Wardrop I/II), Nash? ) Huge search space prohibitively large to be searched exhaustively Dimensions (LC): # (Shopping, Leisure) alternatives (facilities) # Agents + Time dimension → agent interactions Local search + escape local optima Existence and uniqueness of equilibrium
6 Local Search in Our Coevolutionary System Tie together location choice and time choice ( t) p(accept bad solutions) > 0 Day plan Aktivity i - Work Location Start time, duration … Aktivity i+1 - Shopping Duration Aktivity i+2 - Home Location Start time, duration … Location Set: Locations consistent with time choice (t travel ≤ t budget ) Travel time budget Time Geography Hägerstrand Based on PPA- Algorithm Scott, 2006
7 STRC 2008 ZH Scenario: 60K agents
8 First Validation Steps & Utility Function Extensions Improve sim results Consider potential for application of estimated utility maximization models → hypothesis testing MATSim utility maximization framework Starting point for development and introduction of mental modules (such as e.g. location choice) Score → verification
9 Utility Function Extensions (Shopping Activities) Utility function SituationAlternativePerson Strictly time-based → extension (parameters, structure) Store size Stores density in given neighborhood
10 On the Way to Results – ring-shaped PPA Leisure travel <= models of social interaction and sophisticated utility function Not yet productive MATSim longterm goal Activity-based models (chains) → Reasonable shopping location choice model requires sound leisure location choice modeling trip generation/distribution → activity-based multi-agent framework Trip distance distribution MC → act chains (ring-shaped potential path area) Agent population Assignment of travel distances crucial and non-trivial for multi-agent models! Leisure Predictability of leisure travel based on f(agent attributes)? Leisure trip distance ↔ -desired leisure activity duration -working activity activity chains ← f(agent attributes)
11 Results – Avg. Trip Distances Config 0: base case Config 1: leisure PPA Config 2: + shopping activity differentiation (grocery – non-grocery; random assignment) Config 3.1: config 2 + store size Config 3.2: config 2 + stores density Shopping trips (car) Leisure trips (car)
12 Results – Avg. Trip Durations Strong underestimation in general! -Missing intersection dynamics -Access to (coarse) network (parking lots etc) -Freight traffic essentially missing Shopping trips (car)
13 Microcensus bin size ratio (bin 0 / bin 1 ) = 4.22 Config 0 bin size ratio (bin 0 / bin 1 ) = Config 1 bin size ratio (bin 0 / bin 1 ) = 7.08 Config 2 bin size ratio (bin 0 / bin 1 ) = 7.00 Config 3.1 bin size ratio (bin 0 / bin 1 ) = 6.41 Config 3.2 bin size ratio (bin 0 / bin 1 ) = 6.44 Results – Shopping Trip Distance Distributions (Car)
Results – Count Data 18:00 -19:00 Config 0 Config 1 Config 3.1
15 Results – Count Data – 24 h (i, j) (i,j) [%] dist (i,j) [%] 0, … 1, 20.06… 2, , Car shopping trips Config 0 daily : -60.3% Config 1 daily : -36.4% Retest: -... more disaggregated data! -... more stations (now 300 stations for CH) General underestimation of traffic volume dist = upper bound for reduction of error due to increased traffic volume (increased avg. distances) Utf. extensions productive → spatial distribution of trips Weighting by shopping traffic work (#trips * trip length) ≈ 7 % (excl. back to home trips) Reject hypothesis No improvement w/ respect to spatial distribution of trips
16 Conclusions Open research questions... Starting point for validation Strictly time-based utility function → strong underestimation of traffic volume (as expected) Extension of utility function shows expected effects but … Effects very small & difficult to evaluate H 0 : blue eyes → disaggregated evaluation level Reject hypothesis? MATSim hypothesis testing tool?
17 Shopping utility function estimation Future work Choice set generation (boundaries)→ survey: homo oeconomicus vs. real person Further validation steps Disaggregation level of agent-based models Evaluation and modeling level = f (data base) Existence and uniqueness of scheduling equilibrium Inductively: different initial statesPredictability of leisure travel Reducing leisure travel to a cross- sectional sample (e.g. 1 MATSim day)