Transcript A Disaggregate Quasi-Dynamic Park-and

APPLICATION OF A DISAGGREGATE
QUASI-DYNAMIC MODEL OF PARKAND-RIDE LOT CHOICE
John Gibb
DKS Associates
Transportation Solutions
The Park-and-Ride Problem for
Transit Auto Access:
 Which park-and-ride transit stop for a trip
 Getting level of service “skim” values for auto
and transit legs
 Assigning auto and transit legs
Customary Solutions
(Trip-Based)
 Zone-Station links by auto access “shed”
 Capacity restraint by art, trial and error
 Drive legs not assigned
 Intermediate zone
 EMME triple-index (convolution)
 Multinomial logit
 Capacity restraint by shadow-price
Individual trip modeling
- as in activity-based model
 Heterogeneous choice sets & behavior
 Time-specific
 Sub-mode choice
 Single outcome per choice
 Determines auto & transit trips in both directions
“Real world”: Parking
available to all until full
 Time-dependent choice set
 Arrival time determines individual’s priority
(not drive distance or analyst’s judgment)
 Commuter behavior:
 Know when lots fill
 No frustrated arrivals to full lots
Original Sacramento
Application: Chronological
Order
 One-pass algorithm:
 Sort trips by presumed departure time
 Choose best-utility among available lots
 Accumulate parking loads; make unavailable
when full
Limitations of the one-pass
method
 Loss of choices
 Departure & parking-arrival time varies
among alternatives
 One can leave earlier to beat a lot’s fill-time
Improved method for Sacramento
update and new Seattle ABM in
progress…
Crawford-Knoer matching
algorithm (1981)
 Generalizes Gale-Shapely (1962)
 Hospital-residents, college admissions, stable
marriage problems
 Iterative rounds of “proposals” until
constraints satisfied.
 In C-K, rejected proposals are
adjusted & resubmitted
C-K algorithm for parking,
briefly
 Iterative rounds
 Parking choice
 Latecomer rejection
 Rejectees adjust departure time to that lot a unit-step
earlier
 Departure-time adjustment counts against utility
 Choice may repeat
 Trip “accepted” may be “bumped” in a later round
 Stop when no parking oversubscribed
Crawford-Knoer properties
 User-optimal equilibrium
 Escalation of early arrival times
 Last-minute arrival rush
 No denial of choice
 Gradual adjustment avoids problems, can use
efficient methods
 Needs an early-departure utility parameter
System equilibration flow
Skim Matrices
Network
Assignment
P&R lot
placement
Lot-Full
Times
Activity-Based
Demand Model
Trips
Thanks!
Questions, requests for reports welcomed at
[email protected]
DKS Associates
TRANSPORTATION SOLUTIONS
www.dksassociates.com