Joint Modelling of Schedule- and Frequency-based Services in
Public Transport Assignment Models
Trafikdage
August 29th 2017 Morten Eltved
Transport Modelling
DTU Management Engineering, Technical University of Denmark
Agenda
• Schedule- and frequency-based assignment
• Relevance of a joint model
• Developed framework
• Results
• Conclusion & outlook
2 29 August
2017
Schedule- vs Frequency-based Models
Schedule
-based Frequency -based
Coding Demanding Simpler
Computation time Long Short
Results Detailed Overall
Forecasting Demanding Simpler
DTU Management Engineering, Technical University of Denmark
Why do we need a joint model?
• Flexible modelling
• Reliable forecasts
• Modelling of future scenarios
• Todays network
4 29 August
2017
Timetables Today
DTU Management Engineering, Technical University of Denmark
Schedule-based -
Schedule-based
Frequency-based -
Frequency-based Schedule-based
-
Frequency-based
Frequency-based -
Schedule-based
Transfers in a Combined Network
6 29 August
2017
Overall Model Framework
Choice set generation
Discrete choice
model
Flow distributio
n
Event dominance with threshold
Multinomial
logit
DTU Management Engineering, Technical University of Denmark
Real Life Test Examples
8 29 August
2017
DTU – CPH Airport DTU – Brønshøj
DTU – CPH Airport Overview
DTU Management Engineering, Technical University of Denmark
Base Results - DTU to CPH Airport
10 29 August
2017
Change in Departure Time
Departure 7.30am Departure 7.35am
DTU Management Engineering, Technical University of Denmark
Change in Preferences
12 29 August
2017
Departure 7.30am Low Train Parameter
DTU – Brønshøj Overview
• 20 lines
– 2 frequency-based lines – 18 schedule-based lines
• Headway 4-20 minutes
DTU Management Engineering, Technical University of Denmark
Reasonable Paths
14 29 August
2017
Base Results DTU - Brønshøj
Path Cost Arr. time
300S 5C 10.50 8.19am
DTU Management Engineering, Technical University of Denmark
Base Results DTU - Brønshøj
16 29 August
2017
Path Cost Arr. time
300S 5C 10.50 8.19am
150S 21 5C 10.65 8.15am
Base Results DTU - Brønshøj
Path Cost Arr. time
300S 5C 10.50 8.19am
150S 21 5C 10.65 8.15am
150S F 350S 10.71 8.15am
DTU Management Engineering, Technical University of Denmark
Base Results DTU - Brønshøj
18 29 August
2017
Path Cost Arr.
time Prob
300S 5C 10.50 8.19am 23%
150S 21 5C 10.65 8.15am 20%
150S F 350S 10.71 8.15am 19%
30E 250S 5C 11.15 8.19am 12%
180 B F 350S 11.44 8.15am 9%
150S 8A 5C 11.85 8.21am 6%
30E 200S 350S 11.96 8.23am 5%
15E 350S 12.17 8.30am 4%
Weighted average 10.97 8.18am 100%
Change in Type of Service (SB/FB)
• Changing high frequent lines:
– 5C (4min)
– 150S (5min)
– 350S (6min)
DTU Management Engineering, Technical University of Denmark
Change in Type of Service (SB/FB)
• Changing high frequent lines:
– 5C (4min) – 150S (5min) – 350S (6min)
20 29 August
2017
Results for Change to FB Lines
DTU Management Engineering, Technical University of Denmark
Frequency Aggregation of Network
22 29 August
2017
Line 1 Line 2
Agg.
Frequency Aggregation of the Network
• Aggregated lines:
– 5C and 350S
– 150S and 15E
DTU Management Engineering, Technical University of Denmark
Results Using Frequency Aggregation
24 29 August
2017
Comparison of Scenarios
Scenario Avg.
cost
Avg. arr.
time
Alter- natives
Base 10.97 8.18am 8
FB lines 11.25 8.20am 8
Aggregated 11.09 8.18am 7
DTU Management Engineering, Technical University of Denmark
Lessons Learned
• No existing models
• A framework has been developed
• Hard to generate proper choice set
• Threshold is hard to define
• No knowledge on passengers perception of timetables
26 29 August
2017
Future Work
• Understanding of passengers route choice – Perception of timetables
– Correlation of overlapping routes
• Estimation of model
– Discrete choice model
– Recursive assignment model?
• Inclusion of:
– Delays
– In-vehicle congestion
DTU Management Engineering, Technical University of Denmark
Thank you for listening
28 29 August
2017