Delft University of Technology
Toward operationally feasible railway timetables (PPT)
Bešinović, Nikola; Goverde, Rob
Publication date 2016
Document Version Final published version
Citation (APA)
Bešinović, N., & Goverde, R. (2016). Toward operationally feasible railway timetables (PPT). 2016 INFORMS Annual Meeting, Nashville, United States.
Important note
To cite this publication, please use the final published version (if applicable). Please check the document version above.
Copyright
Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons. Takedown policy
Please contact us and provide details if you believe this document breaches copyrights. We will remove access to the work immediately and investigate your claim.
Introduction Problem description Methodology Experimental results Conclusions
Towards operationally feasible railway timetables
Nikola Beˇsinovi´c, Rob M.P. Goverde Delft University of Technology, The Netherlands INFORMS 2016, Nashville, Tennessee
Introduction Problem description Methodology Experimental results Conclusions
Outline
1 Introduction 2 Problem description 3 Methodology 4 Experimental results 5 ConclusionsIntroduction Problem description Methodology Experimental results Conclusions
Current state in railway traffic
Constant growth of demand for passenger and freight railway transport
Heavily congested networks
Reaching maximum available infrastructure capacity
Experiencing delays
Existing need for better planning to satisfy a high level of service (ERA, UIC, IMs, RUs...)
Introduction Problem description Methodology Experimental results Conclusions
Current state in railway traffic
Constant growth of demand for passenger and freight railway transport
Heavily congested networks
Reaching maximum available infrastructure capacity
Experiencing delays
Existing need for better planning to satisfy a high level of service (ERA, UIC, IMs, RUs...)
Introduction Problem description Methodology Experimental results Conclusions
Timetable planning
Infrastructure Line plan Timetable Rollingstock Crew
INPUT:
Train line requests (OD, stops, frequencies, rolling stock)
Track topology
Rolling stock with dynamic characteristics
Passenger connections and rolling stock turn-arounds OUTPUT:
Introduction Problem description Methodology Experimental results Conclusions
Timetable planning
Goals:
Efficiency - short travel times and seamless connections
Realizability - scheduled RT > minimum RT
(Operational) Feasibility - no conflicts
Stability - acceptable capacity occupation in corridors and stations
Robustness - cope with system stochasticity
Operationally feasible timetable
An operationally feasible timetable has no conflicts on the microscopic level (block and track detection sections) between train’s blocking times.
Introduction Problem description Methodology Experimental results Conclusions
Timetable planning
Goals:
Efficiency - short travel times and seamless connections
Realizability - scheduled RT > minimum RT
(Operational) Feasibility - no conflicts
Stability - acceptable capacity occupation in corridors and stations
Robustness - cope with system stochasticity Operationally feasible timetable
An operationally feasible timetable has no conflicts on the microscopic level (block and track detection sections) between train’s blocking times.
Introduction Problem description Methodology Experimental results Conclusions
Time-distance diagram
Ut Utl Htn Cl Gdm Zbm Ht Vg Btl Bet Ehv
0 10 20 30 40 50 60
Introduction Problem description Methodology Experimental results Conclusions
Blocking time diagram
Question:
Introduction Problem description Methodology Experimental results Conclusions
Introduction Problem description Methodology Experimental results Conclusions
Minimum headway time
Minimum headway time (Hansen and Pachl, 2014)
A minimum headway time is the time separation between two trains at certain positions that enable conflict-free operation of trains.
Minimum headway time Lij depends on:
infrastructure characteristics: block lengths
signalling system
train engine characteristics
(scheduled) train running times
Introduction Problem description Methodology Experimental results Conclusions
Minimum headway time
Minimum headway time (Hansen and Pachl, 2014)
A minimum headway time is the time separation between two trains at certain positions that enable conflict-free operation of trains.
Minimum headway time Lij depends on:
infrastructure characteristics: block lengths
signalling system
train engine characteristics
(scheduled) train running times
Introduction Problem description Methodology Experimental results Conclusions
Minimum headway time
Minimum headway time (Hansen and Pachl, 2014)
A minimum headway time is the time separation between two trains at certain positions that enable conflict-free operation of trains.
Minimum headway time Lij depends on:
infrastructure characteristics: block lengths
signalling system
train engine characteristics
(scheduled) train running times
Introduction Problem description Methodology Experimental results Conclusions
State-of-the-art
So far: Efficiency Realizability (Operational) Feasibility Stability - Robustness-Introduction Problem description Methodology Experimental results Conclusions
Periodic event scheduling problem (PESP)
Serafini & Ukovich (1989)
Periodic timetable with cycle time T
Periodic events: arrival & departure times πi ∈ [0, T )
Constraints:
lowerBoundij ≤ πj − πi+ zijT ≤ upperBoundij Period shift: zij - define the order of trains
a1 d1 d2 a2 [13,16]T [11,14]T [1,3]T [1,3]T [3,57]T [3,8]T [22,26]T [19,22]T
Introduction Problem description Methodology Experimental results Conclusions
Periodic event scheduling problem (PESP)
Serafini & Ukovich (1989)
Periodic timetable with cycle time T
Periodic events: arrival & departure times πi ∈ [0, T ) Constraints:
lowerBoundij ≤ πj − πi+ zijT ≤ upperBoundij
Period shift: zij - define the order of trains
a1 d1 d2 a2 [13,16]T [11,14]T [1,3]T [1,3]T [3,57]T [3,8]T [22,26]T [19,22]T
Introduction Problem description Methodology Experimental results Conclusions
Periodic event scheduling problem (PESP)
Serafini & Ukovich (1989)
Periodic timetable with cycle time T
Periodic events: arrival & departure times πi ∈ [0, T ) Constraints:
lowerBoundij ≤ πj − πi+ zijT ≤ upperBoundij Period shift: zij - define the order of trains
a1 d1 d2 a2 [13,16]T [11,14]T [1,3]T [1,3]T [3,57]T [3,8]T [22,26]T [19,22]T
Introduction Problem description Methodology Experimental results Conclusions
Periodic event scheduling problem (PESP)
Serafini & Ukovich (1989)
Periodic timetable with cycle time T
Periodic events: arrival & departure times πi ∈ [0, T ) Constraints:
lowerBoundij ≤ πj − πi+ zijT ≤ upperBoundij Period shift: zij - define the order of trains
a1 d1
[13,16]T [1,3]T
[3,57]T [3,8]T
Introduction Problem description Methodology Experimental results Conclusions
Solving PESP
(PESP − N) Min f (π, z) such that lij ≤ πj − πi + zijT ≤ uij ∀(i , j) ∈ A 0 ≤ πi < T , ∀i zij binaryIntroduction Problem description Methodology Experimental results Conclusions
Computing operationally feasible timetables
Solving PESP-N:
Fixed minimum headways
Can be violated when scheduled running time increases
How to include microscopic details in timetable planning models?
Iterative approach
Introduction Problem description Methodology Experimental results Conclusions
Computing operationally feasible timetables
Solving PESP-N:
Fixed minimum headways
Can be violated when scheduled running time increases How to include microscopic details in timetable planning models?
Iterative approach
Introduction Problem description Methodology Experimental results Conclusions
Iterative micro-macro framework (Transp. Res. B, 2016)
Macro Micro
Micro model (Comp-aided Civil and Inf. Eng., 2016):
Compute operational train speed profiles
Conflict detection
Introduction Problem description Methodology Experimental results Conclusions
Integrated approach
Can we add microscopic details directly to the macroscopic level?
Yes. Introduceflexible minimum headways in PESP
Introduction Problem description Methodology Experimental results Conclusions
Integrated approach
Can we add microscopic details directly to the macroscopic level? Yes.
Introduction Problem description Methodology Experimental results Conclusions
Integrated approach
Can we add microscopic details directly to the macroscopic level? Yes. Introduceflexible minimum headways in PESP
Introduction Problem description Methodology Experimental results Conclusions
Integrated approach
(PESP − N) Min f (π, z) such that lij ≤ πj − πi + zij · T ≤ uij ∀(i , j) ∈ A 0 ≤ πi < T , ∀i zij binaryIntroduction Problem description Methodology Experimental results Conclusions
Integrated approach
(PESP −FlexHeadways) Min f (π, z) such that
lij ≤ πj − πi + zij· T ≤ uij ∀(i , j) ∈ Arun∪ Adwell Lij ≤ πj − πi + zij · T ≤Uij ∀(i , j) ∈ Aheadway 0 ≤ πi < T , ∀i
zij binary
Introduction Problem description Methodology Experimental results Conclusions
L
ij= F (running times of two trains)
For each train pair at each timetable point:
vary running speeds = amount of time supplements
compute minimum headway time for each trains-speeds variations
get functional relationship between given time supplements and minimum headways → Lij
Expected: bigger speed difference → bigger minimum headway time
more homogenized running times → smaller minimum headway time
Introduction Problem description Methodology Experimental results Conclusions
L
ij= F (running times of two trains)
For each train pair at each timetable point:
vary running speeds = amount of time supplements
compute minimum headway time for each trains-speeds variations
get functional relationship between given time supplements and minimum headways → Lij
Expected: bigger speed difference → bigger minimum headway time
more homogenized running times → smaller minimum headway time
Introduction Problem description Methodology Experimental results Conclusions
L
ij= F (running times of two trains)
runik - running time supplement of the first train (in %) runjl - running time supplement of the second train (in %)
Rij - relative difference between time supplements of two trains (in %) Rij = runik− runjl
Introduction Problem description Methodology Experimental results Conclusions
L
ij= F (running times of two trains)
runik - running time supplement of the first train (in %) runjl - running time supplement of the second train (in %)
Rij - relative difference between time supplements of two trains (in %) Rij = runik− runjl
Introduction Problem description Methodology Experimental results Conclusions
L
ij= F (running times of two trains)
−0.2 −0.1 0 0.1 0.2 80 90 100 110 120 130 140 150
Minimum headway time L
ij
[s]
Headway relation for train lines 6001 and 16001 at station Cl
1.05 1.1 1.15
Introduction Problem description Methodology Experimental results Conclusions
L
ij= F (running times of two trains)
−0.2 −0.1 0 0.1 0.2 80 90 100 110 120 130 140 150
Running time difference run
i−runj [s]
Minimum headway time L
ij
[s]
Headway relation for train lines 6001 and 16001 at station Cl
1.05 1.1 1.15
Introduction Problem description Methodology Experimental results Conclusions
L
ij= F (running times of two trains)
Linear dependency between runik and runjl Lij = αij · Rij+ l0 αij - slope of Lij
Rij - relative difference between time supplements of two trains (in %) l0 - minimum headway time for runik = runjl
Introduction Problem description Methodology Experimental results Conclusions
Integrated approach
(PESP −FlexHeadways) Min f (π, z) such that
lij ≤ πj − πi + zij · T ≤ uij ∀(i , j) ∈ Arun∪ Adwell αij · Rij+ l0≤ πj − πi + zij· T ≤ uij ∀(i , j) ∈ Aheadway Rij = runik− runjl
0 ≤ πi < T , ∀i zij binary
Introduction Problem description Methodology Experimental results Conclusions
Case studies
Case network: Utrecht - Eindhoven network (two intersecting corridors)
15 stations and junctions
40 trains/h
96 events and 148 activities Minimum running time supplement: 5% Maximum running time supplement: 20% Minimum dwell times: 60-120 s
Introduction Problem description Methodology Experimental results Conclusions
Introduction Problem description Methodology Experimental results Conclusions
Computed timetables
Table: Solutions obtained after the first iteration
Model # of conflicts Total time Scheduled time
[train pairs] in conflicts [s] supplements [s]
Iterative micro-macro* 4 160 10
Integrated PESP-FlexHeadway 0 0 382
*After first iteration
Iterative micro-macro framework finished after 10 iterations
PESP-FlexHeadway allocated more time supplements to satisfy new headways
Introduction Problem description Methodology Experimental results Conclusions
Computed timetables
Table: Solutions obtained after the first iteration
Model # of conflicts Total time Scheduled time
[train pairs] in conflicts [s] supplements [s]
Iterative micro-macro* 4 160 10
Integrated PESP-FlexHeadway 0 0 382
*After first iteration
Iterative micro-macro framework finished after 10 iterations
PESP-FlexHeadway allocated more time supplements to satisfy new headways
Introduction Problem description Methodology Experimental results Conclusions
Computed timetables
Table: Solutions obtained after the first iteration
Model # of conflicts Total time Scheduled time
[train pairs] in conflicts [s] supplements [s]
Iterative micro-macro* 4 160 10
Integrated PESP-FlexHeadway 0 0 382
*After first iteration
Iterative micro-macro framework finished after 10 iterations
PESP-FlexHeadway allocated more time supplements to satisfy new headways
Introduction Problem description Methodology Experimental results Conclusions
Computed timetables
Table: Solutions obtained after the first iteration
Model # of conflicts Total time Scheduled time
[train pairs] in conflicts [s] supplements [s]
Iterative micro-macro* 4 160 10
Integrated PESP-FlexHeadway 0 0 382
*After first iteration
Iterative micro-macro framework finished after 10 iterations
PESP-FlexHeadway allocated more time supplements to satisfy new headways
Introduction Problem description Methodology Experimental results Conclusions
Iterative micro-macro framework
0 10 20 30 40 50
Time distance diagram for corridor Ut−Ehv
Introduction Problem description Methodology Experimental results Conclusions
Introduction Problem description Methodology Experimental results Conclusions
Integrated framework: PESP-FlexHeadway
0 10 20 30 40 50
Introduction Problem description Methodology Experimental results Conclusions
Introduction Problem description Methodology Experimental results Conclusions
Some more headways...
−0.2 −0.1 0 0.1 0.2 80 90 100 110 120 130
Running time difference run
i−runj [s]
Minimum headway time L
ij
[s]
Headway relation for train lines 3501 and 801 at station Htn
1.05 1.1 1.15 −0.2 −0.1 0 0.1 0.2 70 80 90 100 110 120
Running time difference runi−runj [s]
Minimum headway time L
ij
[s]
Headway relation for train lines 800 and 3500 at station Btl
1.05 1.1 1.15 160 170 180 190 ij [s]
Headway relation for train lines 800 and 6000 at station Htn
Introduction Problem description Methodology Experimental results Conclusions
Conclusions
Main observations:
We cancompute operationally feasible timetables
Iterative approach solves within a limited number of iterations
Minimum headway times as afunction of running times
MacroscopicFlexible minimum headway modelformulation generates (almost) operationally feasible solutions
Pursuing the (passenger) happiness
Is linear approximation always good? Piecewise linear?
Include stability and robustness in the objective function
Introduction Problem description Methodology Experimental results Conclusions
Conclusions
Main observations:
We cancompute operationally feasible timetables
Iterative approach solves within a limited number of iterations
Minimum headway times as afunction of running times
MacroscopicFlexible minimum headway modelformulation generates (almost) operationally feasible solutions
Pursuing the (passenger) happiness
Is linear approximation always good? Piecewise linear?
Introduction Problem description Methodology Experimental results Conclusions