Traffic assignment methods available in OmniTRANS transport planning software, categorized using the framework described in https://www.tandfonline.com/doi/abs/10.1080/01441647.2016.1207211 .

1. Luuk Brederode
2018-03-16

2. Contents
• Scope
• Classification of Traffic Assignment models
• What does it mean for the model user? (and implementation in OmniTRANS)
• Ongoing developments
pagina 2

4. Scope
Macroscopic traffic assignment models for motorized private transport in
strategic, large scale transport model systems.
• Traffic assignment model
• Motorized private transport
• Macroscopic
• Strategic
• Large scale
pagina 4

5. Scope (1/4)
Macroscopic traffic assignment models for motorized private transport in
strategic, large scale transport model systems.
• Goal: to calculate link flows and speeds by confronting travel demand with
network supply
• Travel demand is given in form of OD-matrices per vehicle class
• Network supply is given in form of digitized road network
• Links (speed, capacity)
• Nodes
• Turning movements/prohibitions
• Junction configurations (optionally)
pagina 5

6. Scope (1/4) (continued)
Macroscopic traffic assignment models for motorized private transport in
strategic, large scale transport model systems.
• All traffic assignment models consist of two submodels:
pagina 6Adopted from:
Bliemer, M.C.J., Raadsen, M.P.H., Brederode, L.J.N., Bell, M.G.H., Wismans, L.J.J., Smith, M.J., 2017.
Genetics of traffic assignment models for strategic transport planning. Transp. Rev. 37, 56–78.

7. Scope (2/4)
Macroscopic traffic assignment models for motorized private transport in
strategic, large scale transport model systems.
• Only cars/trucks/vans/buses
• Trains/metros/trams/bikes/pedestrians/… are not explicitly modelled
• But their effect on motorized transport might be
pagina 7

8. Scope (3/4)
Macroscopic traffic assignment models for motorized private transport in
strategic, large scale transport model systems.
• Macroscopic representation of flows:
• Demand, supply and results are aggregated over space (odpairs, routes,
links, nodes) and time (minutes, hours, days)
• No random components
pagina 8

9. Scope (4/4)
Macroscopic traffic assignment models for motorized private transport in
strategic, large scale transport model systems.
• Transport model system:
• System with an assignment model integrated with a (OD) demand model
• Strategic:
• Mainly used for long term prognosis
• Primary output are relative differences between scenario’s, not so much
the absolute outputs of those scenario’s
• Large scale:
• ~3000 centroids / 1mln used OD pairs
• ~100.000 links
pagina 9

11. Framework (focusing on network loading model)
pagina 11
Static
Semi-dynamic
Dynamic
Unrestrained Capacity
Restrained
Capacity
Constrained
Capacity & Storage
Constrained
Simplified from:
Bliemer, M.C.J., Raadsen, M.P.H., Brederode, L.J.N., Bell, M.G.H., Wismans, L.J.J., Smith, M.J., 2017.
Genetics of traffic assignment models for strategic transport planning. Transp. Rev. 37, 56–78.
Spatial interaction assumptions
Temporalinteractionassumptions
-least ‘capable’ model
-best scalable / fastest model
-nicest mathematical properties
-most ‘capable’ model
-least scalable / slowest model
-worst mathematical properties

12. Spatial interaction assumptions
pagina 12Adopted from:
Brederode, L.J.N., Pel, A.J., Wismans, L.J.J., de Romph, E., Hoogendoorn, S.P. (2018, in press)
STAQ – Static Traffic Assignment with Queuing: properties and applications. Transport Metrica / A: Transport Science 1–36.
Capacity
Restrained
Capacity
Constrained
Capacity & Storage
Constrained
Link model
Node model None Existent Existent

13. Temporal interaction assumptions
pagina 13Adopted from:
Bliemer, M.C.J., Raadsen, M.P.H., Brederode, L.J.N., Bell, M.G.H., Wismans, L.J.J., Smith, M.J., 2017.
Genetics of traffic assignment models for strategic transport planning. Transp. Rev. 37, 56–78.
Modelled Demand
‘True’ Demand

14. Framework and most used models in practice
pagina 14
Semi-dynamic
Unrestrained Capacity
Restrained
Capacity
Constrained
Capacity & Storage
Constrained
Simplified from:
Bliemer, M.C.J., Raadsen, M.P.H., Brederode, L.J.N., Bell, M.G.H., Wismans, L.J.J., Smith, M.J., 2017.
Genetics of traffic assignment models for strategic transport planning. Transp. Rev. 37, 56–78.
Spatial interaction assumptions
Temporalinteractionassumptions
Static
Dynamic
‘All-Or-Nothing’
(Dijkstra, 1959)
‘Macroscopic Dynamic’
(CTM, Daganzo (1994);
LTM, Yperman (2007))
‘Static Equillibrium’
(Beckmann et al, 1956)

15. Classification of traffic assignment models
pagina 15
Semi-dynamic
Unresponsive to
congestion
Route distribution
due to congestion
Vertical queues
due to congestion
Horizontal queues
due to congestion
Simplified from:
Bliemer, M.C.J., Raadsen, M.P.H., Brederode, L.J.N., Bell, M.G.H., Wismans, L.J.J., Smith, M.J., 2017.
Genetics of traffic assignment models for strategic transport planning. Transp. Rev. 37, 56–78.
Spatial interaction assumptions
Temporalinteractionassumptions
‘All-Or-Nothing’
(Dijkstra, 1959)
‘Macroscopic Dynamic’
(CTM, Daganzo (1994);
LTM, Yperman (2007))
‘Static Equillibrium’
(Beckmann et al, 1956)
Static
Dynamic
‘STAQ squeezing’
(Brederode et al, 2018)
‘STAQ queuing’
(Brederode et al, 2018)

17. 2 route network with bottlenecks on shortest route
pagina 17
Link capacities [pcu/h]
• Demand: 10000 pcu/h
• Free Flow travel time upper route: 4:30 min
• Free Flow travel time lower route: 5:00 min
Flows (width) and VC ratio’s (color, label) AON assignment results

18. 2 route network
pagina 18
All-Or-Nothing
OtTraffic
User Equillibrium
OtTraffic
STAQ-squeezing
OtStreamLine
STAQ-queuing
OtStreamLine
MaDAM
OtStreamLine
Width: flow [veh/h]
Colour: modelled speed as percentage
of maximum speed [%]
80%100%0%
Static - unconstrained
Static – capacity restrained
Static – capacity constrained
Static – capacity and storage constrained
Dynamic – capacity and storage constrained
V/C ratio > 1
Spillback blocks
lower route

19. Pilot studies STAQ
Flanders (Belgium)
•Client: Flemish ministry of mobility
•Models: model of Leuven (500 zones) and provincial model of Vlaams Brabant
(4000 zones)
•Status: Pilot using early beta version, finished 2013
NRM-WEST (Netherlands) (no slides on this project)
•Client: Dutch ministry of infrastructure and environment
•Model: regional model of Randstad metropolitan area (NRM WEST) (2400
zones, MUC)
•Status: Pilot using late beta version, finished in 2014
pagina 19

20. Pilot studies STAQ
Haaglanden
•Client: city of The Hague
•Model: Haaglanden (urban metropolitan area The Hague, 6700 zones)
•Status: pilot using late beta version, finished in 2014
Onderzoekstraject STAQ in BBMA
•Client: Province of Noord Brabant
•Models: BBMA (province wide base model, 3321 zones) and model of Breda
City Region (6043 zones)
•Status: pilot using version 1.0, finished in 2015
pagina 20

21. Added value of capacity constraints: Case Study
pagina 21
added 1 extra lane
Increased junction
capacity
#lanes
Adopted from: Brederode, L.J.N., Heijnickx, M., Koopal, R., 2016. Quasi Dynamic Assignment on the Large Scale
Congested Network of Noord-Brabant. AET 2016 and contributors.

22. Legend:
Bandwith widths: flow (veh/h AM peak)
Bandwith colors: speed (% of free flow speed):
Pie charts: vertical queue sizes (vehicle loss hours)
pagina 22
Added value of capacity constraints: reference scenario
Assignment results from Static
traffic assignment model
Assignment results from STAQ
80% 100%0%
308

23. pagina 23
Added value of capacity constraints: scenario with measures
1. Bottleneck on off-ramp
disappears
2. Bottleneck on freeway
disappears
3. More traffic on Berlicumseweg
4. Bottlenecks and spillback
downstream increase
5. Less delay on beltway Den
Bosch
6. More inflow from beltway Den
Bosch, activation of new
bottleneck on highway
intersection
1
2
4
4
3
5
6
3
4
1
2
Assignment results from Static
traffic assignment model
Assignment results from STAQ
Legend:
Bandwith widths: flow (veh/h AM peak)
Bandwith colors: speed (% of free flow speed):
Pie charts: vertical queue sizes (vehicle loss hours)
80% 100%0%
308

24. Added value of dynamic demand – Flanders STAQ squeezing
pagina 24

25. Added value of dynamic demand – Flanders STAQ queuing
pagina 25

26. Added value of dynamic demand – Flanders MaDAM (08:30)
pagina 26

27. Conclusions from Flanders study: STAQ scalability
•STAQ is approximately 3-8 times slower compared to static and 500-1200
times faster compared to Madam
•For each 100.000 routes about 125 Mb of memory is required
•For each 300.000 routes calculation time is about 1 minute per iteration
•Models up to 6102 zones succesfully run on laptop with 4Gb ram
•Calculation times on my old laptop (quadcore i7 2.67 Ghz, 4Gb):
• 4.5 second per iteration on model Leuven (430 zones)
• 5 minutes per iteration on model Vlaams Brabant (4000 zones)
pagina 27

28. Flanders: STAQ vs MADAM conclusions
Flanders study - specific
+ Congestion seeds on the same locations
+ Congestion seeds available after squeezing fase usefull for detection of network errors
+ Average travel times roughly of same quality
+ Calculation time and RAM usage of STAQ much lower
- Travel times of both Madam and STAQ to low on urban roads
caused by an OD matrix with too low demand, since it has been calibrated using a static traffic assignment model
More general:
+ STAQ scales to the largest models in the Netherlands
- Operational measures within a peak period cannot be modelled:
ramp metering, dynamic route info panels, traffic towards big events, incidentmanagement
+ All tested modelscenarios converge to a unique user equillibrium
6 different (synthetic and realistic) starting solutions tested by intern (Anton Dijkstra, Twente university)pagina 28

29. Haaglanden – congestion patterns PM peak
Static PM STAQ PM
Adopted from: Possel, B. (2015) STAQ in Haaglanden (in dutch), presented at PLATOS conference 2015,
http://www.platos-colloquium.nl/documents/2015/P1.1- 2%20Possel%20-20STAQ%20voor%20Haaglanden.pdf

30. Haaglanden – congestion patterns PM peak
STAQ PMHERE PM
Adopted from: Possel, B. (2015) STAQ in Haaglanden (in dutch), presented at PLATOS conference 2015,
http://www.platos-colloquium.nl/documents/2015/P1.1- 2%20Possel%20-20STAQ%20voor%20Haaglanden.pdf

31. Haaglanden – congestion patterns AM peak
Static AM STAQ AM
Adopted from: Possel, B. (2015) STAQ in Haaglanden (in dutch), presented at PLATOS conference 2015,
http://www.platos-colloquium.nl/documents/2015/P1.1- 2%20Possel%20-20STAQ%20voor%20Haaglanden.pdf

32. Haaglanden – congestion patterns AM peak
STAQ AMHERE AM
Adopted from: Possel, B. (2015) STAQ in Haaglanden (in dutch), presented at PLATOS conference 2015,
http://www.platos-colloquium.nl/documents/2015/P1.1- 2%20Possel%20-20STAQ%20voor%20Haaglanden.pdf

33. Haaglanden: Travel times AM peak
pagina 33

34. Haaglanden: Travel times PM peak
pagina 34

35. Haaglanden: STAQ vs HERE conclusions
STAQ vs HERE:
+ Congestion seeds highways on the same locations
+ Congestion seeds urban roads mostly on same locations (exact definition and
timing of intersections was not available for all intersections)
+ Travel times of most routes within error margins of observed travel times, far
better than static travel times
- Queues on urban roads in STAQ more compressed (probably due to lack of
higher order effects)
More general:
+/- Capacity needs to be modelled more accurate than was done in the static
network. Capacity of buffer spaces, weaving lanes cannot be neglected
- Convergence on the very busy 2020PM scenario is only just acceptable
pagina 35

36. Calculation time vs convergence
(effect of storage constraints, junction modelling and averaging scheme)
pagina 36
Leuven
~75.000 routes
NRM-West
~1.240.000 routes
4 minutes 2 hours
Adopted from:
Brederode, L.J.N., Pel, A.J., Wismans, L.J.J., de Romph, E., Hoogendoorn, S.P. (2018, in press)
STAQ – Static Traffic Assignment with Queuing: properties and applications. Transport Metrica / A: Transport Science 1–36.

37. Calculation time vs convergence
(effect of storage constraints, junction modelling and averaging scheme)
pagina 37
Province of Noord Brabant
~1.272.000 routes
Breda City Region
~2.069.000 routes
3 hours 8 hours
Adopted from:
Brederode, L.J.N., Pel, A.J., Wismans, L.J.J., de Romph, E., Hoogendoorn, S.P. (2018, in press)
STAQ – Static Traffic Assignment with Queuing: properties and applications. Transport Metrica / A, forthcoming

38. pagina 38
The Hague City region
~2.854.000 routes
Province of Flemish Brabant
~3.109.000 routes
9 hours 8 hours
Calculation time vs convergence
(effect of storage constraints, junction modelling and averaging scheme)
Adopted from:
Brederode, L.J.N., Pel, A.J., Wismans, L.J.J., de Romph, E., Hoogendoorn, S.P. (2018, in press)
STAQ – Static Traffic Assignment with Queuing: properties and applications. Transport Metrica / A, forthcoming

39. • Optimal STAQ variation: spillback only in the last iteration, full
junction modelling and the self-regulating averaging scheme
• five of the seven models tested reach convergence (DG 1E-
04)
• calculation times ranging from 23 minutes up to 3 hours on a
regular desktop pc using 2.2 up to 7.6 GB of memory to
achieve equilibrium
• The network of Vlaams Brabant and NVM proved to be too
coarse in relation to its density, creating artificial congestion
locations causing poor convergence properties.
• Dynamic assignment models rarely reach DG values below 0.01…
pagina 39
Nation Wide model of The Netherlands
~4.057.000 routes
16 hours
Calculation time vs convergence
(effect of storage constraints, junction modelling and averaging scheme)
Adopted from:
Brederode, L.J.N., Pel, A.J., Wismans, L.J.J., de Romph, E., Hoogendoorn, S.P. (2018, in press)
STAQ – Static Traffic Assignment with Queuing: properties and applications. Transport Metrica / A, forthcoming

40. Overall conclusions from pilot studies
• STAQ adds capacity (and storage) constraints, thereby:
• Improving model accuracy (congestion patterns, travel times)
• Maintaining acceptable convergence properties
• Scaling to the largest models existing in the Netherlands
• STAQ requires
• More calculation time than capacity restrained models, but still moderate
calculation times compared to DTA models
• Accurate network coding, especially capacity values
• A matrix estimation technique accounting for congestion
pagina 40

42. • Efficient demand matrix estimation using STAQ and observed:
• Link flows
• Congestion patterns
• Route travel times
• Functional prototype
• Applied on NRM west, might be the new
assignment model in NRM/LMS
• For more details, see:
Brederode, L.J.N., Hofman, F., van Grol, R., 2017. Testing of a demand matrix estimation method
Incorporating observed speeds and congestion patterns on the Dutch strategic model system using
an assignment model with hard capacity constraints. Presented at the European Transport
Conference, AET 2017 and contributors.
Brederode, L.J.N., Pel, A.J., Hoogendoorn, S.P., 2014. Matrix estimation for static traffic assignment
models with queuing. hEART 2014 - 3rd symposium of the European association for research of
transportation, Leeds UK.
STAQ and matrix estimation
pagina 42

43. Event Based Generalized Link Transmission Model
• Abbreviated to eGLTM
• Ridiculously fast LTM version implemented into OtStreamline
• In its late beta versions
• Currently adding controls
(ramp metering, VMS, FD
and #lanes actuators)
• For more details, see:
Raadsen, M.P.H., Bliemer, M.C.J., 2018. Continuous-time general link transmission model with simplified
fanning, Part II: Event-based algorithm for networks. Transportation Research Part B: Methodological.
pagina 43

44. Coordinated ramp metering
• Alinea TDI algorithm within StreamLine
• HERO coordination scheme* within StreamLine
• Applied by VLC in Brisbane M1 Paficic motorway (5 on/offramps)
pagina 44
*Papamichail, I., Papageorgiou, M., 2008. Traffic-Responsive Linked Ramp-Metering Control. IEEE
Transactions on Intelligent Transportation Systems 9, 111–121. https://doi.org/10.1109/TITS.2007.908724

45. Ppt AITPM bij Michiel Jagersma opgevraagd t.b.v. animatie
pagina 45
Coordinated ramp metering
More info: Jagersma, M., Brederode, L., Reid, C., 2018. Modelling coordinated ramp metering in strategic transport models.
Presented at the AITPM National Conference, Perth: https://tinyurl.com/ybos33qc

46. Semi dynamic version of STAQ
• Add residual traffic transfer to STAQ-squeezing; not implemented yet
• Relaxes STAQ assumption of empty network before and after study period
pagina 46
Semi-dynamic
Spatial interaction assumptions
Temporalinteractionassumptions
‘All-Or-Nothing’
(Dijkstra, 1959)
‘Macroscopic Dynamic’
(CTM, Daganzo (1994);
LTM, Yperman (2007))
‘Static Equillibrium’
(Beckmann et al, 1956)
Static
Dynamic
‘STAQ squeezing’
(Brederode et al, 2018)
‘STAQ queuing’
(Brederode et al, 2018)
Unrestrained Capacity
Restrained
Capacity
Constrained
Capacity & Storage
Constrained
‘STAQ squeezing
- semi dynamic’