8/7/2019 JeffJohnson
1/72
Complexity Science & Transport Systems
Jeffrey Johnson & Joan Serras
Design, Development, Environment & MaterialsThe Open University
to infinity and beyond
8/7/2019 JeffJohnson
2/72
The Complex Systems Perspective:
Road systems - interconnected systems of dynamic systems.
Many subsystems: Populations, roads, railways, land uses, regulations.All have multilevel structure, e.g. people, families,
neighbourhoods, cities.All have dynamics => emergent dynamics of whole
Subsystems have thousands or millions of partscan only be modelled on computers.
How to represent and modelling huge heterogeneous multilevelroad systems such as the whole of the UK or Europe ?
8/7/2019 JeffJohnson
3/72
You are here You came in here
or hereor here
8/7/2019 JeffJohnson
4/72
You are here
8/7/2019 JeffJohnson
5/72
8/7/2019 JeffJohnson
6/72
8/7/2019 JeffJohnson
7/72
8/7/2019 JeffJohnson
8/72
8/7/2019 JeffJohnson
9/72 and youre still in England Wales !
Theres a lot of stuff
8/7/2019 JeffJohnson
10/72 how can we represent so much stuff ?
Theres a lot of stuff
8/7/2019 JeffJohnson
11/72 how can we represent so much stuff ?
Theres a lot of stuff
8/7/2019 JeffJohnson
12/72from micro to macro ?
Theres a lot of stuff
8/7/2019 JeffJohnson
13/72
Traffic Modelling Conference 31-March-2009 Open University
Everything is becoming more & more connected
MiltonKeynes
Bedford
8/7/2019 JeffJohnson
14/72
Traffic Modelling Conference 31-March-2009 Open University
Everything is becoming more & more connected
MiltonKeynes
Bedford
8/7/2019 JeffJohnson
15/72
Traffic Modelling Conference 31-March-2009 Open University
Everything is becoming more & more connected
MiltonKeynes
Bedford
8/7/2019 JeffJohnson
16/72
Traffic Modelling Conference 31-March-2009 Open University
Everything is becoming more & more connected
MiltonKeynes
Bedford
Cannot divide & rule
- the are connected !
8/7/2019 JeffJohnson
17/72
Traffic Modelling Conference 31-March-2009 Open University
Objectives of the research
To model road system in an holistic way such that:
captures the dynamics of road traffic:
microdynamics: e.g. shock wavesmacrodynamics: e.g. transmission of congestion
8/7/2019 JeffJohnson
18/72
Traffic Modelling Conference 31-March-2009 Open University
Objectives of the research
To model road system in an holistic way such that:
captures the dynamics of road traffic:
microdynamics: e.g. shock wavesmacrodynamics: e.g. transmission of congestion
includes all the smallest level links nothing left out!
aggregates bottom-up dynamics no information loss!
8/7/2019 JeffJohnson
19/72
Traffic Modelling Conference 31-March-2009 Open University
Objectives of the research
To model road system in an holistic way such that:
captures the dynamics of road traffic:
microdynamics: e.g. shock wavesmacrodynamics: e.g. transmission of congestion
includes all the smallest level links nothing left out!
aggregates bottom-up dynamics no information loss!
distributes computation and data naturally acrossadministrative levels
8/7/2019 JeffJohnson
20/72
Traffic Modelling Conference 31-March-2009 Open University
Objectives of the research
To model road system in an holistic way such that:
captures the dynamics of road traffic:
microdynamics: e.g. shock wavesmacrodynamics: e.g. transmission of congestion
includes all the smallest level links nothing left out!
aggregates bottom-up dynamics no information loss!
distributes computation and data naturally acrossadministrative levels
illustrate by huge road systems, e.g. London, Europe
8/7/2019 JeffJohnson
21/72
Traffic Modelling Conference 31-March-2009 Open University
Starting point: it is like this
8/7/2019 JeffJohnson
22/72
Traffic Modelling Conference 31-March-2009 Open University
Starting point: it is like this
Plots of land zones
8/7/2019 JeffJohnson
23/72
Traffic Modelling Conference 31-March-2009 Open University
Starting point: it is like this
Plots of land zones
ActivityLand Uses
8/7/2019 JeffJohnson
24/72
Traffic Modelling Conference 31-March-2009 Open University
Starting point: it is like this
Plots of land zones
ActivityLand Uses
8/7/2019 JeffJohnson
25/72
Traffic Modelling Conference 31-March-2009 Open University
Starting point: it is like this
origin
destination
Trips are madebetween locatedactivities
8/7/2019 JeffJohnson
26/72
Traffic Modelling Conference 31-March-2009 Open University
Starting point: it is like this
origin
destination
8/7/2019 JeffJohnson
27/72
Traffic Modelling Conference 31-March-2009 Open University
Starting point: it is like this
origin
destination
8/7/2019 JeffJohnson
28/72
Traffic Modelling Conference 31-March-2009 Open University
Starting point: it is like this
origin
destination
8/7/2019 JeffJohnson
29/72
Traffic Modelling Conference 31-March-2009 Open University
Starting point: it is like this
origin
destination
8/7/2019 JeffJohnson
30/72
8/7/2019 JeffJohnson
31/72
Traffic Modelling Conference 31-March-2009 Open University
Starting point: it is like this
origin
destination
A route traversesa set of zones atdifferent levels
8/7/2019 JeffJohnson
32/72
Traffic Modelling Conference 31-March-2009 Open University
Starting point: it is like this
Plots aggregateinto larger areas
zones
8/7/2019 JeffJohnson
33/72
Traffic Modelling Conference 31-March-2009 Open University
Zones aggregate into larger zones
8/7/2019 JeffJohnson
34/72
Traffic Modelling Conference 31-March-2009 Open University
Zones aggregate into larger zones
8/7/2019 JeffJohnson
35/72
Traffic Modelling Conference 31-March-2009 Open University
Zones aggregate into larger zones
8/7/2019 JeffJohnson
36/72
Traffic Modelling Conference 31-March-2009 Open University
Larger zones aggregate into even larger zones
8/7/2019 JeffJohnson
37/72
Traffic Modelling Conference 31-March-2009 Open University
Larger zones aggregate into even larger zones
8/7/2019 JeffJohnson
38/72
Traffic Modelling Conference 31-March-2009 Open University
Larger zones aggregate into even larger zones
8/7/2019 JeffJohnson
39/72
Traffic Modelling Conference 31-March-2009 Open University
Even larger zones aggregate into even larger zones
8/7/2019 JeffJohnson
40/72
Traffic Modelling Conference 31-March-2009 Open University
Even larger zones aggregate into even larger zones
8/7/2019 JeffJohnson
41/72
Traffic Modelling Conference 31-March-2009 Open University
and so on to zones for London, England, Europe,
8/7/2019 JeffJohnson
42/72
Traffic Modelling Conference 31-March-2009 Open University
We assume there is a hierarchical set of zones
ZN-3 Z N-1 ZN ZN+2 ZN+k-1 ZN+k
8/7/2019 JeffJohnson
43/72
Traffic Modelling Conference 31-March-2009 Open University
We assume there is a hierarchical set of zones
link between junction nodes
kitchen
houseplot,road
segment
ZN-3 Z N-1 ZN ZN+2 ZN+k-1 ZN+k
8/7/2019 JeffJohnson
44/72
Traffic Modelling Conference 31-March-2009 Open University
We assume there is a hierarchical set of zones
link between junction nodes
kitchen
houseplot,road
segment
ZN-3 Z N-1 ZN ZN+2 ZN+k-1 ZN+k
Abstract higher level lines(to be explained)
8/7/2019 JeffJohnson
45/72
Traffic Modelling Conference 31-March-2009 Open University
Define a set of nodes -
where a vehicle can cross aroad zone boundary
8/7/2019 JeffJohnson
46/72
Traffic Modelling Conference 31-March-2009 Open University
Define a set of nodes -
where a vehicle can cross aroad zone boundary
A link is defined by a pair of nodes
8/7/2019 JeffJohnson
47/72
Traffic Modelling Conference 31-March-2009 Open University
Define a set of nodes -
where a vehicle can cross aroad zone boundary
A link is defined by a pair of nodes
conventional links exist at Level N
8/7/2019 JeffJohnson
48/72
Traffic Modelling Conference 31-March-2009 Open University
A
B
A Level-N route is a set of Level N links
8/7/2019 JeffJohnson
49/72
Traffic Modelling Conference 31-March-2009 Open University
A
B
There are many N-Level routes between A and B
8/7/2019 JeffJohnson
50/72
Traffic Modelling Conference 31-March-2009 Open University
A
B
There are many N-Level routes between A and B
8/7/2019 JeffJohnson
51/72
Traffic Modelling Conference 31-March-2009 Open University
A
B
There are many N-Level routes between A and B
8/7/2019 JeffJohnson
52/72
Traffic Modelling Conference 31-March-2009 Open University
A
B
Let this set of routes be a Level N+2 Link
8/7/2019 JeffJohnson
53/72
Traffic Modelling Conference 31-March-2009 Open University
There are ~8000 Level-N routes between A and B
AB
~ 20 routes ~ 20 routes ~ 20 routes
8/7/2019 JeffJohnson
54/72
Traffic Modelling Conference 31-March-2009 Open University
Put a N+2-level nodes on the boundaries
AB
8/7/2019 JeffJohnson
55/72
Traffic Modelling Conference 31-March-2009 Open University
Make Level N+2 links and a Level N+2 route
AB
8/7/2019 JeffJohnson
56/72
Traffic Modelling Conference 31-March-2009 Open University
Make Level N+2 links and a Level N+2 route
AB
8/7/2019 JeffJohnson
57/72
Traffic Modelling Conference 31-March-2009 Open University
Make Level N+2 links and a Level N+2 route
AB
8/7/2019 JeffJohnson
58/72
Traffic Modelling Conference 31-March-2009 Open University
We now have 20 + 20 + 20 + 1 = 61
8/7/2019 JeffJohnson
59/72
Traffic Modelling Conference 31-March-2009 Open University
We now have ~ 100,000 possible routes !
AB
~ 50 routes ~ 50 routes ~ 50 routes
8/7/2019 JeffJohnson
60/72
Traffic Modelling Conference 31-March-2009 Open University
We now have ~150 + 9 = ~159
8/7/2019 JeffJohnson
61/72
Traffic Modelling Conference 31-March-2009 Open University
If the routes cross a Level N+3 Zone ..
AB
8/7/2019 JeffJohnson
62/72
Traffic Modelling Conference 31-March-2009 Open University
AB
make a Level N+3 link as the set of N+2 routes
8/7/2019 JeffJohnson
63/72
Traffic Modelling Conference 31-March-2009 Open University
AB
Hierarchical routes used for routing trips
Level N+2 linksLevel N+1 linksLevel N+1
Level N linksLevel N links
Within hierarchical zones, always use the highestlevel links available
8/7/2019 JeffJohnson
64/72
Traffic Modelling Conference 31-March-2009 Open University
AB
Higher level links have lower variance !
Level N+2 linksLevel N+1 linksLevel N+1
Level N linksLevel N links
relative
frequenc
y
travel time, minutes travel time travel time, hours
100%
relative
frequenc
y
relative
frequency
8/7/2019 JeffJohnson
65/72
Traffic Modelling Conference 31-March-2009 Open University
AB
Hierarchical routes used for routing trips
Level N+2 linksLevel N+1 linksLevel N+1
Level N linksLevel N links
Note: higher level links have more stable flowand travel time statistics than lower level links initialisation heuristics ?
8/7/2019 JeffJohnson
66/72
Traffic Modelling Conference 31-March-2009 Open University
TRANSIMS from Micro simulation to Macro dynamics
8/7/2019 JeffJohnson
67/72
Traffic Modelling Conference 31-March-2009 Open University
TRANSIMS from Micro simulation to Macro dynamics
8/7/2019 JeffJohnson
68/72
Traffic Modelling Conference 31-March-2009 Open University
TRANSIMS from Micro simulation to Macro dynamics
8/7/2019 JeffJohnson
69/72
Traffic Modelling Conference 31-March-2009 Open University
TRANSIMS from Micro simulation to Macro dynamics
8/7/2019 JeffJohnson
70/72
Traffic Modelling Conference 31-March-2009 Open University
TRANSIMS from Micro simulation to Macro dynamics
8/7/2019 JeffJohnson
71/72
Traffic Modelling Conference 31-March-2009 Open University
Level N+7
Level N+6
Level N+5
Level N+4
Level N+3
Level N+2
Level N+1
Level N
Level N-1
Level N-2
Level N-3
The World !
Region (e.g. the Americas, Asia )
Nation (e.g. USA, UK )
State (e.g. Virginia, Buckinghamshire)
City, rural region (e.g. New River Valley)
Small town (e.g. Blacksburg, Milton Keynes)
Neighbourhood (e.g. VT Campus)
Conventional Links and zones
Road segment, plot of land
House, garden, garage
Rooms
8/7/2019 JeffJohnson
72/72
Research in Progress
1. Devise new microlevel simulation
2. Implement at microlevel for Milton Keynes
3. Implement hierarchical routing schemes
4. New synthetic micropopulation all properties
5. Run the system for Milton Keynes
6. extend to UK7. extend to Europe
8. extend to infinity, and beyond !