2005 Annual Meeting of the Association of American Geographers,
Denver, Colorado, April 5-9
Micro-simulation and visualization
of individual space-time paths
within a GIS
A bouquet of alternatives
(G) Arnaud Banos, Pau University/CNRS, France
(CS) Bruno Jobard, Pau University, France
(S) Sylvain Lassarre, INRETS, France
(CS) Julien Lesbegueries, Pau University, France
(G) Pierpaolo Mudu, WHO, Italy
(CS) Karine Zeitouni, Versailles University, France
G : Geographer ; CS : Computer Scientist ; S : Statistician
Contents

Urban daily mobility
 Simulation
 “What

if.. ?” scenarios
Hägerstrand conceptual framework
 Monte-Carlo
approach to diffusion : Macro
level
 Time-Geography : Micro level

From concepts to methods and techniques
“A Monte-Carlo approach to urban rythms”
O/D matrix
(time period, mode, activity)
[T1]
D1
D2
D3

O1
 O1
O2
 O2
O3
 O3
 D1
 D2
 D3

GIS

Monte-Carlo
Banos & Thévenin, 2001
Limits

Global view of urban “pulses” based on a
very segmented approach of mobility :
 focused
on independent activities
 loosing trip chaining
 loosing the very basic dimension of urban
systems : INDIVIDUALS
Time Geography
Space-time cube
Space-time path
Trip chaining
Typical data available in France
08:00
08:10
08:35
08:38
Zone 1
Zone 2
Zone 3
Zone 3
1
2
Lille :
• 1 million inhabitants
• 13000 sample survey
3
4
Can we simulate their space-time paths ?
Generic problem in Monte-Carlo simulation
of individual daily space-time activities

Simulating activity
scheduling by picking at
random in time
distributions, under
flexible spatial
constraints, to ensure
global trends to be
respected (O/D matrix)
A systematic Time
Geographic approach
Potential Path Area
[Miller, 2003]
Potential Path Area
Area :
30 km2
10000 cells
Network :
100 000 nodes
From Land use to probability Field
Area :
30 km2
25000 objects
Network :
100 000 nodes
Various probability fields
Residences : RPF
 Work places : WPF
 Shops : SPF

H
W
S
T1
T2
08:00
08:10
Zone 1
P
P11
P12
P13
P14
…
P1n
T3
17h30
17:45
Zone 2
RPF
Cells
Z11
Z12
Z13
Z14
…
Z1n
H
WPF
t1
t2
tn
Cells
Z21
Z22
Z23
Z24
…
Z2n
P
P21
P22
P23
P24
…
P2n
t1
t2
tn
18:30
19h
Zone 1
Zone 1
SPF
RPF
Cells
Z11
Z12
Z13
Z14
…
Z1n
P
P11
P12 t2
P13
P14
…
P1n
t1
Z13
tn
RP(Z11, Z12, Z13, Z1n)
RP[(t1, t2, t3, tn) = T1+- e]
R{[(t1, t2, t3, tn) = T2+- e] INTERSECT [(t, t2, t3, tn) = T3+- e]}
Shortest path
Perspectives
Straightforward translation of concepts
into methods
 HUGE COMPUTATION BURDEN !!!
(10
000 cells, 100 000 nodes)

A swarming approach
Stigmergy
Food
Ants
Ants Nest
Pheromones Trail
Netlogo
http://ccl.northwestern.edu/netlogo/
Prototype
Zone 2
Zone 3
Forward Ants
Backward Ants
Zone 1
Tour to realize :
Z2 --> Z3 --> Z4 --> Z2
Distances to respect :
30 --> 30 --> 44
Zone 4
Swarming Algorithm (Dorigo, 1996)


Locate N/2 forward and N/2
backward ants on node i in
Zone m=0
Each ant k :

Random proportional rule
Move at time t to a connected
node j using a probabilistic
action choice rule :
1
d ij
Pheromone trail
p (t ) 
k
ij
[ ij (t )]  [ij ]


[

(
t
)]

[

]
 ij
ij
if j  N ik
lN ik
Feasible neighbourhood
of ant k ant node i
Updating pheromones trails
Pheromones = pheromones deposit – pheromones evaporation
Amount of pheromones
at edge ij
Reinforcement learning scheme
to favour better solutions
m
 ij  (1  r )  ij    ijk
k 1
Pheromones decay
parameter (0<r<1)
1

k
if
(ij)

tour
done
by
ant
k
and
cumd
 cumdm  0
ij
 cumd k  cumd
ij
m


where  ijk  1
if (ij)  tour done by ant k and cumdijk  cumdm  0


0
otherwise

Actual situation (debugging !)
What comes next ?
GeoVisualisation ?
Mei-Po Kwan, 2000
A bouquet of alternatives based on
mobile objects
GIS : Grass, Postgis (PostgreSQL)
 Visualization : VTK

Banos, Jobard, Lesbegueries (ICC 2005)
Applications ?
Exposure of citizens to urban
transport hazards
Tomorrow afternoon : Session 5505, Applied Transportation Research Projects
Sylvain LASSARRE (5:05)
T5
Origin
T4
Destination
T3
T3 – T5
T3
Origin
T2
Destination
T1
Y
T1-T3
X
Simulation of Artificial Urban Life


MIRO project, French Ministry of Transportation
Agent Based Modelling :

Heterogeneous cognitive agents (Von BDI)
 Limited knowledge (CFOS) and computation capacities
 Interacting locally with their urban environment and with other agents
 Having to program their daily calendar of activities and to perform their
activities in a moving urban environment (traffic conditions, other agents,
time schedule of urban opportunities, public transport availability…)


Goal : testing “what if…?” scenarios by modifying the opportunity
constraints at a global level (public transport, opening/closing time of
public services, schools, universities, shops…) : leave the system
show us how agents react to these various time geographic
constraints (capacity, conjunction, authority constraints)
MORE at CUPUM’05, London
Perspectives





Applying Time Geography is still a challenge…
…what is more when dealing with large
populations !
Various methodological and technological
translations, and more to be invented !
No one best way ! (Herbert Simon)
Time Geo is still alive and remains a major
concern!
Links

HEARTS
 http://www.euro.who.int/hearts

MIRO
 http://lifc.univ-fcomte.fr/~lang/MIRO

Animations
 Http://www.univ-pau.fr/~banos/banos.html