The Dynamic Geometry of Geographical Vector Agents

Download Report

Transcript The Dynamic Geometry of Geographical Vector Agents

The Dynamic Geometry of
Geographical Vector Agents
Yasser Hammam, Antoni Moore,
Peter Whigham and Claire
Freeman*
Spatial Information Research Centre,
Department of Information Science,
*Department of Geography
University of Otago, New Zealand
Overview
Rationale
The Vector Agent
Model Implementation and
Experimental Results
Discussion
Conclusions
Introduction
Like Geographic Automata Systems,
Vector Agents aim to introduce a bit of
geographic realism to agent modelling
But adds a systematic framework to the
geometric element (“georeferencing
convention”)
With emphasis on irregular and dynamic
aspects
Aim to use boundary manipulation through
simple controls to generate vector objects
of a wide variety of shapes and
complexities
Pertinence to real world object characteristics
are key to effectiveness
The Vector Agent
Uses irregular fractal-like process to
generate vector objects
Also more direct boundary manipulation
Aims
To represent any discrete geographic
phenomena through an irregular (or
regular) data structure
May move “bodily”, either based on a
real world object, or is “born” with a nondeterministic shape boundary
Abstracted so that it is able to define its
own location in space
Regular – irregular
Static – dynamic
Geometric manipulation
Conventional midpoint displacement
Pnew = 0.5 (P1 + P2) + µơ02-lh
Where P1 and P2 are the start and end
points of the line segment
µ is a random number from a Gaussian
ơ0 is the S.D. of the Gaussian
l is the level of recursivity
h is the Hurst exponent governing
roughness (= 2 – F.D.)
Point displacement (not nec.midpoint)
Pnew = (1 – r)P1 + rP2 + µơ02-lh
Where r is proportion along line segment
Edge / vertex displacement
P = P + µơ0
Sequence of growth
a
b
c
g
d
h
e
i
f
j
Results show evidence of
both irregular and regular
Graphs
Shape control schematic
MIDPOINT
DISPLACEMENT
high
Control
of
shape complexity
EDGE
DISPLACEMENT
high
SHAPE GROWTH RATE
low
low
VERTEX
DISPLACEMENT
Cities with
similar
shapes
Conclusion
Vector agents are geometry-led
agents
Interplay of midpoint, edge and vertex
displacement
Evolved polygons have shape and
complexity characteristics of real-world
objects
Able to be controlled by alteration of
simple parameters
Conclusion
Like GAS, each object has its own
identity, an improvement on a group
of contiguous CA cells
But complexity and processing speed
ramped up
Next
Test the system on a specific urban area
Having tested on classical urban models 1st
Build in other elements of GAS
States, transition rules, neighbourhood,
neighbourhood rules