The Computer in the Space Planning Process
Bric Teicholz
Associate Director
Laboratory for Computer Graphics and Spatial Analysis,
Graduate School of besign
Harvard University
Cambridge, Massachusetts 02138
3aThe paper* discuS#8S{Sheral approaches
to using a computer for space plannin
problens in an architectural, urban and
Fegional context, Prograns are classi-
fied as under- and over-constrained, of
the. inproveaent= or construction-type and
various techniques eaployed for space
planning are described and illustrated,
wo programs, GRASP and SETUP, are used
to illustrate the space planning process.
The paper concludes: that! there’ fs! n0
unique solution for space planning prob-
Tens and that additional. research is
needed, priaarily to develop a set of
heuriséie decision rules for internally
modifying the search radius for all of
the different types of space planning
applications that exist.
LyTRODUCTION_AND_ BACKGROUND
Computer-aided space planning is defined
as the use of a digital computer to aid
the designer in the planning of architec-
tural, urban and regional spaces. At a
regional scale, space planning consists
of determining the optimum location and
distribution of various land use patterns,
transportation facilities, warehouses and
other elonents of a regional scale. At
an urban scale, space planning might in-
Clude the distribution of urban trans-
portation systems, the location of city
Services in a metropolitan area and the
layout of buildings in a large urban
complex. At an architectural scalo,
space planning usually refers to the op-
tinal distribution of physical spaces
within a building.
At any scale, conputer-aided space plan-
ning deternines a relationship between
@ physical olement and its onvironment.
‘There also exists a subset of the prob-
lem which is to determine and evaluate
the feasibility of the relationship ac-
Sording to some explicit criteria.
Thore are several problens associated
with automating the space planning pro-
€ess at a regional, urban or architec-
tural scale, Some’of these include:
1, Tho process is often implicit rather
than explicit.
2. Very few evaluation models exist.
3. There is a lack of data pertaining
tO most aspects of the problea ~ in
cluding an existing data base of present
conditions.
*This article is an abridgonent of a pa
per entitled "Computer Aided Space Plan-
hing” published by the Laboratory for
Conputer Graphics and spatial Analysis,
Graduate School of lesign, llarvard
University, Septenber 1974, 25pp.
32
4. There is an inability to rapidly mani-
pulate large amounts of data for the rep-
Tesentation of alternative development
patterns,
5, Data sources generated by groups such
a5 transportation agencies, schools, hos-
pitals, etc., are rarely geographically
Congruént (making for non-uniform refer-
ence areas) and the agencies do not update
their data at sinilar intervals.
6. There is no accepted explicit methodo-
logy to test the differential relation-
ships of such systens either with respect
to themselves or to the ontire study are
7, Qualitative attributes such as esthe-
tics, ethics and value judgements are ext-
renely difficult to quantify.
In spite of these problems (which are
mostly applicable to space planning at an
urban and regional scale) there are analo-
gies that can be made to space allocation
Rethodologies used in the design of phys~
ical facilities (at an architectural
scale). Facilities design, or layout
design, is an extremely well researched
(Moore’1973, Willer 1970, Teicholz,
Stewart and’Lec, 1970) and developed area
by industrial engineers, operations
researchers and conputet scientists--as
well as by architects, Each user group
has examined the problem from its own
point of view and has made "process"
Gontributions in his own ai Economists,
for example, have for many years been
interested in the problen of the optinal
location of economic activities. susi-
nessnen want to optimize plant locations
and architects are interested in develo-
ping a conputer approach to physical space
and room Layout design.
In general, space allocation, whether for
lanners, urban designers or architects,
is-an ilf-structured problem. This means
that the problem cannot be described
mathenatically, that there does not exist
an algorithn for finding the "best fit”
Solution, or that the goal to be attained
Cannot be made explicit and quantifiable.
It is important to state that a program
can only augnent the human design process
and attempt to make explicit decisions
that are otherwise made implicitly within
this process. “The process of design,
according to D'Arcy Thompson!, is thé
resolution, at one instant in’timo, of
many forces that are governed by rates of
change. Design is thorofore iterative: and
the most successful allocation programs
are dynamic, interactive and graphic in
nature. In’addition, prograns are most
Successful when emphasis on man/machine
interaction is present rather than man or
machine action alone.In the case of computer augmented space
planning at an urban scale, successful
Pyograng such as Topaz? Pius “and
EWPIRIC! usually have at additional ability
to define the utilization of land in terms
of physical and geographical significance
(sueh’as land types including residential,
commercial, educational, industrial, open
space, etc.). A second’ requisite for be~
ing able to successfully use a computer
for urban space planning is that the urban
fabric must consist of an aggregation of
homogenous areas which are classified into
land use categories (and sub-categories),
each of which can be associated with an
environnental "archetype" (Albert and
Seaman, 1975).
11, _TDER-AND_OVER-CONSTRAINED PROGRAMS,
One way of classifying space planning pro-
blens ts according to whether’ the planning
situation is "under:" or Tover-cons~
trained." Design criteria are satisfied
by many possible alternatives and have
pany solutions. tnder-constrained
Situations will usually, therefore,
involve randon nunber generators to supply
aissing information: fesign alternatives”
fuse be evaluated using Nineuitive"
means usually by selection criteria not
explicit of Mot easily defined to a
machine.
In the over-constrained situation, which
is the more conmon one in design, 'no form
or physical solution can satisfy’all the
Criteria, In the case of planning, sone
of the criteria might be non-quantifiable,
This would be the case, for exanple, for
socio-economic patterns where it is’possi-
ble, at best, to assenble a statistical
model of the geographic area under consid~
eration. The generating mechanism for
this situation does a a search, often
heuristic, for the solution that best re-
laxes constraints to the point of least
Nftiction,"” As a system grows, events or
decisions become increasingly determinis-
tic end alternatives are diminished. In
over-constrained situations, conputer
prograns can present designers with a:wider
Tange of alternative problen solutions
and, in some cases, evaluate then accor-
Ging to specific etiteria. Because desipn
assumptions are frequently included on
the part of the planners, sensitivity
analyses are sonetines used to supply
the problen solver with information
Fegarding the degrees of latitude he has
with errors.
INPROVENENT AD CONSTRUCTION
UL
In order to further classify space alloca
tion programs and relevant conputer tech~
niques beyond the under and over-cons~
trained categories, we will return to the
analogy of facilities, or plant Layout,
333
design for analyzing ané allocating both
tual itative and quantitative data,” In the
Case of plant design Layout, for example,
worker safety and norgley efthetics and
Sther qualities must be evaluated just as
productian schedules, trip distances
Between departments, and other quantities
must be collected afd ueilized tte
design process.
[A second classification system for
Computer prograns used in facility design
are called inprovenent-type and construct
Hlon-type prograns.. ‘These categories are
independent of the under= and over-cons~
trained classification, Inprovement-type
or permutational algorithms. start with an
existing plan or layout ag input and
Aetenpe. to rearrange and iverace, through
alternative plans according to the optin~
Ulatlon of specific criteria. the addi
tive of Construction-type aligcation
propran, on the other herd, will attempt
Eorgenctate spaces or aroas into a tayout
or Plan according to sono generat criteria
but’ starting with an onpty site or floor
plans. The construction: type algorithn
TAIT usually not produce results. that are
entirely acceptiie to the designer. The
Solutions generated will, however, often
form the basis of additional work’ and per-
haps provide a wide range of altemative
designs and evaluations that will stim
Yate"the designer to produce new and
Improved solutions,
‘The improvenont-type program is perhaps
hore Connon in urban planning in’ that
conranhie locations wsualiy, except per~
fiape inthe case of mew communities, exist
for many Land use funetions. Even then
this"is the case, however, & proper objec-
ive or design approach must be explicitly
Hately “in the case of planing, this
could be transiated into the stazonent
that environmental data only becomes use~
ful as a basis of Isnd development. deci s-
ion making when the purpose oF nurposes
folwhich that land will be put sre known.
Iv. _Tecimrques
There are a variety of techniques used
Tor space allocation that are incorporated
into aigorithns for computer programs.
Host of the current, sophisticated pro-
frans are hybrid" in that nore than one
fechnique is’used ina single progran,
Briefly, space allocation prograns can be
neneraiized as-foliowss
1, Neighbor-Searching
This constructive technique will often
use a proximinity or relationship natrix
as input. Scores or Interactions are
calculated for each of the clenents and
the most active (interactive) element is
placed in the middle of the layout space.
Each successive activity unit is