<s>
Subsumption	B-Architecture
architecture	I-Architecture
is	O
a	O
reactive	O
robotic	O
architecture	O
heavily	O
associated	O
with	O
behavior-based	O
robotics	O
which	O
was	O
very	O
popular	O
in	O
the	O
1980s	O
and	O
90s	O
.	O
</s>
<s>
Subsumption	O
has	O
been	O
widely	O
influential	O
in	O
autonomous	O
robotics	O
and	O
elsewhere	O
in	O
real-time	B-General_Concept
AI	B-Application
.	O
</s>
<s>
Subsumption	B-Architecture
architecture	I-Architecture
is	O
a	O
control	O
architecture	O
that	O
was	O
proposed	O
in	O
opposition	O
to	O
traditional	O
AI	B-Application
,	O
or	O
GOFAI	B-General_Concept
.	O
</s>
<s>
Instead	O
of	O
guiding	O
behavior	O
by	O
symbolic	O
mental	O
representations	O
of	O
the	O
world	O
,	O
subsumption	B-Architecture
architecture	I-Architecture
couples	O
sensory	O
information	O
to	O
action	O
selection	O
in	O
an	O
intimate	O
and	O
bottom-up	O
fashion	O
.	O
</s>
<s>
Because	O
a	O
robot	O
must	O
have	O
the	O
ability	O
to	O
"	O
avoid	O
objects	O
"	O
in	O
order	O
to	O
"	O
wander	O
around	O
"	O
effectively	O
,	O
the	O
subsumption	B-Architecture
architecture	I-Architecture
creates	O
a	O
system	O
in	O
which	O
the	O
higher	O
layers	O
utilize	O
the	O
lower-level	O
competencies	O
.	O
</s>
<s>
Subsumption	B-Architecture
architecture	I-Architecture
attacks	O
the	O
problem	O
of	O
intelligence	O
from	O
a	O
significantly	O
different	O
perspective	O
than	O
traditional	O
AI	B-Application
.	O
</s>
<s>
Instead	O
of	O
modelling	O
aspects	O
of	O
human	O
intelligence	O
via	O
symbol	O
manipulation	O
,	O
this	O
approach	O
is	O
aimed	O
at	O
real-time	B-General_Concept
interaction	O
and	O
viable	O
responses	O
to	O
a	O
dynamic	O
lab	O
or	O
office	O
environment	O
.	O
</s>
<s>
Situatedness	O
–	O
A	O
major	O
idea	O
of	O
situated	B-General_Concept
AI	I-General_Concept
is	O
that	O
a	O
robot	O
should	O
be	O
able	O
to	O
react	O
to	O
its	O
environment	O
within	O
a	O
human-like	O
time-frame	O
.	O
</s>
<s>
Brooks	O
argues	O
that	O
situated	B-General_Concept
mobile	O
robot	O
should	O
not	O
represent	O
the	O
world	O
via	O
an	O
internal	O
set	O
of	O
symbols	O
and	O
then	O
act	O
on	O
this	O
model	O
.	O
</s>
<s>
Yet	O
,	O
each	O
module/behavior	O
still	O
models	O
the	O
world	O
,	O
but	O
on	O
a	O
very	O
low	O
level	O
,	O
close	O
to	O
the	O
sensorimotor	O
signals	O
.	O
</s>
<s>
Embodiment	O
–	O
Brooks	O
argues	O
building	O
an	O
embodied	B-General_Concept
agent	I-General_Concept
accomplishes	O
two	O
things	O
.	O
</s>
<s>
The	O
second	O
is	O
that	O
it	O
can	O
solve	O
the	O
symbol	O
grounding	O
problem	O
,	O
a	O
philosophical	O
issue	O
many	O
traditional	O
AIs	B-Application
encounter	O
,	O
by	O
directly	O
coupling	O
sense-data	O
to	O
meaningful	O
actions	O
.	O
</s>
<s>
Also	O
,	O
by	O
rejecting	O
top-down	O
representations	O
as	O
a	O
viable	O
starting	O
point	O
for	O
AI	B-Application
,	O
it	O
seems	O
that	O
"	O
intelligence	O
is	O
determined	O
by	O
the	O
dynamics	O
of	O
interaction	O
with	O
the	O
world.	O
"	O
</s>
<s>
Emergence	O
–	O
Conventionally	O
,	O
individual	O
modules	B-Architecture
are	O
not	O
considered	O
intelligent	O
by	O
themselves	O
.	O
</s>
<s>
It	O
is	O
the	O
interaction	O
of	O
such	O
modules	B-Architecture
,	O
evaluated	O
by	O
observing	O
the	O
agent	O
and	O
its	O
environment	O
,	O
that	O
is	O
usually	O
deemed	O
intelligent	O
(	O
or	O
not	O
)	O
.	O
</s>
<s>
The	O
ideas	O
outlined	O
above	O
are	O
still	O
a	O
part	O
of	O
an	O
ongoing	O
debate	O
regarding	O
the	O
nature	O
of	O
intelligence	O
and	O
how	O
the	O
progress	O
of	O
robotics	O
and	O
AI	B-Application
should	O
be	O
fostered	O
.	O
</s>
<s>
Each	O
layer	O
is	O
made	O
up	O
by	O
a	O
set	O
of	O
processors	O
that	O
are	O
augmented	O
finite-state	B-Architecture
machines	I-Architecture
(	O
AFSM	O
)	O
,	O
the	O
augmentation	O
being	O
added	O
instance	B-Application
variables	I-Application
to	O
hold	O
programmable	O
data-structures	O
.	O
</s>
<s>
A	O
layer	O
is	O
a	O
module	B-Architecture
and	O
is	O
responsible	O
for	O
a	O
single	O
behavioral	O
goal	O
,	O
such	O
as	O
"	O
wander	O
around.	O
"	O
</s>
<s>
There	O
is	O
no	O
central	O
control	O
within	O
or	O
between	O
these	O
behavioral	O
modules	B-Architecture
.	O
</s>
<s>
After	O
testing	O
and	O
debugging	O
the	O
combined	O
behavior	O
,	O
this	O
process	O
can	O
be	O
repeated	O
for	O
(	O
theoretically	O
)	O
any	O
number	O
of	O
behavioral	O
modules	B-Architecture
.	O
</s>
<s>
The	O
following	O
is	O
a	O
small	O
list	O
of	O
robots	O
that	O
utilize	O
the	O
subsumption	B-Architecture
architecture	I-Architecture
.	O
</s>
<s>
the	O
emphasis	O
on	O
iterative	O
development	O
and	O
testing	O
of	O
real-time	B-General_Concept
systems	I-General_Concept
in	O
their	O
target	O
domain	O
;	O
</s>
<s>
the	O
lack	O
of	O
large	O
memory	O
and	O
symbolic	O
representation	O
,	O
which	O
seems	O
to	O
restrict	O
the	O
architecture	O
from	O
understanding	B-General_Concept
language	I-General_Concept
;	O
</s>
<s>
When	O
subsumption	B-Architecture
architecture	I-Architecture
was	O
developed	O
,	O
the	O
novel	O
setup	O
and	O
approach	O
of	O
subsumption	B-Architecture
architecture	I-Architecture
allowed	O
it	O
to	O
be	O
successful	O
in	O
many	O
important	O
domains	O
where	O
traditional	O
AI	B-Application
had	O
failed	O
,	O
namely	O
real-time	B-General_Concept
interaction	O
with	O
a	O
dynamic	O
environment	O
.	O
</s>
<s>
The	O
lack	O
of	O
large	O
memory	O
storage	O
,	O
symbolic	O
representations	O
,	O
and	O
central	O
control	O
,	O
however	O
,	O
places	O
it	O
at	O
a	O
disadvantage	O
at	O
learning	O
complex	O
actions	O
,	O
in-depth	O
mapping	O
,	O
and	O
understanding	B-General_Concept
language	I-General_Concept
.	O
</s>
