<s>
Iterative	B-Protocol
deepening	I-Protocol
A*	I-Protocol
(	O
IDA*	B-Protocol
)	O
is	O
a	O
graph	B-Application
traversal	O
and	O
path	B-Algorithm
search	I-Algorithm
algorithm	O
that	O
can	O
find	O
the	O
shortest	O
path	O
between	O
a	O
designated	O
start	O
node	O
and	O
any	O
member	O
of	O
a	O
set	O
of	O
goal	O
nodes	O
in	O
a	O
weighted	O
graph	B-Application
.	O
</s>
<s>
It	O
is	O
a	O
variant	O
of	O
iterative	B-Algorithm
deepening	I-Algorithm
depth-first	I-Algorithm
search	I-Algorithm
that	O
borrows	O
the	O
idea	O
to	O
use	O
a	O
heuristic	O
function	O
to	O
evaluate	O
the	O
remaining	O
cost	O
to	O
get	O
to	O
the	O
goal	O
from	O
the	O
A*	B-Protocol
search	I-Protocol
algorithm	I-Protocol
.	O
</s>
<s>
Since	O
it	O
is	O
a	O
depth-first	O
search	B-Application
algorithm	I-Application
,	O
its	O
memory	O
usage	O
is	O
lower	O
than	O
in	O
A*	O
,	O
but	O
unlike	O
ordinary	O
iterative	B-Algorithm
deepening	I-Algorithm
search	I-Algorithm
,	O
it	O
concentrates	O
on	O
exploring	O
the	O
most	O
promising	O
nodes	O
and	O
thus	O
does	O
not	O
go	O
to	O
the	O
same	O
depth	O
everywhere	O
in	O
the	O
search	O
tree	B-Application
.	O
</s>
<s>
Unlike	O
A*	O
,	O
IDA*	B-Protocol
does	O
not	O
utilize	O
dynamic	B-Algorithm
programming	I-Algorithm
and	O
therefore	O
often	O
ends	O
up	O
exploring	O
the	O
same	O
nodes	O
many	O
times	O
.	O
</s>
<s>
While	O
the	O
standard	O
iterative	B-Algorithm
deepening	I-Algorithm
depth-first	I-Algorithm
search	I-Algorithm
uses	O
search	O
depth	O
as	O
the	O
cutoff	O
for	O
each	O
iteration	O
,	O
the	O
IDA*	B-Protocol
uses	O
the	O
more	O
informative	O
,	O
where	O
is	O
the	O
cost	O
to	O
travel	O
from	O
the	O
root	O
to	O
node	O
and	O
is	O
a	O
problem-specific	O
heuristic	O
estimate	O
of	O
the	O
cost	O
to	O
travel	O
from	O
to	O
the	O
goal	O
.	O
</s>
<s>
IDA*	B-Protocol
is	O
beneficial	O
when	O
the	O
problem	O
is	O
memory	O
constrained	O
.	O
</s>
<s>
A*	B-Protocol
search	I-Protocol
keeps	O
a	O
large	O
queue	O
of	O
unexplored	O
nodes	O
that	O
can	O
quickly	O
fill	O
up	O
memory	O
.	O
</s>
<s>
By	O
contrast	O
,	O
because	O
IDA*	B-Protocol
does	O
not	O
remember	O
any	O
node	O
except	O
the	O
ones	O
on	O
the	O
current	O
path	O
,	O
it	O
requires	O
an	O
amount	O
of	O
memory	O
that	O
is	O
only	O
linear	O
in	O
the	O
length	O
of	O
the	O
solution	O
that	O
it	O
constructs	O
.	O
</s>
<s>
for	O
all	O
nodes	O
and	O
all	O
neighbors	O
of	O
;	O
they	O
conclude	O
that	O
compared	O
to	O
a	O
brute-force	O
tree	B-Application
search	O
over	O
an	O
exponential-sized	O
problem	O
,	O
IDA*	B-Protocol
achieves	O
a	O
smaller	O
search	O
depth	O
(	O
by	O
a	O
constant	O
factor	O
)	O
,	O
but	O
not	O
a	O
smaller	O
branching	O
factor	O
.	O
</s>
<s>
Recursive	O
best-first	O
search	O
is	O
another	O
memory-constrained	O
version	O
of	O
A*	B-Protocol
search	I-Protocol
that	O
can	O
be	O
faster	O
in	O
practice	O
than	O
IDA*	B-Protocol
,	O
since	O
it	O
requires	O
less	O
regenerating	O
of	O
nodes	O
.	O
</s>
<s>
Applications	O
of	O
IDA*	B-Protocol
are	O
found	O
in	O
such	O
problems	O
as	O
planning	B-Application
.	O
</s>
