<s>
Incremental	B-Algorithm
heuristic	I-Algorithm
search	I-Algorithm
algorithms	O
combine	O
both	O
incremental	O
and	O
heuristic	O
search	O
to	O
speed	O
up	O
searches	O
of	O
sequences	O
of	O
similar	O
search	O
problems	O
,	O
which	O
is	O
important	O
in	O
domains	O
that	O
are	O
only	O
incompletely	O
known	O
or	O
change	O
dynamically	O
.	O
</s>
<s>
Heuristic	O
search	O
algorithms	O
,	O
often	O
based	O
on	O
A*	B-Protocol
,	O
use	O
heuristic	O
knowledge	O
in	O
the	O
form	O
of	O
approximations	O
of	O
the	O
goal	O
distances	O
to	O
focus	O
the	O
search	O
and	O
solve	O
search	O
problems	O
potentially	O
much	O
faster	O
than	O
uninformed	O
search	O
algorithms	O
.	O
</s>
<s>
The	O
resulting	O
search	O
problems	O
,	O
sometimes	O
called	O
dynamic	O
path	O
planning	O
problems	O
,	O
are	O
graph	O
search	O
problems	O
where	O
paths	O
have	O
to	O
be	O
found	O
repeatedly	O
because	O
the	O
topology	B-Architecture
of	O
the	O
graph	O
,	O
its	O
edge	O
costs	O
,	O
the	O
start	O
vertex	O
or	O
the	O
goal	O
vertices	O
change	O
over	O
time	O
.	O
</s>
<s>
So	O
far	O
,	O
three	O
main	O
classes	O
of	O
incremental	B-Algorithm
heuristic	I-Algorithm
search	I-Algorithm
algorithms	O
have	O
been	O
developed	O
:	O
</s>
<s>
The	O
first	O
class	O
restarts	O
A*	B-Protocol
at	O
the	O
point	O
where	O
its	O
current	O
search	O
deviates	O
from	O
the	O
previous	O
one	O
(	O
example	O
:	O
Fringe	O
Saving	O
A*	B-Protocol
)	O
.	O
</s>
<s>
approximate	O
distance	O
to	O
goal	O
)	O
from	O
the	O
previous	O
search	O
during	O
the	O
current	O
search	O
to	O
make	O
them	O
more	O
informed	O
(	O
example	O
:	O
Generalized	O
Adaptive	O
A*	B-Protocol
)	O
.	O
</s>
<s>
The	O
third	O
class	O
updates	O
the	O
g-values	O
(	O
distance	O
from	O
start	O
)	O
from	O
the	O
previous	O
search	O
during	O
the	O
current	O
search	O
to	O
correct	O
them	O
when	O
necessary	O
,	O
which	O
can	O
be	O
interpreted	O
as	O
transforming	O
the	O
A*	B-Protocol
search	I-Protocol
tree	O
from	O
the	O
previous	O
search	O
into	O
the	O
A*	B-Protocol
search	I-Protocol
tree	O
for	O
the	O
current	O
search	O
(	O
examples	O
:	O
Lifelong	B-Algorithm
Planning	I-Algorithm
A*	I-Algorithm
,	O
D*	B-Algorithm
,	O
D*	B-Algorithm
Lite	O
)	O
.	O
</s>
<s>
All	O
three	O
classes	O
of	O
incremental	B-Algorithm
heuristic	I-Algorithm
search	I-Algorithm
algorithms	O
are	O
different	O
from	O
other	O
replanning	O
algorithms	O
,	O
such	O
as	O
planning	O
by	O
analogy	O
,	O
in	O
that	O
their	O
plan	O
quality	O
does	O
not	O
deteriorate	O
with	O
the	O
number	O
of	O
replanning	O
episodes	O
.	O
</s>
<s>
earlier	O
systems	O
)	O
or	O
D*	B-Algorithm
Lite	O
(	O
current	O
systems	O
)	O
,	O
two	O
different	O
incremental	B-Algorithm
heuristic	I-Algorithm
search	I-Algorithm
algorithms	O
.	O
</s>
