Information Technology Reference
In-Depth Information
20
mars rover
depots
restaurant
blocks
o
ce
mars rover
depots
restaurant
blocks
o
ce
mars rover
depots
restaurant
blocks
o
ce
0
.
15
15
400
0
.
1
10
200
5
·
10
−
2
5
0
0
0
0
.
5
1
0
.
5
1
0
.
5
1
removed facts
removed facts
removed facts
(a) Average number of planning and
execution phases.
(c) Average CPU time of a single plan-
ning phase.
Fig. 6.
System behavior for a decreasing amount of initial knowledge about dynamic aspects of
the domain
(b) Average CPU time of the over-
all planning and reasoning process.
The experiments where conducted on a 64-bit Intel Core 2 Quad Q9400 with 4 GB
memory.
Results.
ACogControl was able to correctly perform the given task for all domains
and all runs—even in situations where all facts were removed from the domain model
of the agent. The average number of necessary planning and execution phases is show
in Fig. 6(a). The average number of planning and execution phases increases with a
decreasing number of initial information, since the agent needs to stop the planning
process and execute knowledge acquisition activities more often. We also expected the
overall CPU time of the reasoning and planning process to increase for all domains
with a decreasing amount of initial knowledge. However, Fig. 6(b) shows that this is
only true for the rover, the office and the restaurant domain. The blocks and the depots
domain show a different behavior. For these domains the overall CPU time increases
until 60 respectively 80 percent of the facts are removed from the domain model of
the agent and then decreases until all facts are removed. The results shown in Fig. 6(c)
might give an explanation for this. They show that the average time for a planning phase
decreases with a decreasing amount of information that initially is available for the
agent. Together with the results shown in Fig. 6(a) these results indicate that the more
planning phases are performed the shorter are the individual phases. Thus, the proposed
continual planning system, so to speak, partitions the overall planning problem into a
set of simpler planning problems. Moreover, the depots and the blocks world domain
indicate that the sum of the individual planning phases can be lower even if the number
of planning phases is higher as shown by Fig. 6(b).
5
Related Work
Most of the previous approaches that are able to generate plans in partially known envi-
ronments generate conditional plans—or policies—for all possible contingencies. This
includes
conformant
,
contingent
or
probabilistic
planning approaches [13,4]. Several
planning approaches that generate conditional plans, including [1,3,6,7], use
runtime
variables
for the purpose of representing unknown information. Runtime variables can
