Biomedical Engineering Reference
In-Depth Information
Table 3. Frequency of failure of simulated robot performance considering all experiments from different
starting positions
Coordination
approach
% Causes
of failure
Unsuccessful
code
Causes of
failures
(Freq.)
% All
failed
trials
6.7%
14%
Layered
(ruled-based)
1
2
0%
0%
2
0
23.3%
50%
3
7
16.7%
36%
4
5
AIS
1
0
0%
0%
100%
2
10
33.3%
3
0
0%
0%
4
0
0%
0 %
Table 4. Summary of statistics for time (CPU cycles) to reach the goal using the simulator considering
only successful cases (goal reached)
Mean
Standard
deviation
Co-
ordination
approach
95% confidence interval
P2
P0
P1
P2
P0
P1
P2
P0
P1
Layered
185.43
27.6
15.1
[212.01;287.3]
[159.8; 211]
233.7
249.7
21.14
[200.1;267.2]
AIS
98.9 1
19.8 6
9.6
56.4 4
1.4
13.38
[58,5;139.3] [
90.2;149.4]
[60;79.2]
problem is difficult because the dimensions of the
environment requires a high degree of precision
to steer towards the centre of the small gate and
the robot needs to be able to reach the goal in a
safe way.
The following Tables 3 and 4 summarize
experimental results. In them, it was found that
the performance of the behaviour coordination
depends on the correct choice of the following
parameters:
•
B
: the simple behaviours implemented to
emerge a more complex behaviour
A trial and error procedure was used to deter-
mine these parameters in order to obtain accept-
able results in previous experiments, because there
is not a straightforward method for determining
them. Although adequate parameter selection was
not relevant to this research it is an important is-
sue in mobile robotics. In Krautmacher & Dilger
(2004) found that AIS code was dependent on the
choice of free parameter used.
Table 4 shows that the performance rate was
good over all experiments with AIS. In average,
the robot takes 96 CPU cycles to approach to the
target, failing in none of 30 trials. The worst per-
formance rate obtained for layered coordination
may be found in the following fact. Layered co-
•
D
: the distance between the robot an ob-
stacles (when it is small, there are more
frequent collisions with walls and obstacles
and then frequent trapped situations tend to
appear);
•
S
: the scope of sensor readings, mainly those
related with obstacle avoidance;
Search WWH ::
Custom Search