Improving Area Center Robot Navigation Using a Novel Range Scan Segmentation Method - Foundations on Natural and Artificial Computation - page 239

1

3

4

2

5
6
7
8

15

14

9

13

12

10

11

Fig. 3. Pictures sequence showing the application of the estimation and merging proce-
dure described in section 5. World is composed by plain walls and one hedge where the
noise is 10 times higher, world dimensions are m. Red(black) circle in a)

represents robot and the arrow at robot center indicates orientation ϕ =0 . A picture
of the ideal world used to generate the measures is shown for comparison purposes.

17 . 5 m

× 12 . 5

parameters of regressions over different sets of data. In fig. 3 an example of the
whole process is shown.
To analyze the performance on a computer a preliminary C++ implementa-
tion of the procedure, yet in development, was tested on laptop with a Core Duo
2 processor at 2.28 Ghz. New robots are being shipped with similar processors
and several well known old models can be now upgraded. For 181 measures with
noise standard deviation σ m. the procedure takes about 2.5

milliseconds, for 360 about 6 ms. and for 720 about 9.5 ms., these times suggest
a complexity near linear order. For σ

=0 . 002 , 2 cm. at 10

.
Taken into account that an usual laser sampling period takes 100 ms or longer,
proposed procedure execution only takes a bit of this time.

=0 . 02 , times increase about 25 − 30%

6 Improving Area Center Navigation

The basic area center method for robot navigation was proposed previously in
[1,2,13]. Basically the robot follows the center of area of its perceived free area
while it is accessible. Robot velocity vector is proportional to distance of area
center in robot local coordinates. A basic concept in area center navigation
is the split points, where the perceived free area is split in two sectors when
area center becomes inaccessible and one of them is selected to follow its area