1, 2
1
2
2
8
Emergence of a garbage-collection behavior by multi-robots with
a nonlinear sensory-motor mapping
Kou Iwata1 Yasushi Honda2
1 Division of Information and Electronic Engineering, Muroran Institute of Technology2 College of Information and Systems, Muroran Institute of Technology
Abstract
We developed a robot with a sensory-motor mapping using ultra-sonic sensors. A nonlinear
function is used as the mapping. It is found that a combination of two kinds of hyperbolic
function enables the robot to avoid obstacles like a wall or garbage. We found requirements
in both of body structure and sensory-motor mapping for emergence of a garbage collection.
In the conducted experiments with eight robots, the positional information of the object is
obtained using the motion capture and is compared with initial position of the objects. It
is found that the robots emerge cooperative behavior that is a garbage-collection behavior
avoiding wall and other robots.
1
” ” [1]
” ”
2
Swiss Robot Maris M
Pfeifer [2][3]
[4] 1
2
2
1
[2][3] [4]
4 2 2
3 5 1 8
1m/s 0.2m/s
20 5 20
2.3m, 1 1.8m 2m
2.6m
1:
21
EV3(OS-Linux)
•
•
2
2
1: 2
2.1
1
2
(1) (6) (1)(4)
!10 < r < 10
(2)(5)
2
d (3)(6)
2 3
6 3
3
[1]
2 4
2m
rL, rR [%]
sL, sR [mm]
b1, b2
(300,250)
g1, g2 (0.02,10000)
c1, c2 (20,10)
d (20)
2:
rL1 = c1 · tanh(g1 · (sR ! b1)) (1)
rL2 = c2 · tanh(g2 · (sR ! b2)) (2)
rL = rL1 + rL2 + d (3)
rR1 = c1 · tanh(g1 · (sL ! b1)) (4)
rR2 = c2 · tanh(g2 · (sL ! b2)) (5)
rR = rR1 + rR2 + d (6)
2:
2
-40
-20
0
20
40
60
80
0 100 200 300 400 500
Mot
or-O
utpu
t[%]
USsensor-Value[mm]
sensory-motor mapping
rL1, rR1rL2, rR2rL, rR
3:
4:
2.2
4
•
•
•
•
EV3
1cm 50cm
50 5
5
40mm
80mm
80mm
40mm 2
5:
38
4 ( 35mm
25mm 80mm) 5 5 25
625mm
8
CLUSTERSSH
8 1
20
6 8
3.1 30 120mm
6 20 3,4
30
20 30
-2000
-1000
0
1000
2000
-2000-1000 0 1000 2000
mm
mm
1 minute later
-2000
-1000
0
1000
2000
-2000-1000 0 1000 2000
mm
mm
5 minutes later
-2000
-1000
0
1000
2000
-2000-1000 0 1000 2000
mm
mm
10 minutes later
-2000
-1000
0
1000
2000
-2000-1000 0 1000 2000
mm
mm
20 minutes later
6: 30 120mm(1 5 10 20 )
3
3.2 45 120mm
45
120mm
7
2
-2000
-1000
0
1000
2000
-2000-1000 0 1000 2000
mm
mm
1 minute later
-2000
-1000
0
1000
2000
-2000-1000 0 1000 2000
mm
mm
5 minutes later
-2000
-1000
0
1000
2000
-2000-1000 0 1000 2000
mm
mm
10 minutes later
-2000
-1000
0
1000
2000
-2000-1000 0 1000 2000
mm
mm
20 minutes later
7: 45 120mm(1 5 10 20 )
3.3 45 170mm
1 2 4
7 8 20
-2000
-1000
0
1000
2000
-2000-1000 0 1000 2000
mm
mm
1 minute later
-2000
-1000
0
1000
2000
-2000-1000 0 1000 2000
mm
mm
5 minutes later
-2000
-1000
0
1000
2000
-2000-1000 0 1000 2000
mm
mm
10 minutes later
-2000
-1000
0
1000
2000
-2000-1000 0 1000 2000
mm
mm
20 minutes later
8: 45 170mm(1 5 10 20 )
4
45
120mm
[2][3] 1/5
20
5
2m 8
[1] , : (2006).
[2] Maris, M. and Boekhorst, R., Exploiting phys-ical constraints: heap formation through behav-ioral error in a group of robots , Proceedings of1996 IEEE/RSJ International Conference on Intel-ligent Robots and Systems (1996), pp. 1655-1660.
[3] Pfeifer, R. and Schemer, C., UnderstandingIn-telligence , The MIT Press(1999).
[4]
Vol. 79, No. 800, pp. 1046-1055, 2013.
4