Page 196 - Mechatronics for Safety, Security and Dependability in a New Era

Ch37-I044963.fm Page 180 Tuesday, August 1, 2006 3:12 PM
Ch37-I044963.fm
180
180 Page 180 Tuesday, August 1, 2006 3:12 PM
Different from the previous method, robots cannot recognize obstacles from the fact that human does
not move towards the goal. In this case, robots cannot recognize obstacles correctly. Then we make
a strategy for this method. During human moves straight, it must be the safest way to follow human's
behind. When human turns, there is high possibility that there exists obstacle at the corner. So
robot should bypath the corner. With this strategy, we adopt "local potential map" which describes
possibility of obstacle existence locally around robots.
Local potential map is generated as shown in Figure 4. Robots modify the map by adding this
potential according to its motion. Robots decide their motion to lower the sum of the potential.
Figure 5 and Figure 6 show the result of an experiment. The potential value at the corner is higher
and the robot moves to bypath the corner. Finally the robot has large error in motion direction, but it
correctly generates following path since it depends only on local information.

Potential
Mgh
3
Motion Direction
detect
.ow
Figure 4: Obstacle potential •

1 62

Obstacle
Obstacle

/ Start:
Start :
^ji Goal:
Goal :
Human:
Human :
Robot : —
Robot:
8-4
Figure 5: Experiment result Figure 6: Local potential map
CONCLUSION
We propose two methods that robots recognize obstacles by observation of human motion when they
cooperatively transport an object. Each experiment uses just one robot, but the idea is expandable to
multiple robot transportation. Further, there must exist other applications that use the human ability
of sensing and global path planning. Power assist system may be another type of the application.

REFERENCES
Hirata Y. et al. (2002). Motion control of multiple DR helpers transporting a single object in
cooperation with a human based on map information. Proc. IEEE International Conf. on Robotics
and Automation, 995-1000.
Takubo T. et al. (2001). Human-robot cooperative handling using virtual nonholonomic constraint in
3-D space. Proc. IEEE International Conf. on Robotics and Automation, 2680-2685.