Adaptive Human Aware Navigation Based on Motion Pattern ... - VBN
Adaptive Human Aware Navigation Based on Motion Pattern ... - VBN
Adaptive Human Aware Navigation Based on Motion Pattern ... - VBN
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
is reas<strong>on</strong>able as it gets more difficult to assess human interest<br />
at l<strong>on</strong>g distance.<br />
In all three figures, the vectors in the red color range<br />
(high PI) are dominant when the directi<strong>on</strong> of the pers<strong>on</strong> is<br />
towards the robot, while there is an overweight of vectors not<br />
pointing directly towards the robot in the blue color range<br />
(low PI). This reflects that a pers<strong>on</strong> seeking interacti<strong>on</strong> has<br />
the trajectory moving towards the robot.<br />
Fig. 7 shows the development of PI over time when <strong>on</strong>e<br />
test pers<strong>on</strong> changes behavior. It can be seen how maximum<br />
and minimum values for PI increases as more test pers<strong>on</strong>s<br />
have been evaluated. After evaluating <strong>on</strong>e test pers<strong>on</strong>, the<br />
robot has gathered very little interacti<strong>on</strong> experience, and<br />
thereby has difficulties in determining the corresp<strong>on</strong>dence<br />
between moti<strong>on</strong> pattern and end result - hence PI stays<br />
close to 0.5. After the third test pers<strong>on</strong> has been evaluated,<br />
the robot now has gathered more cases and therefore has<br />
improved estimating the outcome of the behavior. For the<br />
last test pers<strong>on</strong>, the robot is clearly capable of determining<br />
what will be the outcome of the encounter.<br />
Fig. 9 shows the development of PI over time. It can be<br />
seen that PI changes more for the fr<strong>on</strong>tal area and small<br />
area than for all other cases. This is because most cases<br />
will be close to 0.5 at large distances, which affects the<br />
mean result when looking at all cases. Furthermore, most<br />
encounters goes through the fr<strong>on</strong>tal area thereby having the<br />
highest number of updates of PI. Fig. 9 illustrates that the<br />
database quickly starts to adapt to the new envir<strong>on</strong>ment,<br />
when the test pers<strong>on</strong> changes behavior to no interacti<strong>on</strong> after<br />
the first 18 encounters.<br />
By coupling the CBR system with navigati<strong>on</strong>, the result<br />
is an adaptive robot behavior respecting the pers<strong>on</strong>al z<strong>on</strong>es<br />
depending <strong>on</strong> the pers<strong>on</strong>’s willingness to interact - a step<br />
forward from previous studies [15].<br />
VI. CONCLUSION<br />
In this paper, we have described an adaptive system for<br />
natural interacti<strong>on</strong> between mobile robots and humans. The<br />
system forms a basis for human aware robot navigati<strong>on</strong><br />
respecting the pers<strong>on</strong>’s social spaces.<br />
Validati<strong>on</strong> of the system has been c<strong>on</strong>ducted through two<br />
experiments in a real world setting. The first test shows that<br />
the Case <str<strong>on</strong>g>Based</str<strong>on</strong>g> Reas<strong>on</strong>ing (CBR) system gradually learns<br />
from interacti<strong>on</strong> experience. The experiment also shows how<br />
moti<strong>on</strong> patterns from different people can be stored and<br />
generalized in order to predict the outcome of a new humanrobot<br />
encounter.<br />
Sec<strong>on</strong>d experiment shows how the estimated outcome of<br />
the interacti<strong>on</strong> adapts to changes of the behavior of a test<br />
pers<strong>on</strong>. It is illustrated how the same moti<strong>on</strong> pattern can be<br />
interpreted differently after a period of training.<br />
An interesting prospect for future work is elaborati<strong>on</strong>s<br />
of the CBR system, e.g. doing experiments with a variable<br />
learning rate and additi<strong>on</strong>al features in the database.<br />
The presented system is a step forward in creating social<br />
intelligent robots, capable of navigating in an everyday envir<strong>on</strong>ment<br />
and interacting with human-beings by understanding<br />
their interest and intenti<strong>on</strong>. In a l<strong>on</strong>g perspective, the results<br />
could be applied in service or assistive robots in e.g. health<br />
care systems.<br />
REFERENCES<br />
[1] H. J. Andersen, T. Bak, and M. Svenstrup. <str<strong>on</strong>g>Adaptive</str<strong>on</strong>g> robot to pers<strong>on</strong><br />
encounter : by moti<strong>on</strong> patterns. In Proceedings of the Internati<strong>on</strong>al<br />
C<strong>on</strong>ference <strong>on</strong> Research and Educati<strong>on</strong> in Robotics 2008, Lecture<br />
Notes in Computer Science, pages 13–23. Springer-Verlag GmbH,<br />
2008.<br />
[2] A. Bruce, I. Nourbakhsh, and R. Simm<strong>on</strong>s. The role of expressiveness<br />
and attenti<strong>on</strong> in human-robot interacti<strong>on</strong>. In In Proceedings, AAAI Fall<br />
Symposium, 2001.<br />
[3] H. I. Christensen and E. Pacchierotti. Embodied social interacti<strong>on</strong> for<br />
robots. In K. Dautenhahn, editor, AISB-05, pages 40–45, Hertsfordshire,<br />
April 2005.<br />
[4] T. Collett, B. A. MacD<strong>on</strong>ald, and B. P. Gerkey. Player 2.0: Toward<br />
a practical robot programming framework. In C. Sammut,<br />
editor, In Proceedings of the Australasian C<strong>on</strong>ference <strong>on</strong> Robotics<br />
and Automati<strong>on</strong> (ACRA 2005), Sydney, Australia, December 2005.<br />
http://playerstage.sourceforge.net.<br />
[5] K. Dautenhahn. Methodology & themes of human-robot interacti<strong>on</strong>:<br />
A growing research field. Internati<strong>on</strong>al Journal of Advanced Robotic<br />
Systems, 4(1):103–108, 2007.<br />
[6] K. Dautenhahn, M. Walters, S. Woods, K. L. Koay, E. A. Sisbot,<br />
R. Alami, and T. Simé<strong>on</strong>. How may i serve you? a robot compani<strong>on</strong><br />
approaching a seated pers<strong>on</strong> in a helping c<strong>on</strong>text. In HRI <str<strong>on</strong>g>Human</str<strong>on</strong>g><br />
Robot Interacti<strong>on</strong> ’06 (HRI06), Salt Lake City, Utah, USA, 2006.<br />
[7] B. Fajen and W. Warren. The behavioral dynamics of steering, obstacle<br />
avoidance, and route selecti<strong>on</strong>. Journal of Experimental Psychology:<br />
<str<strong>on</strong>g>Human</str<strong>on</strong>g> Percepti<strong>on</strong> and Performance, 29(1):343–362, 2003.<br />
[8] B. Gates. A robot in every home. Scientific American, 296(1):58–65,<br />
2007.<br />
[9] D. A. V. Govea. Incremental Learning for Moti<strong>on</strong> Predicti<strong>on</strong> of<br />
Pedestrians and Vehicles. PhD thesis, Institut Nati<strong>on</strong>al Polytechnique<br />
de Grenoble, 2007.<br />
[10] E. Hall. The Hidden Dimensi<strong>on</strong>. Doubleday, 1966.<br />
[11] N. Hanajima, Y. Ohta, H. Hikita, and M. Yamashita. Investigati<strong>on</strong><br />
of impressi<strong>on</strong>s for approach moti<strong>on</strong> of a mobile robot based <strong>on</strong><br />
psychophysiological analysis. In IEEE Internati<strong>on</strong>al Workshop <strong>on</strong><br />
Robots and <str<strong>on</strong>g>Human</str<strong>on</strong>g> Interactive Communicati<strong>on</strong> ROMAN 2005, pages<br />
79–84, August 2005.<br />
[12] I. Jurisica and J. Glasgow. Applying case-based reas<strong>on</strong>ing to c<strong>on</strong>trol<br />
in robotics. In 3rd Robotics and Knowledge-<str<strong>on</strong>g>Based</str<strong>on</strong>g> Systems Workshop,<br />
St. Hubert Quebec, 1995.<br />
[13] J. Kolodner. Case-based reas<strong>on</strong>ing. Morgan Kaufmann Publishers<br />
Inc., San Francisco, CA, USA, 1993.<br />
[14] M. Michalowski, S. Sabanovic, C. DiSalvo, D. Busquets, L. Hiatt,<br />
N. Melchior, and R. Simm<strong>on</strong>s. Socially distributed percepti<strong>on</strong>: Grace<br />
plays social tag at aaai 2005. Aut<strong>on</strong>omous Robots, 22(4):385–397,<br />
May 2007.<br />
[15] E. Sisbot, A. Clodic, L. Marin U., M. F<strong>on</strong>tmarty, L. Brethes, and<br />
R. Alami. Implementing a human-aware robot system. In The<br />
15th IEEE Internati<strong>on</strong>al Symposium <strong>on</strong> Robot and <str<strong>on</strong>g>Human</str<strong>on</strong>g> Interactive<br />
Communicati<strong>on</strong>, 2006. ROMAN 2006., pages 727–732, 6-8 Sept. 2006.<br />
[16] M. Svenstrup, S. T. Hansen, H. J. Andersen, and T. Bak. Pose<br />
estimati<strong>on</strong> and adaptive robot behaviour for human-robot interacti<strong>on</strong>.<br />
In Proceedings of the 2009 IEEE Internati<strong>on</strong>al C<strong>on</strong>ference <strong>on</strong> Robotics<br />
and Automati<strong>on</strong>. ICRA 2009., pages –, Kobe, Japan, May 2009.<br />
[17] M. L. Walters, K. Dautenhahn, K. L. Koay, C. Kaouri, R. te Boekhorst,<br />
C. Nehaniv, I. Werry, and D. Lee. Close encounters: Spatial distances<br />
between people and a robot of mechanistic appearance. In Proceedings<br />
of 2005 5th IEEE-RAS Internati<strong>on</strong>al C<strong>on</strong>ference <strong>on</strong> <str<strong>on</strong>g>Human</str<strong>on</strong>g>oid Robots,<br />
pages 450–455, Tsukuba, Japan, December 2005.<br />
[18] S. Woods, M. L. Walters, K. Koay, and K. Dautenhahn. Methodological<br />
issues in hri: A comparis<strong>on</strong> of live and video-based methods in<br />
robot to human approach directi<strong>on</strong> trials. In The 15th IEEE Internati<strong>on</strong>al<br />
Symposium <strong>on</strong> Robot and <str<strong>on</strong>g>Human</str<strong>on</strong>g> Interactive Communicati<strong>on</strong><br />
(RO-MAN06), pages 51–58, Hatfield, UK, September 2006.<br />
[19] J. Xavier, M. Pacheco, D. Castro, A. Ruano, and U. Nunes. Fast line,<br />
arc/circle and leg detecti<strong>on</strong> from laser scan data in a player driver. In<br />
Proceedings of the 2005 IEEE Internati<strong>on</strong>al C<strong>on</strong>ference <strong>on</strong> Robotics<br />
and Automati<strong>on</strong>, 2005. ICRA 2005., pages 3930–3935, 18-22 April<br />
2005.<br />
932