4. Homemate v3
●
10 IR range-finders
●
8 motors
●
Robotic hand
●
Stargazer
●
3D camera
●
●
Linux machine controls
HW
Windows notebook is the
brain
–
Ethernet
9. 3D vision and octree cell
representation
ISRC develops cognitive
recognition
system
Octree cells are used for object segmentation
and recognition to make decision
Lee et al. 2012
10. Obstacle detection
●
●
●
I simply assume that
each there is an
obstacle in place of the
octree cell
Mapping of the octree
cells into 2D map
Grid size: 10cm
14. D* lite
●
●
Koenig, S., and M. Likhachev.
“Improved Fast Replanning for
Robot Navigation in Unknown
Terrain.” In IEEE International
Conference on Robotics and
Automation, 2002. Proceedings.
ICRA ’02, 1:968–975 vol.1, 2002.
doi:10.1109/ROBOT.2002.1013481.
Koenig, S., and M. Likhachev. “Fast
Replanning for Navigation in
Unknown Terrain.” IEEE
Transactions on Robotics 21, no. 3
(2005): 354–363.
doi:10.1109/TRO.2004.838026.
18. Moving the robot in reality:
troubles and issues
●
Closed platform : limited control of the robot
–
–
●
●
Move forward/backward X cm
Turn left/right X Degrees
Cannot apply any feedback trajectory tracker →
moving
But feedforward and slipping → very inaccurate
→ limiting the grid size
20. Conclusion
Problems
What I achieved
●
●
D* lite and obstacle
mapping seem to work
well
Robot is most of time
able to navigate to the
goal in the simple
environment but it is
troublesome
●
●
Not obstacle
forgetting
In more complex
environment the robot
fails
21. What did I taken from the project
●
●
●
●
I wish I use a smaller robot
Robots are fun but life is never as easy as the
simulations
It is useful to know Korean when working on
Korean computer :-)
Confirmed again Murphy's laws
–
"Anything that can go wrong, will go wrong".
→ Opt for more theoretical thesis :-)