|
|
Researches
|
|
Distributed object manipulation (Back) |
We are interested in distributed object manipulations because of its futuristic
applications in addition to existing challenges in design of distributed cooperation
and control protocols. Unlike many other multi-robot tasks, distributed object manipulation
requires close coordination. In addition, object manipulation poses great challenges
to learning methods.
In this research our goal is to devise a set of hand designed coordination and cooperation
protocols for distributed object manipulation by mobile robots. In addition, we
try to tailor reinforcement learning for learning required individual skills and
coordination protocols. Our approach to learning distributed object manipulation
is exploiting learned knowledge in individual mode for speeding up learning cooperation
protocols, which is quite effective and promising.
|