| description |
In the problem of motion control for mobile robots, typically only
the kinematic model is used, assuming that there is a dynamic
controller that can produce perfect velocity tracking. However, when
the dynamic model of the robot is considered, exact knowledge about
its parameters is almost unattainable in practical situations. In
this paper, a novel recurrent neural network called Echo-State
Network is used to develop a dynamic-level controller, without
knowledge about the robot parameters. The control approach has been
experimentally tested on an omnidirectional mobile robot available
at the Robotics Lab of the University of Stuttgart.
|