Abstract of FKI-182-93
Title: A Neural Trajectory Storage
Autors: Margit Kinder
Revision Date: January, 1993
Category: Technical Report (Forschungsberichte Künstliche Intelligenz)
Abstract: We present a neural trajectory storage for manipulator
trajectories. Trajectories will be neurally stored as a
sequence of discrete joint positions. Every time a trajectory
needs to be generated, the storage is consulted.
Besides being a ``classical'' retrieval system, the neural
network is able to generalize from stored trajectories to
new trajectories and thus serves as a trajectory planner.
Its generalization abilities arise from an adequate
representation of free space. We recommend the use of
Kohonen's self-organizing feature maps to represent input
and output data, both data of the same free space.
Different to others, we do not compute the winner unit but
instead feed the activities of all units of a Kohonen-Map into
the next layer.
Keywords: neural trajectory storage; redundant manipulator; robotics;
self-organizing feature maps; topologically distributed
Size: 6 pages
Copyright: The ``Forschungsberichte Künstliche Intelligenz''
series includes primarily preliminary publications,
specialized partial results, and supplementary
material. In the interest of a subsequent final
publication these reports should not be copied. All
rights and the responsability for the contents of the
report are with the authors, which would appreciate