@incollection{JanSmaGro1992, Author = {Jansen, Arjan and Smagt, Patrick van der and Groen, Frans}, Title = {High-precision robot control: The nested network}, Year = {1992}, Pages = {583-586}, Month = {Sep.}, Editor = {I. Aleksander and J. Taylor}, Publisher = {North-Holland/Elsevier Science Publishers}, Address = {Amsterdam, The Netherlands}, Booktitle = {Artificial Neural Networks 2}, Keywords = {machine-learning brml}, Abstract = {Traditional robot control is based on precise models of sensors and manipulators. Increasing complexity of required tasks, sensors, and manipulators result in models which are extremely complex to build at the required precision. In the realm of pick-and-place problems, we aim at designing highly adaptive controllers which require minimum knowledge of the manipulator and its sensors. In this area, several models have proven successful in one area or another. The use of a single feed-forward network trained with conjugate gradient back-propagation gives fast and highly adaptive approximation, but needs up to ten feedback steps to get high-precision results. Kohonen networks give a precision up to 0.5cm.~with only two steps, but need thousands of iterations to attain reasonable results. In this paper we introduce the nested network method based on search trees which adapts in real-time and reaches a grasping precision of up to 1mm.~in only three steps.} } @COMMENT{Bibtex file generated on 2018-10-9 with typo3 si_bibtex plugin. Data from https://brml.org/publications/publications/ }