Скачать или смотреть Uncertainty Learning Stochastic Image-based Visual Predictive Control

Image-based visual predictive controllers have attracted
the attention from industry and the research community
alike due to their constraint handling capabilities and optimality.
However, their performance degrades when modeling-errors and
measurement-uncertainties are present. This paper introduces a
computationally-traceable uncertainty-learning-based stochastic
image-based visual predictive control that provides a balance on
system’s performance and the constraint violation probabilities.
We capture the statistical model of the uncertainties through
Gaussian process regressors, and consequently deliver a robust
and constraint-compliant solution. The proposed method was
evaluated via simulation and also on a 6-DOF serial industrial
manipulator.