Premium
Sampling‐Based Motion Planning with a Prediction Model using Fast Gaussian Process Regression
Author(s) -
OKADOME YUTA,
NAKAMURA YUTAKA,
ISHIGURO HIROSHI
Publication year - 2017
Publication title -
electronics and communications in japan
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.131
H-Index - 13
eISSN - 1942-9541
pISSN - 1942-9533
DOI - 10.1002/ecj.11954
Subject(s) - gaussian process , kriging , computer science , motion planning , nonparametric regression , process (computing) , machine learning , artificial intelligence , regression analysis , gaussian , robot , physics , quantum mechanics , operating system
SUMMARY Recently, motion/path planning methods have become popular and been put to use in practical applications such as autonomous cars. Most studies focus on planning based on a given mathematical model, but the development of the data‐driven system identification method is also crucial for practical applications, because the precise model of a control target is not always available in advance. In this research, we propose a motion planning method where a fast Gaussian process regression is used as a model of the control target, since Gaussian process regression is a powerful nonparametric method that is widely used in various practical and complicated applications. Thanks to the Bayesian property of Gaussian process regression, our method can deal with uncertainty in the predictions. We apply our method to the control problem of simple‐pendulum with nonlinearity, and complete the swing‐up task.