Incremental Estimation of Model Parameter for Deformable Object in Manipulation from Sequential Observation

Tomohiro YABUUCHI  Koh KAKUSHO  Michihiko MINOH  

D - Abstracts of IEICE TRANSACTIONS on Information and Systems (Japanese Edition)   Vol.J90-D   No.1   pp.94-105
Publication Date: 2007/01/01
Online ISSN: 1881-0225
Print ISSN: 1880-4535
Type of Manuscript: PAPER
virtualization,  observation,  parameter estimation,  real coded genetic algorithm,  

Full Text(in Japanese): PDF(621.4KB)
>>Buy this Article

This paper discusses modeling real deformable objects for creating a virtual object that reproduces the shapes observed from the reaction of real deformable object under manipulation. For this purpose, mass-spring models with many parameters are introduced and can produce realistic shapes of deformable objects when optimal values are set for their parameters. In this paper, we propose a method of estimating model parameters using Real Coded Genetic Algorithm from the shapes of real object by not all the observation data in advance but sequential observation. Strings and clothes are considered as examples of deformable objects. Experimental results show that the model with optimal values of its parameters estimated from our method can reproduce the shapes which were not used for the estimation.