時変ガウス過程陰関数曲面による近接センサを用いたバケット内ごみの形状推定

抄録

This paper proposes time-varying Gaussian process implicit surface (TV-GPIS) to estimate the time-varying shape of grasped garbage by a waste crane placed in an Energy-from-Waste plant. TV-GPIS can estimate the time-varying shape by integrating the time-varying Gaussian process and the Gaussian process implicit surface. We first investigated the properties of TV-GPIS for three types of shape transitions in 2D simulation. Then, we present a robotic bucket system we developed equipped with 28 proximity sensors and verified the TV-GPIS’s effectiveness by comparing the volume of estimated shape and grabbed garbage by our bucket.