不整地掘削実験のための自動掘削・砂貯蔵システムの構築と評価

抄録

In this study, we developed an automatic sand-pooling and dumping device for data-based modeling and optimization of hydraulic excavators. In order to design an efficient motion trajectory of the excavator bucket for given unlevel terrain and surrounding conditions, we propose an CNN-based approach using the height map of the terrain surface built by a depth-camera, with which we find out an efficient trajectory by sequential optimization. Now a technical issue here is how to collect huge dataset for optimization; we then implement a ROS-controlled automatic data-collection system composed of a robotic bucket, sand-pooling, dumping and re-leveling devices, which enables repetitive experiments of unlevel excavation.