Measuring and update of shape of pile for loading operation by wheel loader

Authors have been conducting research on an unmanned loading system by wheel loader. The system has function of sensing, decision making and control of motion. The operation of the system is autonomous completely without any assist of human operator. As a part of the research, the detection method of location and shape of the pile will be described in this paper. At the beginning of the loading operation, scooping positions should be decided based on the location and shape of the pile. Stereo vision system with two CCD cameras is attached on the loader. For obtaining three dimensional data of the shape of the pile, a correlation method is applied on the images. The 3D data is converted into the column model by quantization. The scooping positions are decided based on the data of the pile and the location and direction of the dump truck for effective operation at the beginning state of the operation. Because the shape of the pile is changed by scooping with advancing the loading operation, the measured shape has to be updated during operation. After several scooping, the images of the pile are captured during travelling between the scooping position and the loading position. It is not necessary to slow down of velocity of the loader up to 8km/h for capturing. 3D shape and column model of the pile is obtained in the same manner. The proposed method is applied on the experimental system for autonomous loading operation and shows good results in the test field trial.