LAUSR

Abstract Wheel-loaders are versatile multi-purpose machines used in construction and mining. Recent studies have shown that wheel-loaders have a significant optimization potential as different operators can account for up to 300% difference in productivity and 150% in fuel efficiency. A better understanding of the different work tasks a wheel-loader performs on-site allows for the implementation of various optimization strategies. In this paper a method for finding and classifying wheel-loader work-cycles is presented. The proposed method uses a dynamic time warping path to align the sensor data to predefined class templates. With this alignment method a phase independent detection and classification of work-cycles can be performed in a robust manner. The proposed method also allows for a short-term prediction of the future signal trajectories, which could be used for an online optimization task of the wheel-loader. The performance of the method was tested with data from field trials and validated with video-recordings.