Abstract
A state estimation-based approach is proposed to determine both the forces exchanged between the soil and the bucket of a hydraulic excavator during digging tasks and the load of the soil accumulated in the bucket. The suggested approach ensures real-time force estimates, which constitutes a necessary preliminary requirement to address the open issue of unmanned earthmoving machine development. The estimates are carried out employing a two-stage estimation approach. It requires the implementation of two observers partially coupled based on kinematic and dynamic excavator models, and measurements which can be collected easily and by inexpensive sensors. An estimation regularization is introduced, which allows to get accurate estimates also in the presence of ill-conditioned system models. The approach is validated by applying it to a very accurate numerical simulator. The sensitivity analysis carried out at varying level of modeling errors highlights that digging and payload forces can be estimated efficiently and effectively in the presence of both relevant model and measurement errors.