Identification of dominant error force component in hydraulic pressure reading for external force detection in construction manipulator

The purpose of this paper is to develop a fundamental external-force-detection framework for construction manipulators. Such an industrial application demands the practicality that satisfies detection requirements such as the accuracy and robustness while ensuring (i) a low cost, (ii) wide applicability, and (iii) a simple detection algorithm. For satisfying (i) and (ii), our framework first adopts a hydraulic sensor as a force sensor. However, hydraulic-pressure readings essentially include error force components. These components depend strongly on the joint kinetic state and differ in the identification difficulty owing to a nonlinear and uncertain hydromechanical system. For satisfying (ii) and (iii), our framework thus focuses on the dominant error-force components classified by the control input states, such as self-weight, cylinder driving, and oscillating forces, and identifies and removes them by using a theoreticalmodel, an experimental estimation, and a waveform analysis without complex modeling, respectively. Experiments were conducted using an instrumented hydraulic arm system. The results of a no-load task indicate that our framework greatly lowers the threshold to determine the on-off state of external force application, independent of the joint kinetic states. The results of an on-load task confirm that our framework robustly identifies the off states in which an external force is not applied to the hydraulic cylinder.