Discrete / continuous task bifurcation by frequency separation of human operation for semi–autonomous excavation

Abstract

Conventional remote operation of excavators is less efficient than direct operation. To address this, we have developed a semi-autonomous control system that combines autonomy (a dynamical system with an attractor) and human action (admittance control), and proposed discrete task selection within the dynamical system. In this paper, we propose a semi-autonomous leader-follower excavation system that achieves continuous task trajectory deformation in a nonlinear dynamical system with an attractor and separates task selection from trajectory deformation. The attractor is designed in a virtual space and transformed into the bucket's state via coordinate transformation to change only the digging position without altering the loading position. Trajectory deformation is estimated by an Extended Kalman Filter based on human operational input and set dynamic characteristics, with task selection operations frequency-separated from the deformation. We implemented and verified the proposed method using a prototype excavation robot.