Semi-autonomous control of excavation using model of human operation and state estimation via Extended Kalman Filter

Abstract

For robotic teleoperation, semi-autonomous control, i.e. autonomous control allowing human interposition will be an effective way overcoming delay of electrical communication and emergency operation. We have proposed semiautonomous control method focusing on excavation in teleoperation environments. In the method, a model of human operation is designed and human internal state is estimated by Extended Kalman Filter. In this paper, experimental evaluations are conducted and reported about "Digging" process. A human skill of excavation is modeled by an attractorbased dynamics embedding shallow and deep digging trajectories for soft and hard soil, respectively. An index parameter is changed based on the load of the bucket, and a seamless motion transition is realized satisfying the restriction by setting a threshold. Moreover, by adding a human operation to the autonomy, the proposed method will cope with the emergency, which is experimentally evaluated.

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

“…Semi-autonomous robotic systems involving continuous deformation of the task trajectory using parameterized dynamical systems have been studied so far. Okada et al (2020) have proposed a semi-autonomous control system for an excavation robot with shallow and deep digging trajectories for soft and hard ground embedded in an attractor-based dynamical system. Seamless transition of digging depth is realized by changing the index parameters according to the bucket load and human operation using a joystick.…”

“…In the context of pHRI (Physical Human-Robot Interaction), Khoramshahi et al (2018) have proposed transforming the task trajectory using a cooperative robot to adapt a parameterized dynamic system to the human intended task online. However, no method has been proposed to combine task selection (as proposed in (Okada et al, 2021;Pistillo et al,2011;Khoramshahi et al, 2019)) and task trajectory deformation (as proposed in (Okada et al, 2020;Khoramshahi et al, 2018)) in nonlinear dynamics in a single robot to achieve operation separation. We contribute to these research areas by proposing a semi-autonomous leader-follower system that deforms the task trajectory in a nonlinear dynamical system with attractor and achieves operational separation between task selection and trajectory deformation.…”