Intuitive Hand Teleoperation by Novice Operators Using a Continuous Teleoperation Subspace

Abstract

Human-in-the-loop manipulation is useful in when autonomous grasping is not able to deal sufficiently well with corner cases or cannot operate fast enough. Using the teleoperator's hand as an input device can provide an intuitive control method but requires mapping between pose spaces which may not be similar. We propose a low-dimensional and continuous teleoperation subspace which can be used as an intermediary for mapping between different hand pose spaces. We present an algorithm to project between pose space and teleoperation subspace. We use a non-anthropomorphic robot to experimentally prove that it is possible for teleoperation subspaces to effectively and intuitively enable teleoperation. In experiments, novice users completed pick and place tasks significantly faster using teleoperation subspace mapping than they did using state of the art teleoperation methods.

EMG-Controlled Non-Anthropomorphic Hand Teleoperation Using a Continuous Teleoperation Subspace

“…Here we use a low dimensional subspace to teleoperate a non-anthropomorphic hand. We have shown that this subspace is useful for teleoperation in previous work [5]; however, our previous formulation required a dataglove for teleoperation control. In this work, we project forearm EMG into this subspace and use it as a control for a fully actuated, multi-degree-of-freedom (DOF) robot ( Figure 1).…”

“…for teleoperation [5]. We briefly review the mapping between pose space and our teleoperation subspace below.…”

“…o, A and δ * represent an offset, a linear mapping and a scaling factor respectively, all of them specific to the robot hand that is being used. Details on how to build these elements of the mapping can be found elsewhere [5].…”

“…The control was trained as follows: we passed EMG signals through a bandpass filter and projected joint angles into T to provide a ground truth values for σ and . We project the joint angles into T using the equation ψ = ((q−o)·A) δ [5]. We trained the regressor on all data where the user was not performing a gesture.…”

Underactuation Design for Tendon-Driven Hands via Optimization of Mechanically Realizable Manifolds in Posture and Torque Spaces

“…One specific idea in this category of robot hand control is to use low-dimensional synergies as a human-robot interface for teleoperation, learning by demonstration, and prosthetics, in order to reduce the required communication bandwidth between the human and the robot. For teleoperation, this method has been shown in the studies from Gioioso et al [5] and Meeker et al [6]. For hand prosthetics, Matrone et al [7] [8] as well as Tsoli and Jenkins [9] presented the aforementioned idea and developed working prototypes.…”