Co-manipulation with a library of virtual guiding fixtures

Abstract

Virtual guiding fixtures constrain the movements of a robot to task-relevant trajectories, and have been successfully applied to, for instance, surgical and manufacturing tasks. Whereas previous work has considered guiding fixtures for single tasks, in this paper we propose a library of guiding fixtures for multiple tasks, and propose methods for 1) Creating and adding guides based on machine learning; 2) Selecting guides on-line based on probabilistic implementation of guiding fixtures; 3) Refining existing guides based on an incremental learning method. We demonstrate in an industrial task that a library of guiding fixtures provides an intuitive haptic interface for joint human-robot completion of tasks, and improves performance in terms of task execution time, mental workload and errors.

A Haptic Shared-Control Architecture for Guided Multi-Target Robotic Grasping

“…With respect to our work, the gripper orientation is fixed and autonomously controlled based on the selected grasp pose. Raiola et al [61] presented a set of virtual fixtures which they demonstrated in a pick-and-place industrial task. Completion time decreased by approximately 15% compared to standard teleoperation.…”

“…A passive switching mechanisim allows the operator to safely switch from one target to another through virtual damping injection. A different approach is proposed in [47] for handling multi-target fixtures when controlling a serial manipulator. The authors implement a control scheme in which all fixtures are simultaneously active, but they are scaled according to the probability that their corresponding task is chosen.…”

“…However, the magnet-like design has only been used for the two or three dimensional case. In applications where more DoFs are considered, a virtual spring-damper system has always been employed to guide the user to the different targets, while an intention recognition algorithm has been in charge of predicting the user's intention and scaling or activating/deactivating fixtures as needed to avoid noise and distracting forces [47], [48], [49]. However, such an implementation is prone to error when the intention of the user is not inferred properly and the on/off switching of the guidance can generate confusion.…”

“…Whenever the robotic gripper comes close to an object, the data generated by the grasping algorithm is used to assist the operator in approaching a suitable grasp candidate. The usual approach when handling 6-DoF multi-target scenarios consists in implementing a virtual spring between the current pose of the robot and the different target poses [46], [47], [48], [49], [56]. In this case, the operator receives a force which increases as the gripper moves away from the target pose and decreases as it approaches it.…”

On Inferring Intentions in Shared Tasks for Industrial Collaborative Robots

“…The robots avoided the physical interaction with the human as long as possible, and when a physical interaction occurred, they remained in a compliant mode so that the force was ignored. Raiola et al [14] addressed the problem of learning virtual guiding fixtures, analogous to the use of a rule when drawing, in human-robot collaboration. Even though there was physical interaction during the task execution, the robot did not use the force based information while guiding the human.…”