A Fast and Reliable Pick-and-Place Application with a Spherical Soft Robotic Arm

Abstract

This paper presents the application of a learning control approach for the realization of a fast and reliable pick-and-place application with a spherical soft robotic arm. The arm is characterized by a lightweight design and exhibits compliant behavior due to the soft materials deployed. A soft, continuum joint is employed, which allows for simultaneous control of one translational and two rotational degrees of freedom in a single joint. This allows us to axially approach and pick an object with the attached suction cup during the pick-and-place application. An analytical mapping based on pressure differences and the antagonistic actuator configuration is introduced, allowing decoupling of the system dynamics and simplifying the modeling and control. A linear parametervarying model is identified, which is parametrized by the attached load mass and a parameter related to the joint stiffness. A gain-scheduled feedback controller is proposed, which asymptotically stabilizes the robotic system for aggressive tuning and over large variations of the parameters considered. The control architecture is augmented with an iterative learning control scheme enabling accurate tracking of aggressive trajectories involving set point transitions of 60 degrees within 0.3 seconds (no mass attached) to 0.6 seconds (load mass attached). The modeling and control approach proposed results in a reliable realization of a pick-and-place application and is experimentally demonstrated.

Offset-free Model Predictive Control: A Ball Catching Application with a Spherical Soft Robotic Arm

“…The spherical soft robotic arm originates from [19] and an improved version is used in [18]. The individual components of the soft manipulator are explained in Fig.…”

“…The three actuator pressures (p A , p B , p C ) form the inputs to the robotic arm. Beside the two orientation degrees of freedom, there is an additional degree of freedom related to the joint stiffness of the system (see [18], [19]). In order to simplify the modeling of the system, an alternative representation of the three inputs is employed using two pressure differences (∆p α , ∆p β ) and a lower pressure bound p (see [18] for a detailed description).…”

“…Beside the two orientation degrees of freedom, there is an additional degree of freedom related to the joint stiffness of the system (see [18], [19]). In order to simplify the modeling of the system, an alternative representation of the three inputs is employed using two pressure differences (∆p α , ∆p β ) and a lower pressure bound p (see [18] for a detailed description). The two pressure differences are related to the orientational degrees of freedom and the lower pressure bound to the joint stiffness.…”

“…1). As the application requires the controller to track a trajectory that is unknown beforehand, previously explored control approaches such as iterative learning control (as used in [18] for realizing a pick and place application) are not feasible for this application. Therefore, an offset-free MPC approach that can deal with an a priori unknown reference trajectory is employed for realizing the ball catching application.…”

A Vision-based Sensing Approach for a Spherical Soft Robotic Arm

“…The spherical robotic arm is closely related to the system presented in Zughaibi et al (2020) and consists of two inflatable links and three fabric-based bellow actuators that are arranged symmetrically around a soft silicone joint connecting the two links. The robot arm has two rotational degrees of freedom, which are described by the extrinsic Euler angles α and β (see Figure 2).…”