Compliant Underwater Manipulator with Integrated Tactile Sensor for Nonlinear Force Feedback Control of an SMA Actuation System.

Design, sensing, and control of underwater gripping systems remain challenges for soft robotic manipulators. Our study investigates these critical issues by designing a shape memory alloy (SMA) actuation system for a soft robotic finger with a directly 3D-printed stretchable skin-like tactile sensor. SMA actuators were thermomechanically trained to assume a curved finger-like shape when Joule heated, and the flexible multi-layered tactile sensor was directly 3D-printed onto the surface of the fingertip. A nonlinear controller was developed to enable precise fingertip force control using feedback from the compliant tactile sensor. Underwater experiments were conducted using closed-loop force feedback from the directly 3D-printed tactile sensor with the SMA actuators, showing satisfactory force tracking ability. Furthermore, a 3D finite element model was developed to more deeply understand the shape memory thermal-fluidic-structural multi-physics simulation of the manipulator underwater. An application for human control via electromyogram (EMG) signals also demonstrated an intuitive way for a person to operate the submerged robotic finger. Together, these results suggested that the soft robotic finger could be used to carefully manipulate fragile objects underwater.

Shape memory alloy tube actuators inherently enable internal fluidic cooling for a robotic finger under force control.

This paper presents the design, control and evaluation of a novel robotic finger actuated by shape memory alloy (SMA) tubes which intrinsically afford an internal conduit for fluidic cooling. The SMA tubes are thennomechanically programmed to flex the robotic finger when Joule heated. A superelastic SMA plate provides a spring return motion to extend the finger when cooling liquid is pumped through the internal channel of the SMA tube actuators. The mechanical design and nonlinear force controller are presented for this unique robotic finger. Sinusoidal and step response experiments demonstrate excellent error minimization when operated below the bandwidth which was empirically determined to be 6 rad s-1. Disturbance rejection experiments are also performed to demonstrate the potential to minimize externally applied forces. This method of internal liquid cooling of Joule heated SMA tubes simultaneously increases the system bandwidth and expands the potential uses of SMA actuators for robotic applications. The results show that this novel robotic finger is capable of precise force control and has a high strength to weight ratio. The finger can apply a force of 4.35 N and has a mass of 30 g. Implementing this design into wearable prosthetic devices could enable lightweight, high strength applications previously not achievable.

Anthropomorphic finger antagonistically actuated by SMA plates.

Most robotic applications that contain shape memory alloy (SMA) actuators use the SMA in a linear or spring shape. In contrast, a novel robotic finger was designed in this paper using SMA plates that were thermomechanically trained to take the shape of a flexed human finger when Joule heated. This flexor actuator was placed in parallel with an extensor actuator that was designed to straighten when Joule heated. Thus, alternately heating and cooling the flexor and extensor actuators caused the finger to flex and extend. Three different NiTi based SMA plates were evaluated for their ability to apply forces to a rigid and compliant object. The best of these three SMAs was able to apply a maximum fingertip force of 9.01N on average. A 3D CAD model of a human finger was used to create a solid model for the mold of the finger covering skin. Using a 3D printer, inner and outer molds were fabricated to house the actuators and a position sensor, which were assembled using a multi-stage casting process. Next, a nonlinear antagonistic controller was developed using an outer position control loop with two inner MOSFET current control loops. Sine and square wave tracking experiments demonstrated minimal errors within the operational bounds of the finger. The ability of the finger to recover from unexpected disturbances was also shown along with the frequency response up to 7 rad s(-1). The closed loop bandwidth of the system was 6.4 rad s(-1) when operated intermittently and 1.8 rad s(-1) when operated continuously.