1 Modular Parallel Manipulator for Long-Term Soft Robotic Data Collection

TL;DR

The platform's ability to be used for policy gradient reinforcement learning directly on hardware in a benchmark 2D manipulation task is validated and compatibility with multiple fingers is demonstrated, as well as the design constraints for compatible extensions.

Abstract

Performing long-term experimentation or large-scale data collection for machine learning in the field of soft robotics is challenging, due to the hardware robustness and experimental flexibility required. In this work, we propose a modular parallel robotic manipulation platform suitable for such large-scale data collection and compatible with various soft-robotic fabrication methods. Considering the computational and theoretical difficulty of replicating the high-fidelity, faster-than-real-time simulations that enable large-scale data collection in rigid robotic systems, a robust soft-robotic hardware platform becomes a high priority development task for the field. The platform's modules consist of a pair of off-the-shelf electrical motors which actuate a customizable finger consisting of a compliant parallel structure. The parallel mechanism of the finger can be as simple as a single 3D-printed urethane or molded silicone bulk structure, due to the motors being able to fully actuate a passive structure. This design flexibility allows experimentation with soft mechanism varied geometries, bulk properties and surface properties. Additionally, while the parallel mechanism does not require separate electronics or additional parts, these can be included, and it can be constructed using multi-functional soft materials to study compatible soft sensors and actuators in the learning process. In this work, we validate the platform's ability to be used for policy gradient reinforcement learning directly on hardware in a benchmark 2D manipulation task. We additionally demonstrate compatibility with multiple fingers and characterize the design constraints for compatible extensions.

Figures (9)

• Figure 1: A) Motion of a 1-DOF soft foam manipulator designed to perform a simple open-loop pushing task. B) Eventual task failure due to material aging from sun exposure leading to dynamics drift. C) Intrinsic time-dependent deformation response of soft material to cyclical loading.
• Figure 2: Left) Rigid body kinematic approximation of five-bar. Right) Approximate forces of soft material five-bar driven by two servos. T1 and T2 are the actuation space of the system, the torques applied by the servos. Springs $k_1, k_2,$ and $k_3$ are the spring constants of living hinge joints of the five-bar, dependent on hinge material and geometry. The force $F_1$ is the reaction force that occurs when the tip of the five-bar contacts another object. The spring constants of the five-bar links are not shown, but do influence behavior.
• Figure 3: A) Full experimental manipulator platform B) Soft five-bar mechanism design with quick-swap connection slots. C) Individual module design.
• Figure 4: A) By mounting the compliant base of the module to mechanical breadboard using a linear arrangement of bolts, the module is able to rock back, pushing the finger out of plane in response to overly high forces. This reduces the chance of breaking the system during extended unmonitored data collection. B) The most likely failure modes are module disassembly to to friction fit module components and soft five-bar failure, especially at living hinges. Even if both of these occur undetected and the system keeps trying to operate, the compliant materials used and construction lowers the chance of long-term damage to the experimental setup.
• Figure 5: A) Motion primitive execution without attached finger, showing servo horn finger connectors moving. B) Execution of that primitive with a urethane finger mounted to module.