Design and Control of Soft Robots Using Differentiable Simulation

Abstract

Purpose of Review We discuss the use of differentiable simulation for computational problems in soft robotics. This includes characterizing the mechanical behavior of soft robots, optimally controlling embedded soft actuators or active materials, and estimating the robot's state from readings of embedded sensors. Moreover, we discuss how design optimization can help to optimally place soft actuators and sensors. Recent Findings We expatiate on the adoption of simulation and optimization tools in the process of designing and controlling soft robots. We include a discussion of rigid-flexible systems and the use of differentiable simulation in combination with machine learning. Summary We review the state of the art in the computational modeling of soft robots and provide a summary of the required mathematical tools. We also review several open questions where computation could help to move the field forward, and discuss the role of differentiable simulation in managing the ever-growing design complexity of next-generation soft robots. Keywords Differentiable simulation • Computational design and control • State estimation • Sensor and actuator design This article is part of the Topical Collection on Soft Robotics

A Recurrent Differentiable Engine for Modeling Tensegrity Robots Trainable with Low-Frequency Data

“…The Constrained Recursive Newton-Euler Algorithm (RNEA) [10], [11] uses differentiable physics engines for rigid-body chain manipulators [19], [6]. Differentiable dynamical constraints can also be applied to path optimization [20], training by image supervision [21], [22], soft object manipulation [23], soft robot control [24], [25] and quantum molecular control [26]. A differentiable engine specifically designed for tensegrity robots [12] provides analytically explainable models for both the robot and the ground, but still requires densely sampled ground-truth trajectories for system identification.…”

Differentiable solver for time-dependent deformation problems with contact

“…Automatic differentiation is more flexible but requires existing simulators to be essentially rewritten and incurs a performance penalty, especially for complex solvers. Surrogate models, though promising dramatic speedups, require huge training sets for even simple design spaces [Gavriil et al 2020], are unsuitable for high-precision applications, and are opaque black boxes [Bächer et al 2021]. To the best of our knowledge, none of these simulators support robust handling of contact and friction for complex geometries, and they only support a subset of the design parameters compared to the generic formulation of this paper.…”

“…Meshfree Methods. Especially for shape optimization, methods like XFEM [Hafner et al 2019;Schumacher et al 2018] and MPM [Hu et al 2019b] that do not maintain conforming meshes might appear attractive to circumvent remeshing-induced discontinuities [Bächer et al 2021]. However, these methods sacrifice accuracy [de Vaucorbeil et al 2019], particularly for stress minimization problems [Sharma and Maute 2018].…”

Recent advances in biomimetic soft robotics: fabrication approaches, driven strategies and applications

“…To improve the preparation efficiency, it is necessary to combine some appropriate preparation methods. Some reviews about soft robots have been published in recent years, including their design and fabrication process, 3,16,53,54 selected materials, [55][56][57][58][59] manufacturing methods, [60][61][62][63] actuation technologies, 64 and application opportunities. 65 However, a comprehensive overview of the field of soft robotics is necessary and significant for readers to quickly understand this field.…”