Traditional robot designs typically employ rigid body parts and high-torque servo motors. This helps to obtain simple reproducible models and, therefore, to facilitate control. Soft robotics is a new research field that deliberately expands the design toolbox to a wide range of smart and, often, soft materials. This approach is generally inspired by the remarkable performance of biological systems, which use soft structures to interact successfully with noisy and hard-to-model environments and, as a result, outperform state-of-the-art robots in open-world scenarios. However, using soft bodies comes with a significant disadvantage. Soft materials often have complex and nonlinear dynamics, which makes them hard to model and therefore difficult to control. To fulfill the potential of soft robotics to achieve performances close to biological systems, this control problem has to be solved. A promising solution is another bioinspired principle called morphological computation, which proposes to outsource functionality directly to the body morphology. From this point of view, the seemingly undesired nonlinear dynamics become a resource for implementing nonlinear functionalities. This extends the control design problem to the question of how to design the body morphology of the robot. While there exist proofs of concept that demonstrate the potential of this approach, the existing work (for the most part) is lacking mathematical rigor and a general framework. We believe that the control community has the right set of tools to support the development of a design framework for morphological computation. The goal of this article (see “Summary”) is to provide an introduction to the concepts of soft robotics and morphological computation, explain how they can work together, and (with the help of examples) illustrate their potential for control. The hope is to inspire members of the control community to develop novel control frameworks for the next generation of soft robots.

中文翻译：

---

利用形态计算控制软机器人：从自然中学习来控制软机器人


传统的机器人设计通常采用刚性身体部件和高扭矩伺服电机。这有助于获得简单的可重复模型，从而促进控制。软机器人技术是一个新的研究领域，它有意将设计工具箱扩展到各种智能材料（通常是软材料）。这种方法通常受到生物系统卓越性能的启发，生物系统使用软结构与嘈杂且难以建模的环境成功交互，因此在开放世界场景中优于最先进的机器人。然而，使用软体有一个明显的缺点。软材料通常具有复杂的非线性动力学特性，这使得它们难以建模，因此难以控制。为了发挥软机器人技术的潜力，实现接近生物系统的性能，必须解决这个控制问题。一个有前途的解决方案是另一种称为形态计算的生物启发原理，它建议将功能直接外包给身体形态。从这个角度来看，看似不需要的非线性动力学成为实现非线性功能的资源。这就将控制设计问题延伸到了如何设计机器人的身体形态的问题。虽然存在证明这种方法潜力的概念证明，但现有工作（大部分）缺乏数学严谨性和通用框架。我们相信控制社区拥有正确的工具集来支持形态计算设计框架的开发。 本文的目标（参见“摘要”）是介绍软机器人和形态计算的概念，解释它们如何协同工作，并（借助示例）说明它们的控制潜力。希望能够激励控制社区的成员为下一代软机器人开发新颖的控制框架。