Designing robotic systems that can change their physical form factor as well as their compliance to adapt to environmental constraints remains a major conceptual and technical challenge. To address this, we introduce the Granulobot, a modular system that blurs the distinction between soft, modular, and swarm robotics. The system consists of gear-like units that each contain a single actuator such that units can self-assemble into larger, granular aggregates using magnetic coupling. These aggregates can reconfigure dynamically and also split up into subsystems that might later recombine. Aggregates can self-organize into collective states with solid- and liquid-like properties, thus displaying widely differing compliances. These states can be perturbed locally via actuators or externally via mechanical feedback from the environment to produce adaptive shape shifting in a decentralized manner. This in turn can generate locomotion strategies adapted to different conditions. Aggregates can move over obstacles without using external sensors or coordinate to maintain a steady gait over different surfaces without electronic communication among units. The modular design highlights a physical, morphological form of control that advances the development of resilient robotic systems with the ability to morph and adapt to different functions and conditions.

翻译：设计能够改变物理形态以及自身柔顺性以适应环境约束的机器人系统，仍是一项重大的概念与技术挑战。为此，我们提出了Granulobot——一种模糊了软体机器人、模块化机器人与群体机器人之间界限的模块化系统。该系统由齿轮状单元组成，每个单元包含一个单独的执行器，使得单元能够利用磁耦合自组装成更大的颗粒状聚合体。这些聚合体可以动态重构，也能分裂成子系统，随后再重新组合。聚合体能够自组织成具有固态和液态特性的集体状态，从而展现出差异极大的柔顺性。这些状态可通过执行器进行局部扰动，或通过环境机械反馈进行外部扰动，从而以去中心化的方式产生适应性形变。这进而能够生成适应不同条件的运动策略。聚合体无需使用外部传感器即可越过障碍物移动，或在单元间无电子通信的情况下协调维持不同表面上的稳定步态。该模块化设计凸显了一种物理形态控制方式，推动了能够变形并适应不同功能与条件的弹性机器人系统的发展。