More Like this

1. Learning Differentiable Tensegrity Dynamics using Graph Neural Networks

   Chen, N; Wang, K; Johnson_III, W; Kramer-Bottiglio, R; Bekris, Kostas E; Aanjaneya, M (November 2024, Conference on Robot Learning (CoRL))

   Tensegrity robots are composed of rigid struts and flexible cables. They constitute an emerging class of hybrid rigid-soft robotic systems and are promising systems for a wide array of applications, ranging from locomotion to assembly. They are difficult to control and model accurately, however, due to their compliance and high number of degrees of freedom. To address this issue, prior work has introduced a differentiable physics engine designed for tensegrity robots based on first principles. In contrast, this work proposes the use of graph neural networks to model contact dynamics over a graph representation of tensegrity robots, which leverages their natural graph-like cable connectivity between end caps of rigid rods. This learned simulator can accurately model 3-bar and 6-bar tensegrity robot dynamics in simulation-to-simulation experiments where MuJoCo is used as the ground truth. It can also achieve higher accuracy than the previous differentiable engine for a real 3-bar tensegrity robot, for which the robot state is only partially observable. When compared against direct applications of recent mesh-based graph neural network simulators, the proposed approach is computationally more efficient, both for training and inference, while achieving higher accuracy. Code and data are available at https://github.com/nchen9191/tensegrity_gnn_simulator_public 

   more » « less
2. Learning Differentiable Tensegrity Dynamics using Graph Neural Networks

   Chen, N; Wang, K; Johnson_III, W; Kramer-Bottiglio, R; Bekris, Kostas E; Aanjaneya, M (November 2024, Conference on Robot Learning (CoRL))

   Tensegrity robots are composed of rigid struts and flexible cables. They constitute an emerging class of hybrid rigid-soft robotic systems and are promising systems for a wide array of applications, ranging from locomotion to assembly. They are difficult to control and model accurately, however, due to their compliance and high number of degrees of freedom. To address this issue, prior work has introduced a differentiable physics engine designed for tensegrity robots based on first principles. In contrast, this work proposes the use of graph neural networks to model contact dynamics over a graph representation of tensegrity robots, which leverages their natural graph-like cable connectivity between end caps of rigid rods. This learned simulator can accurately model 3-bar and 6-bar tensegrity robot dynamics in simulation-to-simulation experiments where MuJoCo is used as the ground truth. It can also achieve higher accuracy than the previous differentiable engine for a real 3-bar tensegrity robot, for which the robot state is only partially observable. When compared against direct applications of recent mesh-based graph neural network simulators, the proposed approach is computationally more efficient, both for training and inference, while achieving higher accuracy. Code and data are available at https://github.com/nchen9191/tensegrity_gnn_simulator_public 

   more » « less
3. Learning Differentiable Tensegrity Dynamics using Graph Neural Networks

   Chen, N; Wang, K; Johnson_III, W; Kramer-Bottiglio, R; Bekris, Kostas E; Aanjaneya, M (November 2024, Conference on Robot Learning (CoRL))

   Tensegrity robots are composed of rigid struts and flexible cables. They constitute an emerging class of hybrid rigid-soft robotic systems and are promising systems for a wide array of applications, ranging from locomotion to assembly. They are difficult to control and model accurately, however, due to their compliance and high number of degrees of freedom. To address this issue, prior work has introduced a differentiable physics engine designed for tensegrity robots based on first principles. In contrast, this work proposes the use of graph neural networks to model contact dynamics over a graph representation of tensegrity robots, which leverages their natural graph-like cable connectivity between end caps of rigid rods. This learned simulator can accurately model 3-bar and 6-bar tensegrity robot dynamics in simulation-to-simulation experiments where MuJoCo is used as the ground truth. It can also achieve higher accuracy than the previous differentiable engine for a real 3-bar tensegrity robot, for which the robot state is only partially observable. When compared against direct applications of recent mesh-based graph neural network simulators, the proposed approach is computationally more efficient, both for training and inference, while achieving higher accuracy. Code and data are available at https://github.com/nchen9191/tensegrity_gnn_simulator_public 

   more » « less
4. Robots that Can Survive the Egg Drop

   https://doi.org/10.3389/frym.2024.1452937  

   Johnson, William R; Kramer-Bottiglio, Rebecca (November 2024, Frontiers for Young Minds)

   If you ever did the egg drop challenge, you know it is hard to build something that can protect a fragile egg from crashing into the ground and breaking. Engineers are building soft robots called tensegrity robots, which are designed to survive harsh crashes. The word tensegrity comes from “tension” and “integrity”. It means the robot is made of stiff bars held together with stretchy cables. This flexible structure helps a tensegrity robot absorb the impact from crashes. Someday, these robots might be used to explore dangerous places like deep caves or other planets. These robots could fall off cliffs or into craters. Right now, engineers are making tensegrity robots better and easier to control. In this article, we will explain how tensegrity robots work. We will discuss their advantages, their disadvantages, and what they can be used for. 

   more » « less
5. A Shape-Changing Wheeling and Jumping Robot Using Tensegrity Wheels and Bistable Mechanism

   https://doi.org/10.1109/TMECH.2023.3276933  

   Spiegel, Sydney; Sun, Jiefeng; Zhao, Jianguo (August 2023, IEEE/ASME Transactions on Mechatronics)

   Tensegrity structures made from rigid rods and elastic cables have unique characteristics, such as being lightweight, easy to fabricate, and high load-carrying to weight capacity. In this article, we leverage tensegrity structures as wheels for a mobile robot that can actively change its shape by expanding or collapsing the wheels. Besides the shape-changing capability, using tensegrity as wheels offers several advantages over traditional wheels of similar sizes, such as a shock-absorbing capability without added mass since tensegrity wheels are both lightweight and highly compliant. We show that a robot with two icosahedron tensegrity wheels can reduce its width from 400 to 180 mm, and simultaneously, increase its height from 75 to 95 mm by changing the expanded tensegrity wheels to collapsed disk-like ones. The tensegrity wheels enable the robot to overcome steps with heights up to 110 and 150 mm with the expanded and collapsed configuration, respectively. We establish design guidelines for robots with tensegrity wheels by analyzing the maximum step height that can be overcome by the robot and the force required to collapse the wheel. The robot can also jump onto obstacles up to 300-mm high with a bistable mechanism that can gradually store but quickly release energy. We demonstrate the robot's locomotion capability in indoor and outdoor environments, including various natural terrains, like sand, grass, rocks, ice, and snow. Our results suggest that using tensegrity structures as wheels for mobile robots can enhance their capability to overcome obstacles, traverse challenging terrains, and survive falls from heights. When combined with other locomotion modes (e.g., jumping), such shape-changing robots can have broad applications for search-and-rescue after disasters or surveillance and monitoring in unstructured environments. 

   more » « less