arxiv: 2605.12654 · v1 · submitted 2026-05-12 · 💻 cs.RO

Recognition: unknown

COSMIC: Concurrent Optimization of Structure, Material, and Integrated Control for robotic systems

Qinsong Guo , Liwei Wang

Authors on Pith no claims yet

Pith reviewed 2026-05-14 20:37 UTC · model grok-4.3

classification 💻 cs.RO

keywords concurrent optimizationtruss-lattice robotdifferentiable simulationlocomotion designco-designneural controllertopology optimization

0 comments

The pith

A gradient-based framework co-optimizes robot topology, material distribution, and control policy to discover superior locomotion.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper establishes that simultaneously optimizing the structure, materials, and control of truss-lattice robots through gradients yields locomotion strategies that outperform those found by optimizing each element in isolation. This matters because most current robots design these components separately, in contrast to the integrated evolution seen in nature, often producing suboptimal results. The approach converts mixed topological and material choices into a continuous space, pairs them with a neural controller inside a differentiable simulator, and applies constrained optimization to explore the resulting non-convex landscape. Case studies confirm the method finds diverse effective gaits while also surfacing insights on how the three design elements interact to shape performance.

Core claim

The framework simultaneously optimizes the topology, material distribution, and control policy of a truss-lattice robot by embedding mixed-type variables into a continuous design space, integrating a neural network controller within a differentiable simulator, and applying constrained optimization to navigate the non-convex landscape, consistently discovering diverse locomotion strategies that outperform baselines obtained through separated design.

What carries the argument

Gradient-based concurrent optimization that embeds mixed topological and material variables continuously and couples them to a neural controller inside a differentiable simulator.

If this is right

The method produces diverse locomotion strategies that exceed performance from separate optimization of structure, material, and control.
The same framework adapts to varied functional requirements and boundary conditions.
Extracted insights reveal both individual and collective contributions of topology, material, and control to overall robot performance.
The approach supplies a computational route toward autonomous co-design of robots capable of reconfiguration and complex behaviors.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

If the simulator matches physical behavior closely enough, the method could directly generate printable robot designs without extensive manual iteration.
The co-design loop might extend to other robot morphologies and tasks where structure and control are strongly coupled.
Design insights from the framework could inspire new manual heuristics that account for interactions among the three entities.

Load-bearing premise

The differentiable simulator must accurately capture the coupled physical effects of topology, material distribution, and control without large gaps from real-world dynamics or fabrication limits.

What would settle it

Building and testing a physical prototype of one optimized design and observing locomotion performance that falls short of or fails to match the simulated results would falsify the practical value of the co-design.

Figures

Figures reproduced from arXiv: 2605.12654 by Liwei Wang, Qinsong Guo.

**Figure 2.** Figure 2: Design representation: (a) Each edge Eij is associated with a continuous, relaxed design vector zij . (b) Through a softmax projection scaled by β, (c) we project it to a one-hot-encoded material state vector ˜zij , that dictates the mix ratio between discrete material types. (d) The Multi-Layer Perceptron controller receives synchronized clock signals from a Central Pattern Generator, the instantaneous no… view at source ↗

**Figure 3.** Figure 3: Illustration of the differentiable simulation. At integration time [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗

**Figure 4.** Figure 4: Penalty fields of (a) the global binarization constraint, (b) the pairwise orthogonal constraint, and (c) the [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗

**Figure 5.** Figure 5: Optimization history of the co-design process corresponding to the design in Figure. 6b. (a–d) Intermediate [PITH_FULL_IMAGE:figures/full_fig_p009_5.png] view at source ↗

**Figure 6.** Figure 6: The baseline and optimization results with varying initial designs. (a) Vertical actuation layout baseline [PITH_FULL_IMAGE:figures/full_fig_p012_6.png] view at source ↗

**Figure 7.** Figure 7: Combinatorial ablation study results. Robots are optimized from an identical initialization. A [PITH_FULL_IMAGE:figures/full_fig_p013_7.png] view at source ↗

**Figure 8.** Figure 8: Quatitative ablation results. Grouped by design entities, the average with and without performance for each [PITH_FULL_IMAGE:figures/full_fig_p014_8.png] view at source ↗

**Figure 9.** Figure 9: Alternative tasks and environments: (a) upward incline; (b) [PITH_FULL_IMAGE:figures/full_fig_p015_9.png] view at source ↗

read the original abstract

Replicating and surpassing the autonomy of natural organisms remains a long-standing goal in robotics. Yet most robotic systems have their structure, materials, and control designed separately, in sharp contrast to the co-evolution in nature. This separation often leads to suboptimal designs, and we still have a limited understanding of the individual and collective contributions of these design entities. In this work, we propose a gradient-based co-design framework that simultaneously optimizes the topology, material distribution, and control policy of a truss-lattice robot. The framework embeds mixed-type topological and material variables into a continuous design space and integrates a neural network controller within a differentiable simulator, capturing their interactions and enabling efficient gradient calculation via automatic differentiation. Furthermore, we develop a constrained optimization to navigate the highly non-convex design landscape and jointly optimize all design entities. Case studies demonstrate that the proposed framework consistently discovers diverse locomotion strategies that outperform baselines obtained through separated design. The framework is also flexible to accommodate different functional requirements and boundary conditions. Using this framework, we further extract design insights that reveal the individual and collective effects of different entities on robotic performance. The proposed framework provides a computational foundation for the autonomous co-design of robotic systems, capable of reconfiguration, locomotion, and other complex autonomous behaviors.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

COSMIC puts topology, material, and neural control into one differentiable loop for truss robots and shows joint optimization beating separate baselines in simulation.

read the letter

The paper's main move is to relax mixed topological and material variables into a continuous space, then back-propagate through a neural controller inside a differentiable simulator so that all three can be optimized together with gradients. They add a constrained optimizer to handle the non-convex landscape and run a few locomotion case studies on truss-lattice robots. The joint designs find varied gaits that outperform the sequential baselines they compare against, and they pull out some observations on how the design variables interact. That integration is the part that feels new relative to the separated-design papers they cite. The framework also looks reasonably flexible for changing boundary conditions or objectives. The obvious limitation is that everything stays in simulation. There is no hardware fabrication, no actuator characterization, and no direct check on whether the contact or buckling models match real truss behavior. If those models have systematic error, the reported margins and the extracted insights could shrink. The abstract does not give quantitative metrics or ablation details, so the strength of the outperformance claim is still hard to judge without the full numbers. This is for people already working on gradient-based co-design or lattice robots who want a concrete implementation to build on. A reader looking for a practical starting point on concurrent optimization would get something usable here. I would send it to peer review. The core idea is timely and the code path looks reproducible enough that referees can check the simulator assumptions and the exact metrics.

Referee Report

2 major / 2 minor

Summary. The manuscript presents COSMIC, a gradient-based co-design framework for truss-lattice robots that simultaneously optimizes topology, material distribution, and neural-network control policies. Design variables are relaxed to a continuous space inside a differentiable simulator; automatic differentiation supplies gradients, and a constrained optimizer navigates the resulting non-convex landscape. Case studies on locomotion tasks report that the co-optimized designs discover diverse gaits and outperform baselines obtained by sequential structure-then-control optimization; the framework is also shown to accommodate varying boundary conditions and to yield interpretable design insights.

Significance. If the reported performance margins prove robust, the work supplies a concrete computational route to integrated robotic design that mirrors biological co-evolution. The use of end-to-end differentiability for mixed discrete-continuous variables is a clear technical strength and could accelerate exploration of high-dimensional design spaces that remain intractable under separate optimization pipelines.

major comments (2)

[§4] §4 (Case Studies) and associated tables/figures: the central claim that co-optimized designs “consistently outperform” separated-design baselines rests on simulation results whose quantitative metrics, error bars, ablation controls, and post-processing discretization steps are not reported; without these the magnitude and reliability of the claimed advantage cannot be assessed.
[§3.2] §3.2 (Differentiable Simulator): the forward model relaxes topology and material variables to continuous parameters and back-propagates through contact and actuation; no sensitivity analysis, friction-model validation, or comparison against measured truss-lattice dynamics is supplied, yet every performance margin and design insight depends on the fidelity of this model.

minor comments (2)

[§3.3] The description of the constrained optimizer (Eq. (X) in §3.3) would benefit from an explicit statement of how the penalty or barrier parameters are scheduled across iterations.
[Figures] Figure captions for locomotion trajectories should include the numerical values of the final optimized material and topology parameters for direct reproducibility.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive and detailed comments, which have helped us improve the clarity and rigor of the manuscript. We address each major comment point by point below and have revised the paper accordingly.

read point-by-point responses

Referee: [§4] §4 (Case Studies) and associated tables/figures: the central claim that co-optimized designs “consistently outperform” separated-design baselines rests on simulation results whose quantitative metrics, error bars, ablation controls, and post-processing discretization steps are not reported; without these the magnitude and reliability of the claimed advantage cannot be assessed.

Authors: We agree that the original presentation of results lacked sufficient statistical detail. In the revised manuscript we now report mean performance metrics with standard-error bars computed over 10 independent optimization runs with different random seeds. We have added an ablation study that isolates the contribution of concurrent optimization versus sequential structure-then-control pipelines, including paired t-tests for statistical significance. The post-processing discretization procedure (thresholding of relaxed density and material variables followed by connectivity filtering) is now described in detail in §4.1 with pseudocode and sensitivity checks on the threshold value. revision: yes
Referee: [§3.2] §3.2 (Differentiable Simulator): the forward model relaxes topology and material variables to continuous parameters and back-propagates through contact and actuation; no sensitivity analysis, friction-model validation, or comparison against measured truss-lattice dynamics is supplied, yet every performance margin and design insight depends on the fidelity of this model.

Authors: We acknowledge the importance of simulator fidelity. The revised §3.2 now contains a sensitivity analysis on friction coefficient, contact stiffness, and actuation bandwidth, showing that the discovered gaits remain qualitatively consistent across a ±20 % parameter range. We have added a validation subsection that compares simulated truss-lattice deformation under gravity and periodic forcing against published experimental data from the literature on similar lattice structures. Direct physical experiments with fabricated prototypes lie outside the current simulation-centric scope; we explicitly note this limitation and outline it as future work. revision: partial

Circularity Check

0 steps flagged

No significant circularity in co-optimization framework

full rationale

The paper defines a differentiable simulator embedding continuous relaxations of topology/material variables plus a neural controller, then applies gradient descent and constrained optimization to maximize a performance objective. Case-study results compare this joint optimization against separated-design baselines run inside the identical simulator; the outperformance claim is therefore a direct numerical comparison of two optimization procedures rather than a self-referential fit or renamed input. No load-bearing self-citations, uniqueness theorems, or ansatzes imported from prior author work appear in the derivation. The framework is self-contained against its own simulation benchmark.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The framework rests on the assumption that a differentiable physics simulator can faithfully represent the coupled effects of discrete topology choices, continuous material fields, and neural control; no free parameters or new entities are explicitly named in the abstract.

axioms (1)

domain assumption Physical interactions among structure, material, and control can be accurately captured inside a differentiable simulator
This assumption enables the gradient flow that drives the entire co-optimization.

pith-pipeline@v0.9.0 · 5518 in / 1154 out tokens · 42050 ms · 2026-05-14T20:37:00.026986+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

58 extracted references · 6 canonical work pages · 1 internal anchor

[1]

Katherine Fu, Diana Moreno, Maria Yang, and Kristin L. Wood. Bio-Inspired Design: An Overview Investigating Open Questions From the Broader Field of Design-by-Analogy.Journal of Mechanical Design, 136(111102), October 2014

2014
[2]

Morphological design methodolo- gies of soft robots.The Innovation Informatics, 1(1):100012, 2025

Beijia Zhang, Jie Shen, Zhenyu He, Qingyue Wu, Jing Xiong, and Zeyang Xia. Morphological design methodolo- gies of soft robots.The Innovation Informatics, 1(1):100012, 2025

2025
[3]

Truby, John William Boley, Timothy J

Arda Kotikian, Ryan L. Truby, John William Boley, Timothy J. White, and Jennifer A. Lewis. 3D Printing of Liquid Crystal Elastomeric Actuators with Spatially Programed Nematic Order.Advanced Materials, 30(10):1706164, 2018

2018
[4]

A review on reprogrammable bistable structures.Smart materials and structures, 33(9):093001, 2024

Chao Zhang, Xiaofeng Yin, Rongchu Chen, Kaiwen Ju, Yinxuan Hao, Tong Wu, Jian Sun, Huxiao Yang, and Yan Xu. A review on reprogrammable bistable structures.Smart materials and structures, 33(9):093001, 2024

2024
[5]

Webster-Wood, and Lining Yao

Jianzhe Gu, Ziwen Ye, Tucker Rae-Grant, Shuhong Wang, Ding Zhao, Josiah Hester, Victoria A. Webster-Wood, and Lining Yao. Optimization and control of actuator networks in variable geometry truss systems using genetic algorithms.Nature Communications, 16(1):8432, September 2025. 16 COSMICA PREPRINT

2025
[6]

A review on locomotion robophysics: the study of movement at the intersection of robotics, soft matter and dynamical systems.Reports on Progress in Physics, 79(11):110001, sep 2016

Jeffrey Aguilar, Tingnan Zhang, Feifei Qian, Mark Kingsbury, Benjamin McInroe, Nicole Mazouchova, Chen Li, Ryan Maladen, Chaohui Gong, Matt Travers, Ross L Hatton, Howie Choset, Paul B Umbanhowar, and Daniel I Goldman. A review on locomotion robophysics: the study of movement at the intersection of robotics, soft matter and dynamical systems.Reports on Pr...

2016
[7]

Robots that evolve on demand.Nature Reviews Materials, 9(11):822–835, November 2024

Robert Baines, Frank Fish, Josh Bongard, and Rebecca Kramer-Bottiglio. Robots that evolve on demand.Nature Reviews Materials, 9(11):822–835, November 2024

2024
[8]

Macrae Montgomery, Masato Tanaka, Liang Yue, Changyoung Yuhn, Yuki Sato, Atsushi Kawamoto, H

Hiroki Kobayashi, Farzad Gholami, S. Macrae Montgomery, Masato Tanaka, Liang Yue, Changyoung Yuhn, Yuki Sato, Atsushi Kawamoto, H. Jerry Qi, and Tsuyoshi Nomura. Computational synthesis of locomotive soft robots by topology optimization.Science Advances, 10(30):eadn6129, July 2024

2024
[9]

Eiben, Danielle Frances Kennedy, Jean-Baptiste Mouret, Philip Valencia, and Dave Winkler

David Howard, Agoston E. Eiben, Danielle Frances Kennedy, Jean-Baptiste Mouret, Philip Valencia, and Dave Winkler. Evolving embodied intelligence from materials to machines.Nature Machine Intelligence, 1(1):12–19, January 2019

2019
[10]

Interactive robogami: An end-to-end system for design of robots with ground locomotion.The International Journal of Robotics Research, 36(10):1131–1147, September 2017

Adriana Schulz, Cynthia Sung, Andrew Spielberg, Wei Zhao, Robin Cheng, Eitan Grinspun, Daniela Rus, and Wojciech Matusik. Interactive robogami: An end-to-end system for design of robots with ground locomotion.The International Journal of Robotics Research, 36(10):1131–1147, September 2017

2017
[11]

A linking invariant for truss robot motion planning.IEEE Robotics and Automation Letters, 7(2):1424–1430, 2022

Alexander Spinos and Mark Yim. A linking invariant for truss robot motion planning.IEEE Robotics and Automation Letters, 7(2):1424–1430, 2022

2022
[12]

Yuntao Guan, Zheming Zhuang, Ze Zhang, and Jian S. Dai. Design, Analysis, and Experiment of the Origami Robot Based on Spherical-Linkage Parallel Mechanism.Journal of Mechanical Design, 145(081701), May 2023

2023
[13]

Co-Design of the Morphology and Actuation of Soft Robots for Locomotion

Merel van Diepen and Kristina Shea. Co-Design of the Morphology and Actuation of Soft Robots for Locomotion. Journal of Mechanical Design, 144(083305), June 2022

2022
[14]

Bertossa

Rinaldo C. Bertossa. Morphology and behaviour: functional links in development and evolution.Philosophical Transactions of the Royal Society B: Biological Sciences, 366(1574):2056–2068, July 2011

2056
[15]

Efficient automatic design of robots.Proceedings of the National Academy of Sciences, 120(41):e2305180120, October 2023

David Matthews, Andrew Spielberg, Daniela Rus, Sam Kriegman, and Josh Bongard. Efficient automatic design of robots.Proceedings of the National Academy of Sciences, 120(41):e2305180120, October 2023

2023
[16]

Generating freeform endoskeletal robots.arXiv preprint arXiv:2412.01036, 2024

Muhan Li, Lingji Kong, and Sam Kriegman. Generating freeform endoskeletal robots.arXiv preprint arXiv:2412.01036, 2024

work page arXiv 2024
[17]

Soft robots modeling: A structured overview.IEEE Transactions on Robotics, 39(3):1728–1748, 2023

Costanza Armanini, Frédéric Boyer, Anup Teejo Mathew, Christian Duriez, and Federico Renda. Soft robots modeling: A structured overview.IEEE Transactions on Robotics, 39(3):1728–1748, 2023

2023
[18]

Prabhat Kumar, Anupam Saxena, and Roger A. Sauer. Computational Synthesis of Large Deformation Compliant Mechanisms Undergoing Self and Mutual Contact.Journal of Mechanical Design, 141(012302), October 2018

2018
[19]

A scalable pipeline for designing reconfigurable organisms.Proceedings of the National Academy of Sciences, 117(4):1853–1859, January 2020

Sam Kriegman, Douglas Blackiston, Michael Levin, and Josh Bongard. A scalable pipeline for designing reconfigurable organisms.Proceedings of the National Academy of Sciences, 117(4):1853–1859, January 2020

2020
[20]

Karacakol, Yunus Alapan, Sinan O

Alp C. Karacakol, Yunus Alapan, Sinan O. Demir, and Metin Sitti. Data-driven design of shape-programmable magnetic soft materials.Nature Communications, 16(1):2946, March 2025

2025
[21]

Accelerated co-design of robots through morphological pretraining.arXiv preprint arXiv:2502.10862, 2025

Luke Strgar and Sam Kriegman. Accelerated co-design of robots through morphological pretraining.arXiv preprint arXiv:2502.10862, 2025

work page arXiv 2025
[22]

Evolution and learning in differentiable robots

Luke Strgar, David Matthews, Tyler Hummer, and Sam Kriegman. Evolution and learning in differentiable robots. InProceedings of Robotics: Science and Systems, Delft, Netherlands, July 2024

2024
[23]

Alacoque, and Kai A

Anurag Bhattacharyya, Jinyoung Kim, Lee R. Alacoque, and Kai A. James. Design Synthesis of a 4D-Printed Self-Tying Knot With Programmable Morphology.Journal of Mechanical Design, 146(063303), December 2023

2023
[24]

Charles Schaff, Audrey Sedal, and Matthew R. Walter. Soft robots learn to crawl: Jointly optimizing design and control with sim-to-real transfer. InProceedings of Robotics: Science and Systems, June 2022

2022
[25]

DiffuseBot: Breeding Soft Robots With Physics-Augmented Generative Diffusion Models.Advances in Neural Information Processing Systems, 36:44398–44423, December 2023

Tsun-Hsuan Johnson Wang, Juntian Zheng, Pingchuan Ma, Yilun Du, Byungchul Kim, Andrew Spielberg, Josh Tenenbaum, Chuang Gan, and Daniela Rus. DiffuseBot: Breeding Soft Robots With Physics-Augmented Generative Diffusion Models.Advances in Neural Information Processing Systems, 36:44398–44423, December 2023

2023
[26]

Evenchik, Jared M

Liwei Wang, Alexander L. Evenchik, Jared M. Yang, Ryan L. Truby, and Wei Chen. Autonomous codesign and fabrication of multistimuli-responsive material systems.Science Advances, 11(37):eadx4409, September 2025

2025
[27]

Adams, Vincent Tournat, and Ka- tia Bertoldi

Giovanni Bordiga, Eder Medina, Sina Jafarzadeh, Cyrill Bösch, Ryan P. Adams, Vincent Tournat, and Ka- tia Bertoldi. Automated discovery of reprogrammable nonlinear dynamic metamaterials.Nature Materials, 23(11):1486–1494, November 2024. 17 COSMICA PREPRINT

2024
[28]

O’Hern, Rebecca Kramer-Bottiglio, and Josh Bongard

Atoosa Parsa, Corey S. O’Hern, Rebecca Kramer-Bottiglio, and Josh Bongard. Gradient-based design of computational granular crystals.arXiv preprint arXiv:2404.04825, 2024

work page arXiv 2024
[29]

The material point method for simulating continuum materials

Chenfanfu Jiang, Craig Schroeder, Joseph Teran, Alexey Stomakhin, and Andrew Selle. The material point method for simulating continuum materials. InACM SIGGRAPH 2016 Courses, SIGGRAPH ’16, New York, NY , USA, 2016. Association for Computing Machinery

2016
[30]

Tenenbaum, Daniela Rus, and Chuang Gan

Tsun-Hsuan Wang, Pingchuan Ma, Andrew Everett Spielberg, Zhou Xian, Hao Zhang, Joshua B. Tenenbaum, Daniela Rus, and Chuang Gan. Softzoo: A soft robot co-design benchmark for locomotion in diverse environments. arXiv preprint arXiv:2303.09555, 2023

work page arXiv 2023
[31]

Chandler, and Pietro Valdastri

Joshua Davy, Peter Lloyd, James H. Chandler, and Pietro Valdastri. A framework for simulation of magnetic soft robots using the material point method.IEEE Robotics and Automation Letters, 8(6):3470–3477, 2023

2023
[32]

Advanced soft robot modeling in ChainQueen.Robotica, 41(1):74–104, January 2023

Andrew Spielberg, Tao Du, Yuanming Hu, Daniela Rus, and Wojciech Matusik. Advanced soft robot modeling in ChainQueen.Robotica, 41(1):74–104, January 2023

2023
[33]

A review of differentiable simulators.IEEE Access, 12:97581–97604, 2024

Rhys Newbury, Jack Collins, Kerry He, Jiahe Pan, Ingmar Posner, David Howard, and Akansel Cosgun. A review of differentiable simulators.IEEE Access, 12:97581–97604, 2024

2024
[34]

Tianju Xue, Shuheng Liao, Zhengtao Gan, Chanwook Park, Xiaoyu Xie, Wing Kam Liu, and Jian Cao. JAX-FEM: A differentiable GPU-accelerated 3D finite element solver for automatic inverse design and mechanistic data science.Computer Physics Communications, 291:108802, October 2023

2023
[35]

arXiv preprint arXiv:1910.00935 , year=

Yuanming Hu, Luke Anderson, Tzu-Mao Li, Qi Sun, Nathan Carr, Jonathan Ragan-Kelley, and Frédo Durand. Difftaichi: Differentiable programming for physical simulation.arXiv preprint arXiv:1910.00935, 2020

work page arXiv 1910
[36]

Diffpd: Differentiable projective dynamics.ACM Trans

Tao Du, Kui Wu, Pingchuan Ma, Sebastien Wah, Andrew Spielberg, Daniela Rus, and Wojciech Matusik. Diffpd: Differentiable projective dynamics.ACM Trans. Graph., 41(2), November 2021

2021
[37]

FEniTop: a simple FEniCSx implementation for 2D and 3D topology optimization supporting parallel computing.Structural and Multidisciplinary Optimization, 67(8):140, August 2024

Yingqi Jia, Chao Wang, and Xiaojia Shelly Zhang. FEniTop: a simple FEniCSx implementation for 2D and 3D topology optimization supporting parallel computing.Structural and Multidisciplinary Optimization, 67(8):140, August 2024

2024
[38]

Computational co-design of structure and feedback controller for locomoting soft robots.Structural and Multidisciplinary Optimization, 68(9):185, September 2025

Yuki Sato, Changyoung Yuhn, Hiroki Kobayashi, Atsushi Kawamoto, and Tsuyoshi Nomura. Computational co-design of structure and feedback controller for locomoting soft robots.Structural and Multidisciplinary Optimization, 68(9):185, September 2025

2025
[39]

4d topology optimization: Integrated optimization of the structure and self-actuation of soft bodies for dynamic motions

Changyoung Yuhn, Yuki Sato, Hiroki Kobayashi, Atsushi Kawamoto, and Tsuyoshi Nomura. 4d topology optimization: Integrated optimization of the structure and self-actuation of soft bodies for dynamic motions. Computer Methods in Applied Mechanics and Engineering, 414:116187, 2023

2023
[40]

Zhang, Kui Wu, Andrew Spielberg, Robert K

Pingchuan Ma, Tao Du, John Z. Zhang, Kui Wu, Andrew Spielberg, Robert K. Katzschmann, and Wojciech Matusik. DiffAqua: a differentiable computational design pipeline for soft underwater swimmers with shape interpolation.ACM Trans. Graph., 40(4):132:1–132:14, July 2021

2021
[41]

Learning-In- The-Loop Optimization: End-To-End Control And Co-Design Of Soft Robots Through Learned Deep Latent Representations

Andrew Spielberg, Allan Zhao, Yuanming Hu, Tao Du, Wojciech Matusik, and Daniela Rus. Learning-In- The-Loop Optimization: End-To-End Control And Co-Design Of Soft Robots Through Learned Deep Latent Representations. InAdvances in Neural Information Processing Systems, volume 32. Curran Associates, Inc., 2019

2019
[42]

Liwei Wang. Co-design of magnetic soft robots with large deformation and contacts via material point method and topology optimization.Computer Methods in Applied Mechanics and Engineering, 445:118205, October 2025

2025
[43]

Lumpe and Kristina Shea

Thomas S. Lumpe and Kristina Shea. Computational Design of Multi-State Lattice Structures With Finite Mechanisms for Shape Morphing.Journal of Mechanical Design, 145(071701), March 2023

2023
[44]

A versatile co-design approach for dynamic legged robots

Traiko Dinev, Carlos Mastalli, Vladimir Ivan, Steve Tonneau, and Sethu Vijayakumar. A versatile co-design approach for dynamic legged robots. In2022 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pages 5562–5569. IEEE, 2022

2022
[45]

Multi-legged robots: progress and challenges.National Science Review, 10(5):nwac214, May 2023

Yue Gao, Bo Su, Lei Jiang, and Feng Gao. Multi-legged robots: progress and challenges.National Science Review, 10(5):nwac214, May 2023

2023
[46]

Design and control of compliant tensegrity robots through simulation and hardware validation

Ken Caluwaerts, Jérémie Despraz, Atıl I¸ sçen, Andrew P Sabelhaus, Jonathan Bruce, Benjamin Schrauwen, and Vytas SunSpiral. Design and control of compliant tensegrity robots through simulation and hardware validation. Journal of the royal society interface, 11(98), 2014

2014
[47]

Shah, Joran W

Dylan S. Shah, Joran W. Booth, Robert L. Baines, Kun Wang, Massimo Vespignani, Kostas Bekris, and Rebecca Kramer-Bottiglio. Tensegrity Robotics.Soft Robotics, 9(4):639–656, August 2022

2022
[48]

Biohybrid bipedal robot powered by skeletal muscle tissue.Matter, 7(3):1026–1041, 2024

Ryotaro Kinjo, Yosuke Shimizu, Yuya Morimoto, and Shoji Takeuchi. Biohybrid bipedal robot powered by skeletal muscle tissue.Matter, 7(3):1026–1041, 2024. 18 COSMICA PREPRINT

2024
[49]

Biohybrid robots: recent progress, challenges, and perspectives

Victoria A Webster-Wood, Maria Guix, Nicole W Xu, Bahareh Behkam, Hirotaka Sato, Deblina Sarkar, Samuel Sanchez, Masahiro Shimizu, and Kevin Kit Parker. Biohybrid robots: recent progress, challenges, and perspectives. Bioinspiration & Biomimetics, 18(1):015001, November 2022

2022
[50]

Topology optimization of hard-magnetic soft materials.Journal of the Mechanics and Physics of Solids, 158:104628, January 2022

Zhi Zhao and Xiaojia Shelly Zhang. Topology optimization of hard-magnetic soft materials.Journal of the Mechanics and Physics of Solids, 158:104628, January 2022

2022
[51]

Central pattern generators for locomotion control in animals and robots: a review.Neural networks, 21(4):642–653, 2008

Auke Jan Ijspeert. Central pattern generators for locomotion control in animals and robots: a review.Neural networks, 21(4):642–653, 2008

2008
[52]

M. P. Bendsøe. Optimal shape design as a material distribution problem.Structural optimization, 1(4):193–202, December 1989

1989
[53]

Elsevier, 2000

Michel Fortin and Roland Glowinski.Augmented Lagrangian methods: applications to the numerical solution of boundary-value problems, volume 15. Elsevier, 2000

2000
[54]

Estimating or Propagating Gradients Through Stochastic Neurons for Conditional Computation

Yoshua Bengio, Nicholas Léonard, and Aaron Courville. Estimating or propagating gradients through stochastic neurons for conditional computation.arXiv preprint arXiv:1308.3432, 2013

work page internal anchor Pith review Pith/arXiv arXiv 2013
[55]

Roth, Raymond L

Frank L. Roth, Raymond L. Driscoll, and William L. Holt. Frictional Properties of Rubber.Rubber Chemistry and Technology, 16(1):155–177, March 1943

1943
[56]

Biohybrid robotics with living cell actuation.Chemical Society Reviews, 49(12):4043–4069, 2020

Lingyu Sun, Yunru Yu, Zhuoyue Chen, Feika Bian, Fangfu Ye, Lingyun Sun, and Yuanjin Zhao. Biohybrid robotics with living cell actuation.Chemical Society Reviews, 49(12):4043–4069, 2020

2020
[57]

Physical reservoir computing with emerging electronics.Nature Electronics, 7(3):193–206, 2024

Xiangpeng Liang, Jianshi Tang, Yanan Zhong, Bin Gao, He Qian, and Huaqiang Wu. Physical reservoir computing with emerging electronics.Nature Electronics, 7(3):193–206, 2024

2024
[58]

Mechanical neural networks: Architected materials that learn behaviors.Science Robotics, 7(71):eabq7278, 2022

Ryan H Lee, Erwin AB Mulder, and Jonathan B Hopkins. Mechanical neural networks: Architected materials that learn behaviors.Science Robotics, 7(71):eabq7278, 2022. 19

2022