This project is a collaboration with research groups at Penn (Mark Yim, Daniel Koditschek, Douglas Jerolmack) and at USC (Feifei Qian). The aim for this project is to develop methods for teams of robots to jointly overcome environmental hazards on the Moon by attaching to each other to form larger and more stable, maneuverable structures. The robots will use their interactions with the ground to form a map of safe and risky terrain, attach to each other as support when the ground traversal risk is high, move in a coordinated fashion once joined, and, once the maneuver has been successfully completed, separate to continue their original individual missions.
The TRUSSES project will develop algorithms that provide robots with two main capabilities: 1) estimate robot-to-regolith interactions in order to plan safe maneuvers, and 2) plan truss formations and coordinated motions for robots to push and pull each other to safe locations. The system will be evaluated to verify risk estimation, risk avoidance, risk mitigation, and recovery from failure.
Our group primarily develops planning and localization methods for coordinated, risk-aware maneuvers in heterogeneous teams of ground robots operating with a partially known map. This map is built proprioceptively by an exploring quadruped and encodes spatial parameters for a robot-ground interaction model, enabling force prediction. It can also be transformed into a risk map that reflects uncertainty (unexplored regions) and poor terrain (where robots may get stuck). Heterogeneity is advantageous, as each robot interacts differently with the terrain, offering unique capabilities in team operations. This diversity requires high-level planning to assign roles aligned with each robot’s strengths, producing team configurations that support the rescue mission while avoiding high-risk zones. Once a target configuration is defined, robots reposition themselves using a reactive navigation planner and form truss connections. A truss-planner then computes joint motions so the connected network can move collectively to safer areas, where the robots can disconnect and resume individual tasks like exploration or transport.
@conference{liu2025scout,
title = {Scout-rover cooperation: Online terrain strength mapping and traversal risk estimation for planetary-analog explorations},
author = {Shipeng Liu and J. Diego Caporale and Ethan Fulcher and Wilson Hu and Natalie Cavallo and Yifeng Zhang and Xingue Liao and Cynthia Sung and Feifei Qian},
year = {2025},
date = {2025-03-11},
booktitle = {Lunar and Planetary Science Conference (LPSC)},
keywords = {},
pubstate = {published},
tppubtype = {conference}
}
@conference{sung2024trusses,
title = {TRUSSES: Temporarily, Robots Unite to Surmount Sandy Entrapments, then Separate},
author = {Douglas Jerolmack and Daniel Koditschek and Feifei Qian and Mark Yim and Cynthia Sung},
year = {2024},
date = {2024-04-23},
booktitle = {Lunar Surface Innovation Consortium Spring Meeting},
keywords = {},
pubstate = {published},
tppubtype = {conference}
}
The work is supported by Lunar Surface Technology Research grant #80NSSC24K0127 from NASA’s Space Technology Research Grants Program. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of this organization.
The emergence of smart actuators with tunable material properties has empowered roboticists to create systems capable of transitioning between rigid and soft operational modes. This adaptability allows robots to operate with high precision and enhanced payload capacity in their rigid state, while offering greater safety and flexibility when soft especially in the presence of unpredictable disturbances. We design these adaptive robotic systems and develop corresponding control algorithms to showcase the versatility and potential of this new class of reconfigurable robots.
Fig 1: Collage of projects in the research thrust: Tunable Stiffness in Soft Robots
Coiled Spring Actuator
We have developed several novel tunable-stiffness actuators as part of this research thrust, one of which is the coiled-spring actuator. It is inspired by the mechanics of nested elastic rings, wherein the effective bulk stiffness can be modulated by varying the number of elastic coiled layers. This mechanism enables near-linear stiffness tuning, achieved through electro-mechanical control. The system allows for precise, programmable stiffness adjustments, and our prototype has demonstrated a tunability range of up to 20-fold. We have also constructed a non-dimensional mechanics model for the coiled spring actuator which extends to all such mechanisms of different dimensions and materials. The local stiffness changes from these actuators induce corresponding deformations in a compliant, segmented manipulator constructed as a tower of multiple such modules.
Pneumatic Dual-Bellows Actuator
Another novel tunable stiffness actuator developed in our lab is able to achieve a stiffness gain of 1.43 times (1332 N/m to 1913 N/m) without needing an external air source or valve. The design consists of an air chamber bellows and spring bellows connected to each other in an air-tight manner. Stiffness modulation in the spring bellows is achieved by altering the volume of the air chamber bellows. Due to large achievable stiffnesses, this actuator is suitable for integration in soft robots that are needed to demonstrate dynamic and adaptable behavior.
Hierarchical Algorithms to Optimize Soft Manipulator Mechanics
Stiffness control algorithms are needed for soft manipulators to be able to take effective advantage of embedded novel stiffness actuators. To address this, we have developed a hierarchical policy for stiffness control for a class of soft segmented manipulators. The stiffness changes induce desired deformations in each segment, thereby influencing the manipulator’s end-effector position. The algorithm can be run as an online controller to influence the manipulator’s stable states or offline as a design algorithm to optimize stiffness distributions.
Coupled Learning in Elastic Networks
We are also collaborating with Lauren Altman and Doug Durian in the Physics Department, School of Arts and Sciences, University of Pennsylvania to build an elastic network for coupled learning. This technique tunes the properties of individual elastic elements in the network to achieve specific outcomes, like applying the right force or strain to an output edge. These mechanical networks could be scaled and automated to tackle more complex tasks, opening the door to a new kind of smart metamaterial.
Related Publications
Misra, Shivangi; Sung, Cynthia
Online Optimization of Soft Manipulator Mechanics via Hierarchical Control (Conference)
7th IEEE-RAS International Conference on Soft Robotics (RoboSoft), 2024.
@conference{misra2024tsmb,
title = {Online Optimization of Soft Manipulator Mechanics via Hierarchical Control},
author = {Shivangi Misra and Cynthia Sung},
url = {https://youtu.be/DS2kEvccwYA?feature=shared
https://repository.upenn.edu/handle/20.500.14332/59590},
doi = {10.1109/RoboSoft60065.2024.10522004},
year = {2024},
date = {2024-04-14},
urldate = {2024-04-14},
booktitle = {7th IEEE-RAS International Conference on Soft Robotics (RoboSoft)},
abstract = {Actively tuning mechanical properties in soft robots is now feasible due to advancements in soft actuation technologies. In soft manipulators, these novel actuators can be distributed over the robot body to allow greater control over its large number of degrees of freedom and to stabilize local deformations against a range of disturbances. In this paper, we present a hierarchical policy for stiffness control for such a class of soft manipulators. The stiffness changes induce desired deformations in each segment, thereby influencing the manipulator’s end-effector position. The algorithm can be run as an online controller to influence the manipulator’s stable states – as we demonstrate in simulation – or offline as a design algorithm to optimize stiffness distributions – as we showcase in a hardware demonstration. Our proposed hierarchical control scheme is agnostic to the stiffness actuation method and can extend to other soft manipulators with nonuniform stiffness distributions.},
keywords = {},
pubstate = {published},
tppubtype = {conference}
}
Actively tuning mechanical properties in soft robots is now feasible due to advancements in soft actuation technologies. In soft manipulators, these novel actuators can be distributed over the robot body to allow greater control over its large number of degrees of freedom and to stabilize local deformations against a range of disturbances. In this paper, we present a hierarchical policy for stiffness control for such a class of soft manipulators. The stiffness changes induce desired deformations in each segment, thereby influencing the manipulator’s end-effector position. The algorithm can be run as an online controller to influence the manipulator’s stable states – as we demonstrate in simulation – or offline as a design algorithm to optimize stiffness distributions – as we showcase in a hardware demonstration. Our proposed hierarchical control scheme is agnostic to the stiffness actuation method and can extend to other soft manipulators with nonuniform stiffness distributions.
@conference{chen2024design,
title = {Design and Characterization of a Pneumatic Tunable-Stiffness Bellows Actuator},
author = {Rongqian Chen and Jun Kwon and Wei-Hsi Chen and Cynthia Sung},
doi = {10.1109/RoboSoft60065.2024.10521916},
year = {2024},
date = {2024-04-13},
urldate = {2024-04-13},
booktitle = {IEEE-RAS International Conference on Soft Robotics (RoboSoft)},
abstract = {We introduce a self-contained pneumatic actuator capable of 1.43 times stiffness gain from 1332 N/m to 1913 N/m without needing an external air source or valve. The design incorporates an air chamber bellows and a spring bellows, connected and sealed. Stiffness modulation is achieved by altering the air chamber volume. We present an approach for computing the volume, pressurized force, and stiffness of a single bellows component, as well as methods for composing single bellows models to predict the change in stiffness of the dual bellows actuator as a function of air chamber compression. We detail the fabrication of the actuator and verify the models on the fabricated prototype. This actuator holds promise for future integration in tunable stiffness robots demanding high strength and adaptability in dynamic scenarios.},
keywords = {},
pubstate = {published},
tppubtype = {conference}
}
We introduce a self-contained pneumatic actuator capable of 1.43 times stiffness gain from 1332 N/m to 1913 N/m without needing an external air source or valve. The design incorporates an air chamber bellows and a spring bellows, connected and sealed. Stiffness modulation is achieved by altering the air chamber volume. We present an approach for computing the volume, pressurized force, and stiffness of a single bellows component, as well as methods for composing single bellows models to predict the change in stiffness of the dual bellows actuator as a function of air chamber compression. We detail the fabrication of the actuator and verify the models on the fabricated prototype. This actuator holds promise for future integration in tunable stiffness robots demanding high strength and adaptability in dynamic scenarios.
@conference{misra2023design,
title = {Design and Control of a Tunable-Stiffness Coiled-Spring Actuator},
author = {Shivangi Misra and Mason Mitchell and Rongqian Chen and Cynthia Sung},
url = {https://repository.upenn.edu/grasp_papers/72/
https://youtu.be/52WVMEeGxUk
},
doi = {10.1109/ICRA48891.2023.10161218},
year = {2023},
date = {2023-05-30},
urldate = {2023-05-30},
booktitle = {IEEE International Conference on Robotics and Automation (ICRA)},
abstract = {We propose a novel design for a lightweight and compact tunable stiffness actuator capable of stiffness changes up to 20x. The design is based on the concept of a coiled spring, where changes in the number of layers in the spring change the bulk stiffness in a near-linear fashion. We present an elastica nested rings model for the deformation of the proposed actuator and empirically verify that the designed stiffness-changing spring abides by this model. Using the resulting model, we design a physical prototype of the tunable-stiffness coiled spring actuator and discuss the effect of design choices on the resulting achievable stiffness range and resolution. In the future, this actuator design could be useful in a wide variety of soft robotics applications, where fast, controllable, and local stiffness change is required over a large range of stiffnesses.},
keywords = {},
pubstate = {published},
tppubtype = {conference}
}
We propose a novel design for a lightweight and compact tunable stiffness actuator capable of stiffness changes up to 20x. The design is based on the concept of a coiled spring, where changes in the number of layers in the spring change the bulk stiffness in a near-linear fashion. We present an elastica nested rings model for the deformation of the proposed actuator and empirically verify that the designed stiffness-changing spring abides by this model. Using the resulting model, we design a physical prototype of the tunable-stiffness coiled spring actuator and discuss the effect of design choices on the resulting achievable stiffness range and resolution. In the future, this actuator design could be useful in a wide variety of soft robotics applications, where fast, controllable, and local stiffness change is required over a large range of stiffnesses.
@conference{misra2022forward,
title = {Forward kinematics and control of a segmented tunable-stiffness 3-D continuum manipulator},
author = {Shivangi Misra and Cynthia Sung},
url = {https://repository.upenn.edu/grasp_papers/69/
https://youtu.be/P-vg3Hiuk4M
https://youtu.be/lc6W61xCWuQ},
doi = {10.1109/ICRA46639.2022.9812098},
year = {2022},
date = {2022-05-23},
urldate = {2022-05-23},
booktitle = {IEEE International Conference on Robotics and Automation (ICRA)},
abstract = {In this work, we consider the problem of controlling the end effector position of a continuum manipulator through local stiffness changes. Continuum manipulators offer the advantage of continuous deformation along their lengths, and recent advances in smart material actuators further enable local compliance changes, which can affect the manipulator's bulk motion. However, leveraging local stiffness change to control motion remains lightly explored. We build a kinematic model of a continuum manipulator as a sequence of segments consisting of symmetrically arranged springs around the perimeter of every segment, and we show that this system has a closed form solution to its forward kinematics. The model includes common constraints such as restriction of torsional or shearing movement. Based on this model, we propose a controller on the spring stiffnesses for a single segment and provide provable guarantees on convergence to a desired goal position. The results are verified in simulation and compared to physical hardware.},
keywords = {},
pubstate = {published},
tppubtype = {conference}
}
In this work, we consider the problem of controlling the end effector position of a continuum manipulator through local stiffness changes. Continuum manipulators offer the advantage of continuous deformation along their lengths, and recent advances in smart material actuators further enable local compliance changes, which can affect the manipulator's bulk motion. However, leveraging local stiffness change to control motion remains lightly explored. We build a kinematic model of a continuum manipulator as a sequence of segments consisting of symmetrically arranged springs around the perimeter of every segment, and we show that this system has a closed form solution to its forward kinematics. The model includes common constraints such as restriction of torsional or shearing movement. Based on this model, we propose a controller on the spring stiffnesses for a single segment and provide provable guarantees on convergence to a desired goal position. The results are verified in simulation and compared to physical hardware.
Support for these projects has been provided by the National Science Foundation (NSF) Grant No. 1845339, the Johnson & Johnson WiSTEM2D Scholars Program, and the Army Research Office (ARO) under MURI award #W911NF1810327. Any opinions, findings, conclusions, or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of funding source.
Jet propulsion is a locomotion mode commonly found in biological swimmers, including cephalopods and tunicates such as squids, cuttlefish, and salps. We are developing a soft salp-inspired robotic system to study mechanisms that produce greater locomotion agility and energetic efficiency.
Salps are barrel-shaped marine invertebrates that swim via jet propulsion. They move forward by rapidly changing the volumes of their body cavity, drawing water into their muscular mantle cavity through the front aperture, and then expelling it under high pressure through the rear funnel. We aim to leverage the unique biomechanics of salps to inform the development of energy-efficient, maneuverable underwater robots capable of environmental sensing in complex marine environments.
Salps can swim either as solitary jet-propelled individuals or while physically connected in a multi-jet colony, commonly known as a “salp chain”. Inspired by salps, we develop the SALP (Salp-inspired Approach to Low-energy Propulsion) robot, a soft underwater robot that swims via jet propulsion similarly to a biological salp.
Version 1: Origami Swimmer
Traditional soft robots use flexible materials for shape change but require complex fabrication. To simplify this, we used an origami-inspired design, which allows the robot to fold from flat sheets into 3D shapes, making it easier to store, transport, and assemble in just a few hours. The robot mimics squid locomotion using the origami magic ball pattern, which transforms between an ellipsoid and a sphere. This enables it to expand and contract, pulling in and expelling water to create a jet for propulsion. A tendon mechanism in its spine controls its length. Ongoing effects focus on evaluating different configurations’ effect (e.g. adding a front nozzle) on the performance of the robot and understanding the underlying dynamics of it.
Mechanism demo
Robot swimming
Version 2: Multi-Robot SALP Robot Platform for Long-Term Distributed Sensing
We have optimized the jetting robot design to enable higher thrust and lower drag. The SALP robot is now modular in the sense that the robots can be manually attached to each other in different physical arrangements to study the effect of multi-robot interactions on locomotion performance. Physically connected SALP chains coordinate their jets to achieve various propulsion modes. We are interested in investigating how physical arrangement and jet coordination between two robots affect the swimming performance and energy efficiency of a two-SALP robotic system. We aim to gain insights into the potential hydrodynamic benefits of multi-jet propulsion by exploring how different coordination strategies influence the surrounding flow environment.
Leveraging Fluid-Structure Interactions for Efficient Control in Geophysical Flows
Micro-vehicles are cost-effective platforms for robotics and automation, excelling in maneuverability and adaptability in diverse environments. However, their lightweight and limited computational capacity pose control challenges. Using our underwater platform, we aim to understand fluid-structure interactions to enhance design and control, resulting in more efficient micro-vehicles with extended lifespans. This effort is a collaborative project focusing on fluid dynamics, control theory, and reconfiguration planning. The project aims to leverage environmental forces for power efficiency, investigating morphological adaptations and passive transport properties. It seeks to synthesize motion control strategies considering inertial effects and fluid-structure interactions while exploring efficiency trade-offs. Ultimately, it aims to enhance micro-autonomous vehicles’ capabilities for long-term operations and future large-scale deploy.
@conference{yang2025salp,
title = {Effect of Jet Coordination on Underwater Propulsion with the Multi-Robot SALP System},
author = {Zhiyuan Yang and Yipeng Zhang and Matthew Herbert and M. Ani Hsieh and Cynthia Sung},
url = {https://www.youtube.com/watch?v=mzd1QCXssCk
https://repository.upenn.edu/handle/20.500.14332/61048},
year = {2025},
date = {2025-04-23},
urldate = {2025-04-23},
booktitle = {8th IEEE-RAS International Conference on Soft Robotics (RoboSoft 2025)},
abstract = {Salps, marine invertebrates known for their collective swimming through coordinated jet propulsion, offer a unique model for efficient underwater movement. Inspired by this biological system, we develop the SALP (Salp-inspired Approach to Low-energy Propulsion) robot, a soft underwater robot that swims via jet propulsion similarly to a biological salp. The SALPs can be physically connected into SALP chains and coordinate their jets to achieve various propulsion modes. In our experiments, we compare the swimming performance of the individual SALP with the two-SALP system, focusing on power, acceleration, velocity, and energy efficiency. Results indicate that two SALPs swimming synchronously exhibit a 9.0% increase in steady-state velocity and a 16.6% improvement in transient acceleration compared to a single SALP. Additionally, our analysis of swimming efficiency implies that asynchronous swimming is potentially more energy efficient than the synchronous mode, as reflected by a decrease in the cost of transport (COT).},
keywords = {},
pubstate = {forthcoming},
tppubtype = {conference}
}
Salps, marine invertebrates known for their collective swimming through coordinated jet propulsion, offer a unique model for efficient underwater movement. Inspired by this biological system, we develop the SALP (Salp-inspired Approach to Low-energy Propulsion) robot, a soft underwater robot that swims via jet propulsion similarly to a biological salp. The SALPs can be physically connected into SALP chains and coordinate their jets to achieve various propulsion modes. In our experiments, we compare the swimming performance of the individual SALP with the two-SALP system, focusing on power, acceleration, velocity, and energy efficiency. Results indicate that two SALPs swimming synchronously exhibit a 9.0% increase in steady-state velocity and a 16.6% improvement in transient acceleration compared to a single SALP. Additionally, our analysis of swimming efficiency implies that asynchronous swimming is potentially more energy efficient than the synchronous mode, as reflected by a decrease in the cost of transport (COT).
Drag coefficient characterization of the origami magic ball (Proceedings Article)
In: ASME International Design Engineering Technical Conferences and Computers and Information in Engineering Conference (IDETC/CIE), pp. DETC2023-117182, 2023.
@inproceedings{chen2023drag,
title = {Drag coefficient characterization of the origami magic ball},
author = {Guanyu Chen and Dongsheng Chen and Jessica Weakly and Cynthia Sung},
url = {https://repository.upenn.edu/grasp_papers/73},
doi = {10.1115/DETC2023-117182},
year = {2023},
date = {2023-08-29},
urldate = {2023-08-29},
booktitle = {ASME International Design Engineering Technical Conferences and Computers and Information in Engineering Conference (IDETC/CIE)},
pages = {DETC2023-117182},
abstract = {The drag coefficient plays a vital role in the design and optimization of robots that move through fluids. From aircraft to underwater vehicles, their geometries are specially engineered so that the drag coefficients are as low as possible to achieve energy-efficient performances. Origami magic balls are 3-dimensional reconfigurable geometries composed of repeated simple waterbomb units. Their volumes can change as their geometries vary and we have used this concept in a recent underwater robot design. This paper characterizes the drag coefficient of an origami magic ball in a wind tunnel. Through dimensional analysis, the scenario where the robot swims underwater is equivalently transferred to the situation when it is in the wind tunnel. With experiments, we have collected and analyzed the drag force data. It is concluded that the drag coefficient of the magic ball increases from around 0.64 to 1.26 as it transforms from a slim ellipsoidal shape to an oblate spherical shape. Additionally, three different magic balls produce increases in the drag coefficient of between 57% and 86% on average compared to the smooth geometries of the same size and aspect ratio. The results will be useful in future designs of robots using waterbomb origami in fluidic environments.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
The drag coefficient plays a vital role in the design and optimization of robots that move through fluids. From aircraft to underwater vehicles, their geometries are specially engineered so that the drag coefficients are as low as possible to achieve energy-efficient performances. Origami magic balls are 3-dimensional reconfigurable geometries composed of repeated simple waterbomb units. Their volumes can change as their geometries vary and we have used this concept in a recent underwater robot design. This paper characterizes the drag coefficient of an origami magic ball in a wind tunnel. Through dimensional analysis, the scenario where the robot swims underwater is equivalently transferred to the situation when it is in the wind tunnel. With experiments, we have collected and analyzed the drag force data. It is concluded that the drag coefficient of the magic ball increases from around 0.64 to 1.26 as it transforms from a slim ellipsoidal shape to an oblate spherical shape. Additionally, three different magic balls produce increases in the drag coefficient of between 57% and 86% on average compared to the smooth geometries of the same size and aspect ratio. The results will be useful in future designs of robots using waterbomb origami in fluidic environments.
@article{yang2021origami,
title = {Origami-inspired robot that swims via jet propulsion},
author = {Zhiyuan Yang and Dongsheng Chen and David J. Levine and Cynthia Sung},
url = {https://repository.upenn.edu/grasp_papers/68
https://youtu.be/cZ1nx_kOw3w},
doi = {10.1109/LRA.2021.3097757},
year = {2021},
date = {2021-07-19},
urldate = {2021-07-19},
booktitle = {IEEE Robotics and Automation Letters},
journal = {IEEE Robotics and Automation Letters},
volume = {6},
number = {4},
pages = {7145-7152},
abstract = {Underwater swimmers present unique opportunities for using bodily reconfiguration for self propulsion. Origami-inspired designs are low-cost, fast to fabricate, robust, and can be used to create compliant mechanisms useful in energy efficient underwater locomotion. In this paper, we demonstrate an origami-inspired robot that can change its body shape to ingest and expel water, creating a jet that propels it forward similarly to cephalopods. We use the magic ball origami pattern, which can transform between ellipsoidal (low volume) and spherical (high volume) shapes. A custom actuation mechanism contracts the robot to take in fluid, and the inherent mechanics of the magic ball returns the robot to its natural shape upon release. We describe the design and control of this robot and verify its locomotion in a water tank. The resulting robot is able to move forward at 6.7 cm/s (0.2 body lengths/s), with a cost of transport of 2.0. },
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Underwater swimmers present unique opportunities for using bodily reconfiguration for self propulsion. Origami-inspired designs are low-cost, fast to fabricate, robust, and can be used to create compliant mechanisms useful in energy efficient underwater locomotion. In this paper, we demonstrate an origami-inspired robot that can change its body shape to ingest and expel water, creating a jet that propels it forward similarly to cephalopods. We use the magic ball origami pattern, which can transform between ellipsoidal (low volume) and spherical (high volume) shapes. A custom actuation mechanism contracts the robot to take in fluid, and the inherent mechanics of the magic ball returns the robot to its natural shape upon release. We describe the design and control of this robot and verify its locomotion in a water tank. The resulting robot is able to move forward at 6.7 cm/s (0.2 body lengths/s), with a cost of transport of 2.0.
The project “Leveraging Fluid-Structure Interactions for Efficient Control in Geophysical Flows” is in collaboration with Ani Hsieh’s lab from University of Pennsylvania, Eric Forgoston’s lab from Montclair State University, and Philip Yecko’s lab from The Cooper Union.
These projects have been supported by the National Science Foundation (NSF) Grant No. 2121887 and the Office of Naval Research (ONR) award #N00014-23-1-2068. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of funding source.
