Michael T. Tolley

Soft autonomous underwater vehicles (AUVs) that employ biological swimming mechanisms have demonstrated immense promise in navigating marine environments with increased efficiency and adaptability and decreased environmental disturbances in comparison to their rigid counterparts. However, most AUVs with soft propulsors are poorly suited for precise swimming at low speeds and station holding tasks—two capabilities critical to underwater monitoring and manipulation. In contrast, gymnotid fish undulate an elongated ribbon fin to aptly swim at low-speeds with high agility and have inspired soft AUVs that mimic their swimming mechanics. However, these AUVs typically use numerous actuators to force undulatory kinematics in their ribbon fin rather than leveraging the fluid-structure interaction between the soft ribbon fin and surrounding water. In this work, we developed a sparsely actuated robot, inspired by the black ghost knifefish (<italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">apteronotus albifrons</i>) and capable of highly agile swimming with only two actuated fin rays. Leveraging prior work on the hydrodynamics of ribbon-finned swimming, we constructed a heuristic for expected thrust generation as a function of the ribbon fin's actuation. Using this heuristic, we investigated four parameterized motion primitives for agile in-plane swimming: 1) forward swimming, 2) banked turning maneuvers, 3) pure yaw rotation, and 4) lateral (i.e., sideways) swimming. We developed a single parametric gait to express all of these primitives and experimentally characterized each primitive. Ultimately, we found that by leveraging fluid-structure interaction, our highly underactuated platform could execute the in-plane swimming capabilities necessary for agile locomotion.

• Developed a simplified lumped-parameter model for small heave plate WECs. • Validated model predictions against extensive experimental data. • The simplified model captured key WEC dynamics efficiently. • Linked laboratory behavior to ocean-relevant operational conditions. Wave energy converters (WECs) are increasingly being adapted for smaller-scale applications, such as powering unmanned oceangoing vehicles. For free-floating ocean platforms, there is a need for a simplified modeling framework that can be efficiently applied and used as a foundation for future optimization and design insight. We present the development and validation of a lumped-parameter model representative of a small heave plate WEC designed for short-period waves (1-5 s) typical of moderate wind conditions. The modeled system consists of a negatively buoyant drag plate and a surface buoy connected by a spring-damper power take-off (PTO) that captures the relative motion between the two bodies. To validate the model, a physical prototype of the heave plate WEC was constructed. A linear actuator simulated wave excitation, replicating the idealized vertical forcing experienced by a buoy in a wave field. Experiments varied wave conditions, system mass, and PTO parameters to assess their influence on the model and measured response. Results show that lower non-dimensional spring constants ( K *) improve broadband performance, while higher non-dimensional masses ( M *) yield only marginal increases in mean power. These findings confirm the model’s accuracy and provide guidance for optimizing WEC design for free-floating platforms.

Dexterous manipulation is limited by both control and design, without consensus as to what makes manipulators best for performing dexterous tasks. This raises a fundamental challenge: how should we design and control robot manipulators that are optimized for dexterity? We present a co-design framework that learns task-specific hand morphology and complementary dexterous control policies. The framework supports 1) an expansive morphology search space including joint, finger, and palm generation, 2) scalable evaluation across the wide design space via morphology-conditioned cross-embodied control, and 3) real-world fabrication with accessible components. We evaluate the approach across multiple dexterous tasks, including in-hand rotation with simulation and real deployment. Our framework enables an end-to-end pipeline that can design, train, fabricate, and deploy a new robotic hand in under 24 hours. The full framework and generated robot hands are open-sourced and available on our website.

Soft robots offer considerable advantages relative to traditional, rigid robots, including: versatility, safety, and rich dynamic behavior. A common approach for actuating soft robots is to use pumps and valves to control pressures and flow rates through fluidic channels embedded in a soft material, to cause desired deformations. Usually, an electronic control board carries the pumps, valves, and other components to control this fluid flow. Currently, many existing hydraulic control systems are designed with application-specific performance requirements, forcing researchers to either invest substantial time and effort into developing custom hardware or constrain their soft robotic systems to the limitations of available control platforms. To address these issues, we present an approach to controlling soft robots underwater that can be achieved with a compact, hydraulic control system. We demonstrate this approach with hardware design housed in a 3D-printable enclosure that is water-resistant up to 1 m for 30 min. The system features three independently controlled pumps, each capable of reversible actuation, with a combined maximum pressure of 0.9 MPa and a maximum f low rate of 3.6 L/min. Additionally, we present an approach to f low sensing using onboard RPM sensors, which avoids adding impedance to the fluid channels and, when combined with our direct pump drive approach, enables precise and simultaneous control of multiple soft actuators. This system offers an accessible platform to accelerate the development of future underwater soft robots.

Soft robots have shown immense promise in settings where they can leverage dynamic control of their entire bodies. However, effective dynamic shape control requires a controller that accounts for the robot's high-dimensional dynamics--a challenge exacerbated by a lack of general-purpose tools for modeling soft robots amenably for control. In this work, we conduct a comparative study of data-driven model reduction techniques for generating linear models amendable to dynamic shape control. We focus on three methods--the eigensystem realization algorithm, dynamic mode decomposition with control, and the Lagrangian operator inference (LOpInf) method. Using each class of model, we explored their efficacy in model predictive control policies for the dynamic shape control of a simulated eel-inspired soft robot in three experiments: 1) tracking simulated reference trajectories guaranteed to be feasible, 2) tracking reference trajectories generated from a biological model of eel kinematics, and 3) tracking reference trajectories generated by a reduced-scale physical analog. In all experiments, the LOpInf-based policies generated lower tracking errors than policies based on other models.

Soft robotic hands promise to provide compliant and safe interaction with objects and environments. However, designing soft hands to be both compliant and functional across diverse use cases remains challenging. Although co-design of hardware and control better couples morphology to behavior, the resulting search space is high-dimensional, and even simulation-based evaluation is computationally expensive. In this paper, we propose a Cross-Entropy Method with Reward Model (CEM-RM) framework that efficiently optimizes tendon-driven soft robotic hands based on teleoperation control policy, reducing design evaluations by more than half compared to pure optimization while learning a distribution of optimized hand designs from pre-collected teleoperation data. We derive a design space for a soft robotic hand composed of flexural soft fingers and implement parallelized training in simulation. The optimized hands are then 3D-printed and deployed in the real world using both teleoperation data and real-time teleoperation. Experiments in both simulation and hardware demonstrate that our optimized design significantly outperforms baseline hands in grasping success rates across a diverse set of challenging objects.

Existing robots designed for locomotion in granular media typically excel at a single purpose—either surface travel or subsurface digging—while lacking the ability to perform both within the same platform. In contrast, nature offers various examples of burrowing organisms that exhibit multi-functional digging behaviors by separating their body into two essential parts: a digger for substrate intrusion and rest of the body as anchor for stabilization and controlling digger orientation. Inspired by these biological strategies, we present an extension to an existing Screw Propelled Vehicle (SPV) that incorporates an adjustable body anchor to reduce drag and enable orientation control. This integration allows the robot to transition between horizontal crawling and vertical digging. We also investigate the effect of local fluidization (LF), a bio-inspired technique that temporarily reduces the resistive forces in granular media. Experimental results show that integrating LF improves surface propulsion performance in terms of speed and depth with increment of over 5x compared to the baseline configuration. These findings support the hypothesis that bio-inspired design principles—specifically body–anchor separation and local fluidization—significantly enhance both the functionality and efficiency of granular locomotion robots, providing a pathway toward more versatile, autonomous, and high-performance subsurface exploration.

Abstract Objectives &amp; Scope A new technology for wireless real-time downhole data monitoring was deployed in a brown field off the coast of Sarawak, offshore Malaysia, to monitor and analyze gas well and reservoir performance through one of the gas producers. The candidate was selected to better understand connectivity in a compartmentalized sector of the clastic reservoir. This innovative instrumentation involved a series of pressure and temperature gauges deployed on high-expansion gauge hangers (one main gauge with three repeaters) along the completion tubing via slickline. Once deployed, the downhole system is paired with a surface-mounted receiver and data logger. Methods, Procedures &amp; Process This technology has effectively enabled real-time downhole data monitoring and analysis, which otherwise requires frequent wireline interventions or a workover to replace the Permanent Downhole Gauge (PDG) in a well fitted with one. The main objective of this wireless data acquisition solution is to enable real-time downhole data transmission, whereas conventional slickline-deployed gauges only support memory-recorded data over a limited timeframe. As such, the asset benefits from minimal production deferments caused by frequent interventions, and inadvertently minimizes operational, mechanical and well integrity risks. Additionally, downhole data acquisition programs are no longer constrained to wireline window availability, resulting in a more robust and continuous well and reservoir monitoring capacity, which is critical to ensure optimum reservoir performance over time. Results, Observations &amp; Conclusions Success criteria were set against cost-benefit, battery life, consistency in data transmission via wireless communication, modification of data transmission frequency, and data suitability to support pressure transient analysis objectives. From the pilot run, the wireless solution consistently transmitted real-time downhole data (20% more data points than the target) and successfully delivered a data recovery rate of more than 97% during the pressure build-up test. Novel &amp; Additive Information This paper describes the advantages, limitations and potential benefits of the sonic wireless monitoring system. Additionally, it will cover insights gained from the installation and utilization of the technology to serve as a valuable foundation for facilitating successful replications in the future.

Wave energy converters (WECs) have been used commonly in large-scale applications and are now included in several unmanned oceangoing vehicles. Wave energy is used to power propulsion systems and onboard instruments to increase deployment range and time, expanding the limits of ocean exploration. We describe the development and validation of a lumped parameter model representative of a small-scale WEC. This system can be useful for small marine systems and is motivated by Unmanned Underwater Vehicle (UUV) applications. Key components of the modeled heave-plate WEC include a surface element, a power take-off (PTO) and a negatively buoyant drag plate. The PTO is modeled as a mass-spring-damper and forces acting on the heave plate include nonlinear drag and added mass, forced by harmonic components. The surface element oscillates vertically in response to passing ocean waves, while the motion of the submerged heave plate is damped, resulting in a periodic differential in motion. As the surface float tends to pull the heave plate, an effectively tuned PTO captures this motion and efficiently converts it into electrical energy. The system is tuned for short period waves (1 to 6 sec) that are ubiquitous at moderate winds. Using the system model, the parameter space is explored under varying wave conditions, identifying optimal design configurations to maximize energy production. To validate the model, a physical prototype of the heave-plate generator was constructed and tested in a controlled experimental setting. The experimental apparatus features a linear actuator with an inline force sensor and component motion is assessed via video tracking. Results from these experiments provide critical insights into the performance of the PTO system under varying wave conditions, highlighting areas where the model aligns with and diverges from observed behavior. These findings informed iterative improvements to the model, enhancing its accuracy and reliability for future design iterations for an oceangoing WEC system.

In this paper we propose a framework towards achieving two intertwined objectives: (i) equipping reinforcement learning with active exploration and deliberate information gathering, such that it regulates state and parameter uncertainties resulting from modeling mismatches and noisy sensory; and (ii) overcoming the computational intractability of stochastic optimal control. We approach both objectives by using reinforcement learning to compute the stochastic optimal control law. On one hand, we avoid the curse of dimensionality prohibiting the direct solution of the stochastic dynamic programming equation. On the other hand, the resulting stochastic optimal control reinforcement learning agent admits caution and probing, that is, optimal online exploration and exploitation. Unlike fixed exploration and exploitation balance, caution and probing are employed automatically by the controller in real-time, even after the learning process is terminated. We conclude the paper with a numerical simulation, illustrating how a Linear Quadratic Regulator with the certainty equivalence assumption may lead to poor performance and filter divergence, while our proposed approach is stabilizing, of an acceptable performance, and computationally convenient.

Abstract Soft robots, particularly large-scale pneumatic systems, offer adaptability and safe interaction with humans due to their compliant materials. This makes soft robots attractive for a variety of applications, including large visual arts installations. One example is the Dual Pneuma robot, a large pneumatic quadruped soft robot composed of eight soft segments with antagonistic pneumatic actuators, developed by the artist Chico MacMurtrie, artistic director of Amorphic Robot Works. This work, a collaboration of artists and engineers, presents the modeling and control of the Dual Pneuma robot for artistic performances. The primary challenge addressed is achieving real-time modeling and control of the kinematics of the robot while maintaining sufficient stiffness to hold its own weight. We developed a graphical user interface for controlling the Dual Pneuma using models of the forward and inverse kinematics, enabling real-time adjustments to the robot’s position and orientation. We also evaluated the use of both bending and stiffness sensors for feedback on the positions and quasi-static forces generated by the bending segments of the robot. To minimize the complexity and weight of the system, we implemented a feedback control algorithm using only flex sensors to achieve desired configurations. Comparative experiments demonstrated that flex sensors alone were sufficient for controlling the configuration of the robot and maintaining stiffness for quasi-static motions in art installations.

Subterranean exploration in submerged granular media (GM) presents significant challenges for robotic systems due to high drag forces and the complex physics of GM. This paper introduces a robotic system that combines water-jet-based fluidization for self-burrowing in submerged environments and an untethered, volume-change mechanism for burrowing out. The water-based fluidization approach significantly reduces drag on the robot, allowing it to burrow into GM with minimal force. To burrow out, the robot uses a soft, inflatable bladder that undergoes periodic radial expansion, inspired by natural systems such as razor clams. Experimental results demonstrate that increased water flow rates accelerate the burrowing process, while the unburrowing mechanism is effective at varying depths. Comparisons between pneumatic and hydraulic untethered systems highlight trade-offs in terms of operational time and unburrowing speed. This work advances the capabilities of robots in underwater environments, with potential applications in environmental monitoring and underwater archaeology.

Soft robots, with their compliance and adaptability, are ideal for applications requiring continuously flexible, dynamic movement, making them promising candidates for underwater locomotion. However, current swimming soft robots often rely on electronic power sources and complex, laborintensive manufacturing, limiting their scalability and use in challenging environments. Recent advancements in 3D printing, particularly fused filament fabrication (FFF), offer a practical alternative for fabricating soft robots, enabling monolithic structures that require minimal assembly. In this work, we introduce a pneumatically powered, electronics-free swimming robot, fully fabricated from soft thermoplastic elastomer (TPE) using a desktop FFF 3D printer. Inspired by the morphology of the tadpole, our design incorporates a pneumatic oscillating controller as the “brain” and segmented actuators as the “tail,” enabling autonomous undulating propulsion without electronics. We demonstrate untethered operation using a portable CO₂ canister and characterize two robot configurations optimized for efficient swimming. The robots achieve controlled oscillation and effective underwater movement, reaching a maximum speed of 0.70 body lengths per second (BL/s). This electronics-free, 3Dprinted design represents a step forward in creating low-cost, accessible soft robotic platforms, suited for exploration in aquatic environments where electronics are impractical.

For robot manipulation, both the controller and end-effector design are crucial. Soft grippers are generalizable by deforming to different geometries, but designing such a gripper and finding its grasp pose remains challenging. In this paper, we propose a co-design framework that generates an optimized soft gripper's block-wise stiffness distribution and its grasping pose, using a neural physics model trained in simulation. We derived a uniform-pressure tendon model for a flexure-based soft finger, then generated a diverse dataset by randomizing both gripper pose and design parameters. A neural network is trained to approximate this forward simulation, yielding a fast, differentiable surrogate. We embed that surrogate in an end-to-end optimization loop to optimize the ideal stiffness configuration and best grasp pose. Finally, we 3D-print the optimized grippers of various stiffness by changing the structural parameters. We demonstrate that our co-designed grippers significantly outperform baseline designs in both simulation and hardware experiments. More info: http://yswhynot.github.io/codesign-soft/

In this paper we explore kinematics ranging from anguilliform to thunniform achieved in a self-sensing multi-degree-of-freedom soft robotic fish and analyze the effect of them on the swimming. First, we examine the characteristics of the bending actuators of the robotic fish. Then, we express the kinematics of the fish as a propagating wave parameterized by three bending amplitudes and a wavelength, which are determined by the flow rates and phase shift of the pumps. We capture various motion patterns generated by different actuator inputs and directly measure the thrust generated by each pattern. We observe that the robotic swimmer can reproduce two different modes of propulsion, that are embodied by two distinct morphological patterns in nature: anguilliform and thunniform. When neither of modes are activated, propulsion is zero or even negative. Finally, we estimate the stationary swimming speed by towing the undulating fish, which satisfies the slip condition (with the speed of the body wave matching the swimming velocity). The analysis of a wide range of kinematic patterns in this study, including two extreme cases of anguilliform and thunniform modes, will provide insights for comprehensive understanding the mechanics of efficient swimming.

Tracking human movement is essential during robot-assisted rehabilitation to provide kinematic feedback for monitoring range of motion, counting repetitions of a certain movement, or achieving closed-loop control. Current strategies for tracking and analyzing human motion, however, are costly, not easily accessible, or challenging to integrate with existing soft assistive robots. In this work, we present skin-tight sensing garments using commercial flex sensors that can be rapidly manufactured using a sew-free lamination strategy. We designed and fabricated sensing garments that can be worn underneath an existing soft robotic wrist orthosis to provide kinematic feedback on flexion and extension of the wrist. We conducted preliminary experiments and demonstrated that our sensing garment can track the position of the wrist with less than 5° of root mean squared error (RMSE) when compared to motion capture. We also demonstrated that our garment design and fabrication approach can be easily extended to other joints and rehabilitation devices to track uniaxial motions by developing a sensing garment for the index finger to provide kinematic feedback when used with a commercial soft rehabilitation glove. Our sensing garments are easy to manufacture, consist of inexpensive and readily available materials, and show promise as a method for providing real-time kinematic feedback of the user during robot-assisted rehabilitation activities.

Autonomous Walking Robots A soft, monolithic, air-powered walking robot is designed for autonomous operation. Fabricated in a single step using a desktop fused filament fabrication (FFF) 3D printer, the robot requires no manual postassembly and becomes immediately operational with a constant-pressure air supply. Its electronics-free design ensures robust performance across outdoor terrains and aquatic environments. More details can be found in article number 2400876 by Yichen Zhai, Michael T. Tolley, and co-workers.

Vertical digging into and out of granular media is a challenging task for autonomous systems. Granular media present considerable resistance to vertical penetration due to the high friction forces and large pressure at depths. In this paper, we present a soft robot that is capable of digging into and out of granular media to depths over 10× its body length. Our robot incorporates a vibration motor to locally fluidize the granular media for burrowing, and a soft pneumatic actuator to adjust the volume and hence the density of the robot, allowing it to transition from digging down to digging up. To analyze the performance of the robot, we measure its weight and density, track its location using a motion capture system, and investigate the effect of local fluidization. When the robot is buried and inflated with vibration turned off, it can increase its passive anchoring force by 5.22× (up to 35 N) relative to when the robot is deflated with vibration on. By contrast, by inflating the soft pneumatic bladder and providing vibration the robot is able to actively unburrow.

Anguilliform-swimming soft fluidic robots hold great promise for a range of underwater applications. However, because they leverage the complex dynamics of soft bodies interacting with fluids, it is challenging to use intuition to determine design parameters. Multidisciplinary design optimization offers a promising solution to this challenge by providing an automated and systematic method for leveraging computational models to find optimal design parameters. This study investigates a method for multidisciplinary control co-design optimization of anguilliform-swimming soft robots, using physics-based models to optimize shape and control parameters. The modeling framework includes a geometry-centric approach for geometry modeling and parameterization, a three-dimensional finite element model for structural mechanics, and an unsteady panel method for fluid dynamics. The approach is tested by applying it to the optimization of a pre-existing eel-inspired soft robot. Model parameters are estimated from existing experimental data, and two control co-design optimizations, with high-level and multilevel shape parameterizations, are performed to minimize energy cost. The optimized actuator module is manufactured, and used to collect additional data for re-estimating the structural model parameters. The optimization is performed again with the updated model parameters, showing a simulated energy cost reduction of 45% and the prescribed 128% speed increase compared to the baseline design with optimized control and updated model parameters. These results demonstrate the potential of the proposed optimization approach to advance the performance of anguilliform-swimming soft fluidic robots.

An actuator health estimation algorithm for a soft swimming robot that can perform anguilliform swimming is developed. Due to harsh operational environments of underwater robots, and the common degradation of soft robot materials and actuators, accurate estimation of actuator functionality is necessary for robots to perform their missions as well as return to base in the event of actuator degradation and failure. Termed REACH (Real-time Estimator of Actuator Control and Health), the architecture employs a soft robot model, sigma point filter, and a formal statistical hypothesis test to adequately capture the nonlinearities and changes over time. The performance of REACH using three sensor types (GPS, IMU, and Bend Sensor) with one sensor on each actuator is compared, demonstrating that both bend sensor and IMU are adequate choices. Sensor quantity and placement are evaluated for IMU and bend sensor, showing two sensors are sufficient for IMU, whereas three sensors are needed for bend sensor. Three swimming gaits (linear swimming, wide turning, tight turning) are compared, demonstrating that REACH can successfully predict actuator health for all three gaits, with minimal differences in performance. A filter validation method shows the fault estimation algorithm is statistically consistent in finding the correct degradation. The approach is experimentally evaluated using bend sensor data collected from a fish robot, demonstrating that REACH can successfully estimate actuator health with noisy data and variations in manufacturing.

Many wearable soft robotic devices have been developed to assist with spasticity – the increased muscle tone seen in individuals with neurological conditions such as stroke or cerebral palsy. However, these devices typically lack the ability to individually assist multiple degrees of freedom (DOF), which is essential to performing a broad range of daily tasks. To address this need, we developed a soft robotic thumb orthosis to achieve movement in three DOF. We focused on the thumb as it performs the most complex movement. Our orthosis used three unfolding textile pneumatic actuators to assist with the extension of the interphalangeal joint, combined extension of the metacarpal and carpometacarpal joints, and abduction of the carpometacarpal joint. The orthosis was tested on both a 3D-printed biomimetic test bed that emulated spasticity and a healthy individual. The orthosis met the full range of motion targets for each joint, except for the abduction of the carpometacarpal joint with medium spasticity, and surpassed the range of motion threshold required for activities of daily living. The orthosis was able to: (1) assist each DOF independently, (2) move the tip of a biomimetic thumb to a pre-determined target with total error of less than 1.5 cm when actuated with open-loop control, and, (3) provide individual joint assistance to a user with no motor impairment. This research showcases the ability of a soft robotic orthosis to assist movement in simulated spasticity of individual thumb joints and provides an important step towards a platform to study whether independent joint assistance during rehabilitation improves patient outcomes.

Soft fluidic actuators (SFAs) are widely adopted for soft robotic applications, for example, for the manipulation of delicate objects of various geometries. However, the innate compliance of SFAs prevents them from sustaining large forces when needed. Existing methods to achieve variable stiffness in SFAs necessitate complex mechanisms that require additional control inputs and often limit the flexibility of the actuators. This work explores the use of supercooled sodium acetate trihydrate (SAT) solution as a multifunctional working fluid for SFAs to allow independent control of position and stiffness; the solution is used to both inflate the elastomer chambers while in liquid state and to achieve a dramatic increase in stiffness with rapid solidification. The mechanical properties of crystallized SAT samples in flexible membranes and with embedded materials for reinforcement are first investigated. SFAs that use the SAT solution to greatly increase their stiffness (up to 13 times) are then tested. Furthermore, the use of supercooled SAT is demonstrated to achieve both manipulation and stiffness change with a single control input. The proposed approach is a new way to combine the high manipulability of SFAs with stiffness tunability and opens up new applications for soft robots and manipulators.

Vine robots are soft robots that translate by everting, or growing, from their tips. This mechanism of translation minimizes the application of shear forces on the environment, making them particularly well-suited for surgical tasks, such as colonoscopy. However, steering and retracting vine robots within tortuous and delicate environments presents significant challenges. In this article, we introduce a novel soft robotic system for colonoscopy that consists of an inchworm-inspired device—the InchIGRAB—nested within a vine robot. The InchIGRAB is designed to enable steering of the vine robot along curved paths, as well as to enable controlled retraction of the vine robot after it reaches its target. We present the design, modeling, and fabrication of the robotic system and characterize its performance. Furthermore, we demonstrate the presented robotic system's ability to safely navigate the entire colon length, including several curved sections, during both forward motion and retraction, highlighting its potential as a robotic colonoscope that offers enhanced safety for patients.

The fully automated fabrication of robots has long been a holy grail with the potential to revolutionize various industries, including manufacturing, construction, disaster relief, and space exploration. 3D printing offers a promising approach to automation, but the ability to print entire, complex robots with multiple materials remains limited. Previous approaches have simplified robot manufacturing by using fluidic control circuits, but these rely on labor‐intensive methods like silicone molding and manual assembly, limiting accessibility and replicability. Recent work, including this work, has demonstrated 3D‐printed robotic grippers and crawlers with embedded control circuits, but generating cyclic control outputs for legged locomotion in rough terrain remains challenging. This study addresses the challenge with a monolithic 3D‐printable four‐phase bistable oscillating valve, capable of generating coordinated motion of multiple limbs from a steady source of pressurized air. The ability of the oscillator to control an electronics‐free autonomous legged robot capable of walking on rough terrain, which can be fully fabricated on a desktop 3D printer without postassembly is demonstrated. The robot is operational immediately upon connection to an air supply. This development marks a significant step toward accessible, customizable, and biodegradable autonomous soft robots that can be produced using desktop 3D printers with no human intervention.

Soft robotic fish have significant potential for underwater applications, with eel-inspired robots offering notable advantages due to the efficiency of their anguilliform swimming motion. This efficiency enables extended range and endurance, making them ideal for various missions. However, designing eel-inspired robots poses significant challenges due to the nonlinear, dynamic fluid-structure interactions (FSI) that drive system performance. Multidisciplinary design optimization (MDO) offers a systematic approach to exploring this unintuitive design space, though few studies have applied MDO to eel-inspired soft robots. While there have been many studies exploring multidisciplinary design optimization with fluid-structure interaction, gradient-based design optimizations with dynamic fluid-structure interaction remains relatively unexplored. A previous study investigated a method for optimizing eel-inspired robots using a static structural model and a dynamic fluids model. While promising, mapping the static structural solution to the dynamic fluids mesh introduces an unquantified amount of modeling error. This study seeks to build on the prior work by investigating the method of directly modeling the dynamic hydroelasticity for shape optimization. Furthermore, this work seeks to demonstrate dynamic hydroelastic shape optimization with analytic unsteady adjoint computation. The method applies a geometry-centric approach, dynamic Euler-Bernoulli beam theory to model structural dynamics, and an unsteady panel method to model the fluid dynamics. Additionally, the models are implemented using a graph-based modeling language to automate the unsteady adjoint computation. The proposed method is applied to optimize the efficiency of an existing modular, eel-inspired soft robot. Furthermore, the method explored in previous work is applied to the same optimization result to investigate the impact of the difference in modeling approach. The presented method shows a 73.9\% decrease in cost of transport and 130\% increase in swim speed compared to a control optimization of the baseline design. When compared to the optimization result using the static structural model, the optimal designs exhibit similar design trends but significant differences in the scale of the final design. These findings demonstrate the feasibility of automated unsteady hydroelastic adjoint computation and highlight the trade-offs between static and dynamic modeling fidelity in optimizing soft robotic systems.

Data-driven model reduction methods provide a nonintrusive way of constructing computationally efficient surrogates of high-fidelity models for real-time control of soft robots. This work leverages the Lagrangian nature of the model equations to derive structure-preserving linear reduced-order models via Lagrangian Operator Inference and compares their performance with prominent linear model reduction techniques through an anguilliform swimming soft robot model example with 231,336 degrees of freedom. The case studies demonstrate that preserving the underlying Lagrangian structure leads to learned models with higher predictive accuracy and robustness to unseen inputs.

Lessons from the CubeSat and Mars Exploration programs may guide the infusion of robotics for planetary science and exploration.

Biologically inspired soft anguilliform swimming robots show great promise in underwater exploration. Their soft bodies promise to reduce the chance of harming humans or wildlife with which they come into contact, and also to reduce their chance of becoming stuck in complex environments. Furthermore, the efficiency of anguilliform swimming may enable long duration operation. However, the design and fabrication of soft anguilliform swimming robots remain challenging. Here we present a design concept for a modular soft anguilliform robot. To address the challenge of consistently fabricating modules for this design, we also present a new fabrication method that combines injection molding and lost-core molding for the consistent fabrication of bi-directional fluidic elastomer actuator modules. We evaluate the consistency of the fabrication method through visual inspection, weight measurements, bending angle measurements and the measurement of force output of the actuator. This work represents a step towards an autonomous eel-inspired soft robot, as well as a new fabrication approach that may enable a number of other new soft robotic systems.

Wearable soft robots can be effective tools for rehabilitation due to their inherent safety and compliance. Challenges, however, exist regarding the development of suitable on-body actuation methods. Furthermore, the majority of existing soft wearable devices are not accessible and easy to use for those with physical disabilities. This paper presents the design, fabrication, and characterization of a soft robotic wrist orthosis to achieve flexion and extension for continuous passive motion therapy. First, we developed bending textile pneumatic actuators that could be mechanically programmed to conform to the target joint's anatomy when mounted on the body. The textile pneumatic actuators achieved up to 2.24 Nm of torque at 124 kP <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$a$</tex> . We then embedded the textile pneumatic actuators into a soft wrist orthosis that we designed to ensure it was easy to don/doff without assistance. To determine the operating pressures and range of motion achievable by our soft robotic wrist orthosis, we conducted a device evaluation study with three healthy individuals. Our device achieved over 100 degrees of flexion/extension assistance at operating pressures below 90 kPa. This work takes the first steps towards developing a wearable soft robotic device that can deliver passive therapy at home without the need for a physical therapist or assistant.

Soft robotic systems necessitate accurate and reliable sensor readings to detect environmental interactions and provide precise feedback for control. To be effective, soft sensors must exhibit sensitivity, reliability, repeatability, and flexibility. A versatile approach to sensing for soft robots uses soft air-filled deformable structures with pressure transducers to detect pressure changes due to applied forces. However, the common approach of employing one pressure transducer per sensing chamber limits the scalability of this sensing approach (e.g., for large arrays able to detect touch at many locations). Here we present an approach to the design of pneumatic sensor arrays that reduces the number of required transducers. We develop mathematical models to analyze the pressure variations within the sensor arrays in response to applied forces at various locations. We also introduce a method of rapidly fabricating sensor arrays by laminating elastomeric sheets patterned with laser-cut sacrificial layers. We then use our model to optimize the geometry of the sensors and evaluate the results experimentally. Finally, we devise an algorithm capable of determining the location of multiple touches anywhere within the sensor array. This work represents a step towards the practical application of soft pneumatic sensors, particularly for robotic sensing and haptic devices, enhancing the safety of human-robot interactions.

Retractors for laparoscopic surgery face competing challenges: they must be sufficiently soft/small to be (1) deployed and removed through a small opening and (2) manipulated into a desired configuration, but also (3) sufficiently rigid/wide to affix tissue atraumatically. Existing rigid designs are functional, but present the need for additional incisions and external anchoring. We developed a jamming-based foldable retractor capable of deployment, atraumatic anchoring within the body, retraction, and removal without the need for any additional incisions or surgeon involvement. Through FEA and experiments, including mechanical testing and in-porcine testing, we assessed the effect of different device cross-sectional areas on their ability to retract a liver. A patterned high-friction surface on one side of the device provides atraumatic anchoring to the abdominal wall, we compared different patterns, surface conditions, and preloads experimentally. To facilitate easy (i.e., low-force) removal, the device has a tapered end. When removed via the trocar, the taper causes the device to self-fold, with different tapers resulting in different removal forces. Validated by experimental testing and a in-porcine case study, our device demonstrates the ability of layer-jamming-based approaches to fill a hole in the current surgical toolkit and presents applicability beyond liver retraction.

Data-driven model reduction methods provide a nonintrusive way of constructing computationally efficient surrogates of high-fidelity models for real-time control of soft robots. This work leverages the Lagrangian nature of the model equations to derive structure-preserving linear reduced-order models via Lagrangian Operator Inference and compares their performance with prominent linear model reduction techniques through an anguilliform swimming soft robot model example with 231,336 degrees of freedom. The case studies demonstrate that preserving the underlying Lagrangian structure leads to learned models with higher predictive accuracy and robustness to unseen inputs.

The deep ocean, Earth's untouched expanse, presents immense challenges for exploration due to its extreme pressure, temperature, and darkness. Unlike traditional marine robots that require specialized metallic vessels for protection, deep-sea species thrive without such cumbersome pressure-resistant designs. Their pressure-adaptive forms, unique propulsion methods, and advanced senses have inspired innovation in designing lightweight, compact soft machines. This perspective addresses challenges, recent strides, and design strategies for bioinspired deep-sea soft robots. Drawing from abyssal life, it explores the actuation, sensing, power, and pressure resilience of multifunctional deep-sea soft robots, offering game-changing solutions for profound exploration and operation in harsh conditions.

Robots capable of generating adhesion forces that can achieve free movement in application environments while overcoming their own gravity are a subject of interest for researchers. A robot with controllable adhesion could be useful in many engineered systems. Materials processing equipment, robots that climb walls, and pick-and-place machines are some examples. However, most adhesion methods either require a large energy supply system or are limited by the properties of the contact plane. For example, electromagnetic adhesion requires a ferromagnetic surface and pneumatic adhesion requires a flat surface. Furthermore, nearly all existing approaches are only used to generate adhesion forces and often require additional mechanisms to remove the adhesive component from the surface. In this study, we aimed to develop a simpler method of adhering to a surface while simultaneously moving in directions parallel to the surface, using multiple vibration sources to generate normal adhesion and propulsion. To test our approach, we constructed circular and elliptical models and conducted experiments with various inputs and model parameters. Our results show that such a gas-lubricated adhesive disk could achieve adhesive rotation and displacement in the plane without requiring any auxiliary operating system. Using only vibration sources, we were able to generate the necessary adhesion and propulsion forces to achieve the desired motion of the robot. This work represents a step towards the construction of a small-sized tetherless robot that can overcome gravity and move freely in a general environment.

In this paper we propose a framework towards achieving two intertwined objectives: (i) equipping reinforcement learning with active exploration and deliberate information gathering, such that it regulates state and parameter uncertainties resulting from modeling mismatches and noisy sensory; and (ii) overcoming the computational intractability of stochastic optimal control. We approach both objectives by using reinforcement learning to compute the stochastic optimal control law. On one hand, we avoid the curse of dimensionality prohibiting the direct solution of the stochastic dynamic programming equation. On the other hand, the resulting stochastic optimal control reinforcement learning agent admits caution and probing, that is, optimal online exploration and exploitation. Unlike fixed exploration and exploitation balance, caution and probing are employed automatically by the controller in real-time, even after the learning process is terminated. We conclude the paper with a numerical simulation, illustrating how a Linear Quadratic Regulator with the certainty equivalence assumption may lead to poor performance and filter divergence, while our proposed approach is stabilizing, of an acceptable performance, and computationally convenient.

Most soft robots are pneumatically actuated and fabricated by molding and assembling processes that typically require many manual operations and limit complexity. Furthermore, complex control components (for example, electronic pumps and microcontrollers) must be added to achieve even simple functions. Desktop fused filament fabrication (FFF) three-dimensional printing provides an accessible alternative with less manual work and the capability of generating more complex structures. However, because of material and process limitations, FFF-printed soft robots often have a high effective stiffness and contain a large number of leaks, limiting their applications. We present an approach for the design and fabrication of soft, airtight pneumatic robotic devices using FFF to simultaneously print actuators with embedded fluidic control components. We demonstrated this approach by printing actuators an order of magnitude softer than those previously fabricated using FFF and capable of bending to form a complete circle. Similarly, we printed pneumatic valves that control a high-pressure airflow with low control pressure. Combining the actuators and valves, we demonstrated a monolithically printed electronics-free autonomous gripper. When connected to a constant supply of air pressure, the gripper autonomously detected and gripped an object and released the object when it detected a force due to the weight of the object acting perpendicular to the gripper. The entire fabrication process of the gripper required no posttreatment, postassembly, or repair of manufacturing defects, making this approach highly repeatable and accessible. Our proposed approach represents a step toward complex, customized robotic systems and components created at distributed fabricating facilities.

Granular environments, such as sand, are one of the most challenging substrates for robots to move within due to large depth‐dependent forces, unpredictable fluid/solid resistance forces, and limited sensing capabilities. An untethered robot is presented, inspired by biological diggers like sea turtles, which utilize underactuated appendages to enable propulsion and obstacle sensing in granular environments. To guide the robot's design, experiments are conducted on test appendages to identify the morphological and actuation parameters for forward thrust generation. Obstacle sensing is observed in granular media by measuring the increased force on the moving appendage caused by changes in the granular flow around it. These results are integrated into an untethered robot capable of subsurface locomotion in a controlled granular bed like natural, loosely packed sand. The robot achieves subsurface “swimming” at a speed of 1.2 mm s −1 , at a depth of 127 mm, faster than any other reported untethered robot at this depth, while also detecting obstacles during locomotion via force sensors embedded in the appendages. Finally, subsurface robot locomotion in natural sand at the beach is demonstrated, a feat no other robot has accomplished, showcasing how underactuated structures enable movement and sensing in granular environments with limited limb control.

A fundamental challenge of pneumatically powered soft robotic devices is the scalability of fluidic control systems to address each actuated degree of freedom, as the required electromechanical valves are bulky and expensive. Previous solutions have compromised the reprogrammability and/or the bandwidth of the fluidic system. This article describes and models a fluidic subsystem, a fluidic matrix circuit (FMC) , that enables the independent control of many () actuators using a limited number () of electromechanical valves. The fundamental unit, a fluidic logic module (FLM) , generates a bidirectional pressure signal (vacuum or positive pressure) based on the state of the mass flow through it. Thus and array of pneumatic actuators can be addressed individually using an array of FLMs integrated into a matrix (i.e., an FMC), with electromechanical valves to switch the mass flow through each row and column of the matrix. The resulting refresh rates are an order of magnitude faster than previous approaches. This concept with a prototype FMC able to control 25 actuators using 10 electromechanical valves for applications including a fluidic shape display and a wearable haptic vest is demonstrated. This approach could enable more complex and sophisticated soft robotic devices with scalable control hardware.

A pneumatically powered, fully untethered mobile soft robot is described. Composites consisting of silicone elastomer, polyaramid fabric, and hollow glass microspheres were used to fabricate a sufficiently large soft robot to carry the miniature air compressors, battery, valves, and controller needed for autonomous operation. Fabrication techniques were developed to mold a 0.65 meter long soft body with modified Pneumatic network actuators capable of operating at the elevated pressures (up to 138 kPa) required to actuate the legs of the robot and hold payloads of up to 8 kg. The soft robot is safe to handle, and its silicone body is innately resilient to a variety of adverse environmental conditions including snow, puddles of water, direct (albeit limited) exposure to flames, and the crushing force of being run over by an automobile.