Harnessing Liquid Crystal Elastomers for Locomotion and Mechanical Intelligence in a Soft Robot

Abstract

Recently, there has been an increased interest in endowing intelligent behaviors and features in soft robotic systems. As a prerequisite for intelligence, a system must integrate sensing, information processing, and the ability to act in response to external stimuli. This work presents a soft robotic crawler that demonstrates locomotion using electroactive liquid crystal elastomers (LCEs). By integrating independent components such as a photo-responsive LCE switch into a conductive electromechanical processing network based on sequential logic, the robot can sense optical indicators and process this information to change direction autonomously. This study expands the design of the individual mechanical material subsystems and experimentally showcases the autonomous operation of the soft robot. The embedded bistable mechanism stores the present operational state of the robot and enforces directional locomotion by controlling the position of a mechanical hard stop that interfaces with the legs. The robot exemplifies the advanced potential of soft intelligent material systems for complex autonomous behavior, leveraging the unique properties of LCEs and a mechanical-electrical network for information processing without the need for traditional electronic controllers.

Introduction

Soft robotic systems have shown significant promise in many engineering applications, including navigation in hostile environments,^1–3 medical devices,⁴ and human-machine interfaces.⁵ The characteristics of soft robots, such as their ability for complex motion, resilience to harsh conditions, and adaptability, offer attractive alternatives to traditional robotic systems. To increase the functionality of soft robots, engineers have sought to incorporate elements of intelligence into the design and control of the mechanical system.^6–8 This involves the integration of sensing external stimulus, processing the information, and acting accordingly.⁹ Such intelligent functionality enables increasingly autonomous soft robotic systems capable of adapting and making decisions without direct human control.

A frequent goal of soft robotics research is to cultivate novel locomotion capabilities.^10–13 To this end, researchers have drawn inspiration from biological soft systems to design gait and motion profiles that enable motion.^7,14–16 Soft robotic systems have employed many actuation techniques to achieve motion, including pneumatic pressurization,^17–20 magnetoelastic control,^21,22 and dielectric elastomer actuators.^23–25 Recently, thermally driven liquid crystal elastomer (LCE) actuators have received attention due to their large and reversible actuation, relatively low mass, and versatility in response to external stimuli.^26–28 With Joule heating, complex behavior can be achieved by controlling the electrical signal to the actuator.²⁹

Because locomotion requires careful coordination of actuators, several strategies have been employed to achieve such control. Some have explored the integration of traditional electronic controller,^26,30 which allow for complex gait generation and multimodal locomotion. Yet, traditional controllers may introduce challenges as they require the interfacing of rigid components in a soft robotic system eliminating some of the benefits of the soft system. Mechanical computing techniques have received interest recently due to their ability to embed processing functionality directly into the mechanical system.³¹ Through this technique, complex combinational logic operations can be performed based on the configuration of digital mechanical material systems.^32–35 By integrating actuators to self-update the mechanical platform, autonomous sequential and cyclical operations can also be performed.^36,37 He et al.³⁸ demonstrated how a modular system that integrates both rigid and soft components with materials including LCEs, and hydrogels to allow for autonomous motion that responds to heat, light, and chemical stimuli. It offers a compelling demonstration of how incorporating field-responsive materials can allow for more complex behavior. To achieve more intelligent behavior, mechanical information processing combined with soft locomotion strategies enables the development of soft robotic systems that offer advanced autonomous capabilities and adaptability for complex tasks.

This work introduces a soft robotic crawler capable of autonomous locomotion using electroactive LCEs. Previous research has presented individual components of sensing and sequential logic to showcase materials and strategies to approach intelligence in material systems.^35,36 While Tholen et al.³⁵ demonstrated optomechanical logic gates, and El Helou et al.³⁶ highlighted sequential logic in a soft electroactive platform, the scope of their work revolved around isolated elements of intelligent material systems. This study takes an integrated approach to combine these foundational principles of sensing, processing, and actuation into a cohesive soft robotic platform. Such integration allows the robot presented here to achieve locomotion, a common goal for soft robotic systems. By integrating a photo-responsive LCE switch,³⁵ into a conductive electromechanical processing network based on sequential logic,³⁶ the robot can sense the presence of an optical signal and process the information to change direction. A singular LCE actuator coupled with a mechanical hard stop enforces directional motion. This study explores the design of the robot subsystems, their integration, and the autonomous operation is experimentally demonstrated. The presented robot exemplifies the potential of soft intelligent material systems capable of complex autonomous behavior without the need for traditional electronic controllers through integrated mechanical logic.

Locomotion with Elements of Intelligence

The soft robot in this work consists of a monolithic 3D-printed body using fused deposition modeling (FDM) processes supported with two legs made of thermoplastic polyurethane (TPU) (Filaflex 70A, Recreus) as shown in Figure 1. A network of conductive silver thermoplastic polyurethane (Ag-TPU) elastomer delivers electrical current to electroactive LCE actuators which drive the locomotion of the robot. By autonomously reconfiguring the mechanical structure of the robot in response to optical stimuli in the form of UV light, the conductive Ag-TPU pathway is updated to control the behavior of the individual functional subsystems. With this ability to sense external stimuli, process the information, and act in response to that information, the soft robot here exhibits the fundamental building block of sense-process-act feedback for intelligent behavior. The feedback loop seen in Figure 1A summarizes the relationship between the three requisite functions in the soft robotic system. Figure 1B details the design of the soft robot and the organization of the critical elements in the feedback loop. A photograph of the 3D-printed robot with critical dimensions is shown in Figure 1C.

FIG. 1.

Basis of intelligence in a soft crawling robot. (A) Feedback loop connecting sensing, processing, and action. (B) Schematic of a soft robot with elements of the feedback loop labeled. (C) 3D-printed TPU robot with LCE actuators and photo-responsive LCE. TPU, thermoplastic polyurethane; LCE, liquid crystal elastomer.

The robot consists of three subsystems that correspond to the outlined functions in Figure 1A: sensing, information processing, and acting. To respond to optical stimulus, a photo-responsive LCE switch is integrated into the top surface of the robot. Based on the state of the switch, an electrical current is sent to the processing subsystem through the mechanical-electrical network consisting of the conductive Ag-TPU traces and a self-adaptive, magnetically multistable platform. The multistable mechanism serves as a non-volatile memory that stores the state of the robot and governs the mechanical and electrical response of the robot. The processed electrical signal is then passed to the acting subsystem which activates the locomotive LCE actuator causing one of the legs of the robot to move which propels the LCE in one of two directions. Finally, the locomotion of the robot can also influence the exposure of the optical signal to the photo-responsive LCE thus altering the input signal and behavior of the robot. This control loop is cycled indefinitely while power is supplied to the robotic system. The following subsections elucidate the design and function of each subsystem.

LCEs for sensing and actuation

The photo-responsive LCE switch that allows the soft robot to sense the presence of UV light is adapted from previous work on optomechanical computing from Tholen, et. al.³⁵ LCEs exhibit reversible shape change based on the alignment of liquid crystal mesogens during the material synthesis.³⁹ When heated, the liquid crystals become disordered which facilitates the shape change. In this work, the LCE films used for sensing are aligned in a twisted-nematic orientation which exhibits curling and uncurling behavior when exposed to heat. By infusing light absorbing carbon black material into the LCE mixture during synthesis, the LCE film can respond to UV light as the light energy is converted to thermal energy when absorbed. A conductive pad of Ag-TPU is placed on the switch to bridge conductive traces in the processing mechanical-electrical network.

Figure 2A outlines the fabrication process of the LCE switch used in this work. The LCE film is aligned in a glass cell made of glass slides with orthogonal alignment layers to create the twisted-nematic orientation. The arrows in (i) represent the alignment direction of each layer. The film (i) is polymerized under UV light while being heated at 93°C. Tholen, et. al.³⁵ showed that this elevated polymerization temperature causes small cantilever structures cut into the LCE film to curl at room temperature and flatten when heated. Once polymerized, the cells are separated (ii), and the LCE is laminated to the slide with the alignment layer along the length of the slide. The cell is covered with a laser-cut mask of low-tack artist tape with 200 µm thickness. The pattern is a 7.5 × 4 mm rectangle with the length of the rectangle along the same direction as the alignment layer. Ag-TPU conductive ink is then blade-coated onto the mask. In previous work,³⁵ liquid metal is used to create the trace, yet in this application, Ag-TPU is used to match the material of the conductive traces used throughout the rest of the robot. In (iii) the mask is removed, and the film is heated to evaporate the solvent in the Ag-TPU ink. The solvent in the conductive ink causes some swelling in the LCE which can destroy small features in the film, so it is necessary to evaporate the solvent before cutting the switch geometry. Finally, (iv) the switch is cut from the film using a laser cutter. A more detailed description of the LCE material synthesis and fabrication is found in Supplementary Data.

FIG. 2.

Fabrication and operation of liquid crystal elastomer (LCE) material. (A) Fabrication procedure for photo-responsive LCE switch. Arrows on the slides represent alignment layer orientation causing twist-nematic orientation of LCE. (B) Operation of photo-responsive LCE connecting Ag-TPU traces when exposed to UV light. (C) Fabrication procedure for LCE actuator with Ag-TPU heating elements. (D) Contraction of LCE actuator subject to the 3V input voltage. Ag-TPU, silver thermoplastic polyurethane; LCE, liquid crystal elastomer.

Photographs of the operation of the photo-responsive switch are shown in Figure 2B. The switch consists of the Ag-TPU pad connected to a cantilever section. The section of the switch without silver is used to adhere the switch to the robot body. A silicone adhesive (DAP All-Purpose) is used to join the LCE film to the printed TPU body with the Ag-TPU pad facing down. As seen in Figure 2B, the switch, which is outlined with a dashed yellow line, curls upward which causes a break in the conductive trace connecting power to the memory processing subsystem. When the switch is exposed to UV light, the cantilevered switch bends down and contacts the robot body. By limiting the geometry of the LCE switch to a small cantilever with an area of Ag-TPU at the end, the twisting motion of the film is limited to curling toward the surface of the robot body when actuated by light. The Ag-TPU pad bridges the gap in the conductive trace which allows the electrical signal to pass from the voltage source V_CC to the processing subsystem triggering the reconfiguration of the structure of the robot to move in the opposite direction. This functionality can be roughly compared to a traditional phototransistor or relay where the presence of a light signal allows current to flow through the transistor to the rest of the circuit. Using the LCE films, the switching behavior is realized in a fully soft medium.

The electroactive LCE actuators used to facilitate the toggling behavior and locomotion are fabricated in the process outlined in Figure 2C. Each actuator consists of (i) an Ag-TPU based heating element fabricated through a custom direct-ink-writing process, which is (ii) sandwiched between two partially cross-linked LCE elements. Because the LCE elements are only partially cross-linked, when joined the two layers bond together without the need for additional adhesive. This eliminates the risk of separation of the LCE actuator during repeated operation. To align the LCE mesogens, layers (iii) are stretched to 100% strain and exposed to UV light at the top and bottom of the sample simultaneously. The resulting actuator (iv) has dimensions of 10 mm by 50 mm with an average resistance of 1.25 Ohm. This process is adapted from previous work published by El Helou et. al.³⁶ When a voltage is applied across the actuator, the temperature of the Ag-TPU trace increases due to Joule heating. The increased temperature causes the aligned LCE mesogens to become disordered and the actuator reduces in length. Due to the soft TPU (85A Shore hardness) base of the heating element and the thin serpentine pattern, the Ag-TPU trace can deform with the contracting LCE layers with negligible resistance or fatigue. Photographs of the operation of the LCE actuator are shown in Figure 2D. Previous research has characterized the response to differing voltages and the break force of the LCE actuators.³⁶ The characterization performed in,³⁶ outlines how the voltage affects the time it takes to toggle a single processing unit like the system used in the memory unit in this work. Due to Joule heating, higher voltages increase the induced temperature of the heating element which causes faster actuation of the LCE as the heat conducts rapidly through the LCE material. At lower voltages, the lower induced temperature causes slower heating and slower actuation. The work also characterizes the break force of the LCE actuator which represents the maximum force that the actuator can apply before the material fails. The break force of the LCE actuators used in this work is approximately 4.9N. These LCE actuators are used in both the processing and locomotive subsystems as discussed in subsequent sections. Supplementary Data contains more details of the LCE actuator synthesis and fabrication methods.

Processing subsystem with sequential logic

To process information, the soft robot uses a mechanical-electrical network of conductive Ag-TPU traces which directs the electrical power to the soft LCE actuators to facilitate locomotion and structural reconfiguration. As shown in the feedback loop of Figure 1A, this network processes the photo-responsive signal along with the present state of the robot. The present state of the system is stored in mechanical memory in the form of a bistable mechanism as seen in Figure 3A. The design of the bistable mechanism and accompanying processing network is adapted from mechanical computing platforms introduced in previous works.^33,36 Here the bistable mechanism consists of rotating polygonal elements assembled in parallel. The orientation of N45 Neodymium magnets inserted in the polygon segments results in two stable mechanical configurations when the polygonal elements reach a contact state with a translating segment as illustrated in Figure 3B. The translating segment is connected to a shifting segment underneath the main body of the robot that contacts one of the legs in each configuration. This shifting segment is a directional enforcement hard stop that enables bidirectional motion control. Details of the operation and interaction of this hard stop with the robot legs are covered in the next section (Section 2.3).

FIG. 3.

Description of the memory subsystem. (A) Design of bistable mechanism and mechanical-electrical network. LCE actuators enable toggle behavior when photo-responsive LCE film is actuated by UV light. (B) Photographs of bistable states of the memory subsystem. The dotted cyan line highlights the orientation of the bistable mechanism. (C) Logic diagram and characteristic table of toggle behavior of memory subsystem. LCE, liquid crystal elastomer.

The bistable mechanism is abstracted to represent a 1-bit digital material whose value $Q_{n}$ can be represented by a ‘1’ or ‘0’ binary digit.^32,33 These digital states directly correlate to the direction of locomotion due to the position of the directional enforcement hard stop. When the polygonal elements are rotated counter-clockwise $(Q_{n} = 0)$ , the robot translates to the right. With clockwise rotation $(Q_{n} = 1)$ , the robot translates to the left. The orientation of the polygon segments is highlighted by the dotted cyan line in Figure 3B. By connecting soft, electroactive LCE actuators between the translating element and the robot body as shown in Figure 3A, the bistable mechanism can autonomously toggle between each stable state. The operation of the bistable mechanism is described as a T flip-flop sequential logic element. The logic diagram and characteristic table of a T flip-flop are shown in Figure 3C where $n$ subscript indicates the state index representing the present state in the sequential cycle. When the value of the toggle bit is high $(T = 1)$ , the present value $Q_{n}$ is set to update to the opposite value for the subsequent state $Q_{n + 1}$ . As shown in Figure 3A, the photo-responsive LCE switch serves as the toggle bit $T$ . When no light is on, the electrical power is disconnected $(T = 0)$ . When UV light is present, the switch is connected $(T = 1)$ , and electrical power is passed to update the mechanism to the next state. The updating is done by sending an electrical signal $(Δ Q_{n}, Δ Q_{n} ″)$ to the corresponding LCE actuator to switch to the next state. This toggle behavior will continue indefinitely while the UV light is present $(T = 1)$ at a frequency determined by the clock signal. The clocking signal of the memory subsystem here is based on the intrinsic electro-thermal properties of the LCE actuators.³⁶ The actuators in this work result in a clocking signal of approximately 11 MHz. This clocking signal is distinct from the timing of the locomotion which is discussed in the next section (Section 2.3).

Locomotion with structural reconfiguration

The third element of intelligent behavior in the presented mechanical system is the ability to act in response to processed information. Based on the state of the information processing system, the soft robot in this work moves autonomously in one direction until UV light triggers a shift in the mechanical configuration. Figure 4 illustrates an expanded finite-state diagram of the locomotive operation of the soft robot and the transition between each direction of travel. The robot achieves directional locomotion through repeated actuation and relaxation of a single leg at a time. The two states of the bistable mechanism of the memory subsystem correspond to two directions of travel: right or left. The behavior of the robot in each state is highlighted in green in Figure 4A and C. The transition between states is outlined in Figure 4B. The transition is triggered by the presence of UV light as indicated by the purple arrows. It is important to note that the signal to transition between states is the same for each condition. This indicates a significant difference from traditional field-responsive material systems that typically exhibit a single behavior in response to external stimuli.⁴⁰ In the presented soft robotic system, the response of the system results in a toggling behavior with two possible actions from the same input based on the present state of the system.

FIG. 4.

Summary of locomotive operation. (A) Schematic showing the oscillatory operation of the robot traveling towards the right. (B) The toggling function of the directional enforcement hard stop when triggered by UV light causes a state transition. (C) Oscillatory behavior of the robot traveling towards the left.

Figure 4A shows that the two legs are attached to the bottom of the processing subsystem rectangular frame by a thin section of material that acts as a compliant hinge. Magnets are placed in each leg and the rectangular frame to create a stable closed configuration when the legs are in contact with the processing subsystem. A locomotive LCE actuator fabricated in the same method as Figure 2C connects the two legs. The electrical contacts of the LCE actuator are connected to Ag-TPU traces along the leg and side of the rectangular processing subsystem. One of the conductive pathways leads to the input voltage (V_CC) while the other leads to the ground.

On both legs of the robot, the conductive pathways are completed by bridging the gaps between the top edge of the leg and the bottom edge of the processing layer. An electrical signal can be passed to the locomotive LCE actuator only when both legs are in the closed position shown in the top images of Figure 4A and C. When power is applied, the locomotive LCE begins to heat and apply a contractile force between each leg. As the force exerted by the LCE increases, the magnet inserts maintain the legs in a closed position until the force reaches a threshold of approximately 3.5N which overcomes the magnetic attractive force. When the threshold is reached, a single leg snaps to an open position and breaks the electrical contact of the Ag-TPU traces shown in the bottom images of Figure 4A and C. As the leg opens, the bottom of the leg slides along the surface upon which the robot sits without causing bulk motion of the robot. The break in electrical contact causes the LCE actuator to cool and relax which allows the leg to return to the closed position due to the magnetic attraction and release of strain energy in the compliant hinge. As the leg relaxes, it pushes the robot in one direction which facilitates locomotion. In the closed position, the electrical contact is restored, and the locomotion cycles indefinitely while power is applied.

The direction of travel is dictated by the directional enforcement hard stop as illustrated in Figure 4. This hard stop is a segment connected to the translating portion of the bistable mechanism in the processing subsystem. In each stable position of the bistable mechanism, the hard stop contacts the inner side of one of the legs. The contact introduces a bias to the force needed to open one of the legs. Without the hard stop, the force exerted by the locomotive LCE is experienced by each leg with identical actuation thresholds until one of the legs opens. The bias caused by the hard stop ensures that one leg has a lower threshold to open which allows for predictable actuation. As described in previous sections, shining UV light on the photo-responsive light causes the bistable mechanism to toggle to the opposite position which shifts the hard stop to the other leg which is outlined in Figure 4B. This causes the opposite leg to cyclically actuate which propels the robot in the opposite direction. Thus, the soft robot achieves bidirectional motion with the use of a single oscillating locomotive LCE actuator system enabled by the structural reconfiguration of the processing subsystem.

Experimental Operation of Soft Robotic Crawler

By attaching the soft robot to a constant voltage of 3V, the system autonomously operates as described in Figure 4. Figure 5 outlines the intelligent feedback of the robot as it exhibits an integration of sensing, processing, and acting. Figure 5A shows the timing of the locomotive LCE operation by reporting the voltage across the actuator during the operation of the robot for 10 cycles of heating and cooling of the LCE. When the legs are closed, the voltage applied to the robot is 3V. The LCE begins to heat as indicated by the red-shaded region which pulls the legs together. Once the leg breaks contact, no power reaches the actuator, so the recorded voltage drops to 0V which causes the LCE to begin to cool (blue-shaded section). The frequency of the locomotive cycle is directly related to the actuation and recovery of the LCE actuator. The heating periods record the time it takes for enough heat to disburse through the LCE material to cause the actuator to contract. As it cools, the recovery of the LCE is based on how quickly the material dissipates heat to the environment. Before the UV light is turned on (purple shading), the robot travels toward the right, and after the directional enforcement hard stop toggles, the robot travels to the left. Supplementary Movie shows the operation of the 10 cycles with the change in direction.

FIG. 5.

Summary of robot operation. (A) The timing diagram of leg operation reporting the voltage applied to the locomotive LCE actuator. (B) Photographs of locomotion to the right. (C) Demonstration of the switching behavior triggered by UV light. The dotted red lines show the edge of the leg where the hard stop (cyan box) makes contact. The thicker line shows the side on which the hard top is in the photograph. (D) Top–down view of memory subsystem during sequence shown in (C). LCE, liquid crystal elastomer.

The photographs in Figure 5B highlight the first two full cycles of locomotion towards the right. The green dotted line in each photo indicates where the left leg starts initially. In this sequence, the state of the processing subsystem, $Q_{n} = 0$ , indicates that the robot will travel to the right. When a voltage supply of 3V is applied at 0 s, the locomotive LCE begins to heat and contract. At 52 s, the force of the LCE applied between the legs causes the left leg to snap through to its open position. The bottom of the left leg slides forward, while the right leg remains stationary. Over the next 104 s, the left leg relaxes again to the closed position and pushes the robot to the right 18.1 mm. The next cycle lasts 182 s, and the robot moves 15.1 mm. The robot continues for 1302 s and completes 8 cycles in which the robot travels a total of 78.7 mm as seen in the first photograph of Figure 5C. These results demonstrate an average speed of 3.6 mm/min.

Figure 5C shows photographs of the direction change when the robot is exposed to UV light. The green dotted starting line now indicates the initial position of the right leg starting at 1302 s of operation. The cyan box in each photograph outlines the position of the directional enforcement hard stop. Red dotted lines indicate the edges of the legs where the hard stop makes contact to introduce the directional bias. The thicker line shows the side of the hard stop contacts in the photograph. A UV light (Thorlabs M365L2-C1) with an intensity of 50 mW cm⁻² is directed at the photo-responsive LCE on the top surface of the robot starting in the second photograph of Figure 5C and D shows photographs from the top perspective when the light is turned on. When the light is off, the photo-responsive switch is curled upwards, and the Ag-TPU network on the bistable mechanism is not connected to power. When the light turns on the switch bends down which creates an electrical connection. The left LCE in Figure 5D begins to actuate until the mechanism toggles to the other position at time 1389 s. When the light is turned off, the photo-responsive switch curls upward again which halts the toggle operation of the processing subsystem. In the third photograph of Figure 5C, the directional enforcement hard stop can be seen shifted to the left which allows the right leg to open due to the locomotive LCE. When the right leg closes, the robot is propelled towards the left.

To evaluate the performance and efficiency of soft robots, the cost of transport (CoT) is often used.^26,37,41 The CoT compares the power needed with the motion of the robot with Equation 1. It is important to note that in this demonstration, the soft robot operates with an external power supply. The addition of an onboard power supply would increase the mass and alter the performance of the robot. To accommodate an untethered operation, optimization of the body, legs, and feet of the system may be required. While the lack of onboard power makes comparison to fully untethered robots and biological locomotion difficult, the CoT here provides an initial benchmark to evaluate and elucidate future opportunities for the principles of locomotion with embedded intelligence. $CoT = \frac{\bar{P}}{mgv}$ (1)

In Equation 1, the average power $\bar{P}$ is used because the LCE is only active for part of the cycle while the legs are connected. The total mass $m$ of the robot in this work including the magnets, LCE actuators, and photo-responsive LCE is 77.6 g, gravity $g$ is 9.81 m/s², and the velocity of the robot $v$ is 0.06 mm/s. The average power $\bar{P}$ consumption is calculated using the duty cycle of the LCE actuator which is the ratio of the time the actuator is active and the average cycle time. Assuming a constant applied voltage of 3V, an average resistance of 1.25 Ω, a steady state period of approximately 163 s, and an active time of 55 s, the average power consumption per cycle is 2.4W. Therefore, the CoT of the presented robot is around 52,900. This magnitude of CoT is similar to previous soft robotic systems using thermally responsive actuators.^26,30,42,43 Such a high CoT value is likely due to inefficiencies in the thermal behavior of the LCE actuators and poor transmission of force to the movement of the legs.

Discussion

The primary focus of this study is the integration of intelligent behaviors including sensing, processing, and acting into soft robotic systems. While the current design serves as a proof of concept for such autonomous, intelligent behavior, several areas for improvement can enhance the efficiency and expand the capabilities of these systems. Alternate materials and geometric designs can be explored to alter the friction coefficients and gait of the locomotion which could improve the efficiency and performance of the robot. In addition, the TPU material and scale selected for the body of the robot concept presented here allow for versatility in the manufacturing process using readily available additive manufacturing technologies in a desktop FDM printer (Ender 3-V2). Furthermore, as a polyurethane-based substrate, the Ag-TPU conductive paste bonds effectively with the robot body ensuring reliable electrical networks. By exploring alternative manufacturing technologies, the principles demonstrated here may be applied at different scales.

As explored in the previous section, the CoT of the present robot is high, indicating inefficiencies. One way to improve the CoT of the robot is to increase the velocity. For the robot in this work, the velocity is directly related to the actuation and recovery rate of the LCE actuators. One strategy to increase the rate of actuation and recovery is a heating element that can better distribute the applied heat over the area of the LCE material. With a more efficient heat distribution, a lower temperature is needed to cause the orientation change in the LCE and therefore, it would take less time for the LCE to cool and recover. The rapid recovery would allow the locomotive cycle to increase in frequency causing the robot to move faster.

While the presented concept requires an external power supply, advances in self-powered devices may provide alternatives to create an untethered system. For example, work in soft photovoltaic materials,⁴⁴ could be used to sense optical input and use that input to power subsystems of the robotic system, thus increasing the autonomy of the soft robotic system.

This work has shown how the feedback system of sense-process-act serves as a foundational building block of intelligent material systems. To enhance autonomous intelligent behavior, the sensing and processing subsystems can be scaled to perceive more complex signals and act with more sophisticated decision-making. In the presented system, the sensing subsystem features a singular switch to perceive a binary light signal. Previous work by Tholen, et. al.³⁵ highlights how the integration of multiple switches can be used to understand digital light signals with multiple bits of information. The mechanism used in the memory subsystem can also be expanded by combining bistable mechanisms in series and parallel to store multiple states using techniques outlined by El Helou et. al.³⁶ The increased processing capability could also allow for more complicated self-feedback as the motion of the robot alters incoming signals. Such expansion in sensing and processing functions would require updates to the geometry and structure of the robot. By enhancing these systems, the robot can perform more complex tasks autonomously such as greater directional control, multimodal locomotion, and adaptation to terrain.

Additional opportunities to expand the potential of intelligent mechanical material systems are to enable feedback and interaction with multiple surrounding agents. The interactions can enable emergent behavior that exceeds the capabilities of a single robot. Such emergent behavior has been used to create systems that can learn and adapt,⁴⁵ and understand more about the nature of living plants and animals.⁴⁶ By increasing the intelligent behavior of individual units with methods explored in this work, it may be possible to greatly improve the capability of groups of intelligent systems as they interact one with another.

Conclusion

The presented robotic system encapsulates the fundamental elements of an intelligent mechanical system by integrating the necessary functions of sensing, information processing, memory, and action. The robot can sense optical UV stimuli using an optothermal LCE switch which can trigger a switch in the configuration of a bistable information processing subsystem. The mechanical configuration enforces the direction of locomotion by introducing a biased snap-through response in the locomotive subsystem. While power is applied to the robot, it will continually move in one direction until UV light is sensed which causes the robot to update and toggle to the opposite configuration and change direction. If the same UV light is subsequently applied, the robot will toggle and reverse directions again. The presented robot demonstrates a comprehensive soft intelligent system capable of autonomous operation without the use of traditional electronics. The multi-functional integration and locomotion highlighted in this work represent a step toward more practical intelligent soft robotic systems capable of navigating dynamic environments.

While the presented robot does not match the speed or efficiency of traditional robots, it demonstrates important principles of integrating intelligent functions in the design of the system to achieve more complex behaviors. Future investigations can focus on the optimization of the leg and foot mechanisms to improve the robot’s efficiency and performance. Furthermore, researchers can explore the integration of battery components to realize a fully untethered robotic system. Such intelligent behavior lays a foundation for robotic systems that can operate without the need for external control thus allowing them to operate in diverse environments including deep-sea and space exploration or disaster response scenarios.

Footnotes

Acknowledgment

The authors acknowledge helpful conversations with Ms. Haley Tholen and Dr. Kyung Min Lee and other researchers at the Air Force Research Laboratory over the duration of this work.

Authors’ Contributions

L.P.H., R.L.H., and J.J.B.: Conceptualization, methodology. L.P.H.: Investigation. L.P.H., P.R.B., R.L.H., and J.J.B.: Data analysis, writing—review and editing. R.L.H.: Funding acquisition. R.L.H. and J.J.B.: Project administration. P.R.B., R.L.H., and J.J.B.: Supervision.

Author Disclosure Statement

The Authors declare no conflicts of interest.

Funding Information

This research is supported in part by the Air Force Research Laboratory, the Air Force Office of Scientific Research, the National Science Foundation (CMMI-2314559), and the Army Research Office (W911NF-23-1-0314).

Supplementary Material

References

, Chen

, et al. An underwater bionic crab soft robot with multidirectional controllable motion ability. Ocean Engineering, 2023; 278:114412.

CSX

, Lum

. Untethered soft robots for future planetary explorations? Advanced Intelligent Systems, 2023; 5(3):2100106.

Kurumaya

, Phillips

, Becker

, et al. A modular soft robotic wrist for underwater manipulation. Soft Robot, 2018; 5(4):399–409.

Runciman

, Darzi

, Mylonas

. Soft robotics in minimally invasive surgery. Soft Robot, 2019; 6(4):423–443.

Yasa

, Toshimitsu

, Michelis

, et al. An overview of soft robotics. Annu Rev Control Robot Auton Syst, 2023; 6(1):1–29.

Rus

, Tolley

. Design, fabrication, and control of soft robots. Nature, 2015; 521(7553):467–475.

Whitesides

. Soft robotics. Angew Chem Int Ed Engl, 2018; 57(16):4258–4273.

Del Dottore

, Mondini

, Rowe

, et al. A growing soft robot with climbing plant–inspired adaptive behaviors for navigation in unstructured environments. Sci Robot, 2024; 9(86):eadi5908.

Kaspar

, Ravoo

, van der Wiel

, et al. The rise of intelligent matter. Nature, 2021; 594(7863):345–355.

10.

Jeong

, Kim

, Choi

, et al. Spikebot: A multigait tensegrity robot with linearly extending struts. Soft Robot, 2024; 11(2):207–217.

11.

CSX

, Tan

MWM

, Xu

, et al. Locomotion of miniature soft robots. Advanced Materials, 2021; 33(19):2003558.

12.

Calisti

, Picardi

, Laschi

. Fundamentals of soft robot locomotion. J R Soc Interface, 2017; 14(130):20170101.

13.

Joyee

, Szmelter

, Eddington

, et al. 3D printed biomimetic soft robot with multimodal locomotion and multifunctionality. Soft Robot, 2022; 9(1):1–13.

14.

Coyle

, Majidi

, LeDuc

, et al. Bio-inspired soft robotics: Material selection, actuation, and design. Extreme Mechanics Letters, 2018; 22:51–59.

15.

Wang

, Xiao

, Quan

, et al. Bioinspired soft spine enables small-scale robotic rat to conquer challenging environments. Soft Robot, 2024; 11(1):70–84.

16.

, Gao

, Tan

. Bioinspired 3d-printed snakeskins enable effective serpentine locomotion of a soft robotic snake. Soft Robot, 2023; 10(3):568–579.

17.

Drotman

, Jadhav

, Sharp

, et al. Electronics-free pneumatic circuits for controlling soft-legged robots. Sci Robot, 2021; 6(51):eaay2627.

18.

Shepherd

, Ilievski

, Choi

, et al. Multigait soft robot. Proc Natl Acad Sci USA, 2011; 108(51):20400–20403.

19.

Zou

, Lin

, Ji

, et al. A reconfigurable omnidirectional soft robot based on caterpillar locomotion. Soft Robot, 2018; 5(2):164–174.

20.

Lin

, Siddall

, Schwab

, et al. Modeling and control of a soft robotic fish with integrated soft sensing. Advanced Intelligent Systems, 2023; 5(4):2000244.

21.

, Hu

, Xie

, et al. Reconfigurable magnetic soft robots with multimodal locomotion. Nano Energy, 2021; 87:106169.

22.

Kong

, Zhu

, Yang

, et al. Integrated locomotion and deformation of a magnetic soft robot: Modeling, control and experiements. IEEE Trans Ind Electron, 2021; 68(6):5078–5087.

23.

Gupta

, Qin

, Wang

, et al. Soft robots based on dielectric elastomer actuators: A review. Smart Mater Struct, 2019; 28(10):103002.

24.

Cao

, Qin

, Liu

, et al. Untethered soft robot capable of stable locomotion using soft electrostatic actuators. Extreme Mechanics Letters, 2018; 21:9–16.

25.

Wang

, Zhang

, Tan

, et al. Electroactive polymer-based soft actuator with integrated functions of multi-degree-of-freedom motion and perception. Soft Robot, 2023; 10(1):119–128.

26.

Boothby

, Gagnon

, McDowell

, et al. An untethered soft robot based on liquid crystal elastomers. Soft Robot, 2022; 9(1):154–162.

27.

Xiao

, Jiang

, Zhao

. Liquid crystal polymer-based soft robots. Advanced Intelligent Systems, 2020; 2(12):2000148.

28.

Nie

, Wang

, Yang

. Structure-induced intelligence of liquid crystal elastomers. Chemistry, 2023; 29(38):e202301027.

29.

, Hong

, Zhao

, et al. Caterpillar-inspired soft crawling robot with distributed programmable thermal actuation. Sci Adv, 2023; 9(12):eadf8014.

30.

, Wang

, et al. Electrically controlled liquid crystal elastomer-based soft tubular actuator with multimodal actuation. Sci Adv, 2019; 5(10):eaax5746.

31.

Yasuda

, Buskohl

, Gillman

, et al. Mechanical computing. Nature, 2021; 598(7879):39–48.

32.

El Helou

, Buskohl

, Tabor

, et al. Digital logic gates in soft, conductive mechanical metamaterials. Nat Commun, 2021; 12(1):1633.

33.

El Helou

, Grossmann

, Tabor

, et al. Mechanical integrated circuit materials. Nature, 2022; 608(7924):699–703.

34.

Hyatt

, Harne

. Programming metastable transition sequences in digital mechanical materials. Extreme Mechanics Letters, 2023; 59:101975.

35.

Tholen

, Ambulo

, Lee

, et al. Optomechanical computing in liquid crystal elastomers. Soft Matter, 2023; 19(36):6978–6986.

36.

El Helou

, Hyatt

, Buskohl

, et al. Intelligent electroactive material systems with self-adaptive mechanical memory and sequential logic. Proc Natl Acad Sci U S A, 2024; 121(14):e2317340121.

37.

Thomas

, Germano

, Martinez

, et al. An untethered mechanically-intelligent inchworm robot powered by a shape memory alloy oscillator. Sensors and Actuators: A. Physical, 2021; 332:113115.

38.

, Yin

, Hua

, et al. A modular strategy for distributed, embodied control of electronics-free soft robots. Sci Adv, 2023; 9(27):eade9247.

39.

Schwartz

, Lagerwall

JPF

. Embedding intelligence in materials for responsive built environments: A topical review on liquid crystal elastomer actuators and sensors. Building and Environment, 2022; 226:109714.

40.

Pishvar

, Harne

. Foundations for soft, smart matter by active mechanical metamaterials. Adv Sci (Weinh), 2020; 7(18):2001384.

41.

, Gravish

. Multimodal locomotion in a soft robot through hierarchical actuation. Soft Robot, 2024; 11(1):21–31.

42.

Pang

, Xu

, Wu

, et al. A soft microrobot with highly deformable 3D actuators for climbing and transitioning complex surfaces. Proc Natl Acad Sci U S A, 2022; 119(49):e2215028119.

43.

Messner

PWC

, Paik

, Shepherd

, et al. Energy for biomimetic robots: Challenges and solutions. Soft Robotics, 2014; 1(2):106–109.

44.

Park

, Heo

, Lee

, et al. Self-powered ultra-flexible electronics via nano-grating-patterned organic photovoltaics. Nature, 2018; 561(7724):516–521.

45.

Oliveri

, van Laake

, Carissimo

, et al. Continuous learning of emergent behavior in robotic matter. Proc Natl Acad Sci USA, 2021; 118(21):e2017015118.

46.

Corucci

, Cheney

, Kriegman

, et al. Evolutionary developmental soft robotics as a framework to study intelligence and adaptive behavior in animals and plants. Front Robot AI, 2017; 4:34.