1. Introduction
Encompassing over 70% of the Earth’s surface, the ocean is a treasure trove of abundant resources and diverse ecosystems
[1]. Moreover, the effective management and conservation of marine resources depend on comprehensive data support and monitoring. However, the complexities of the environment, including ocean currents, darkness, and low temperatures
[2],
[3], increase the difficulty of marine surveys and resource development for human beings. Underwater robot emerged as a potential tool for long-duration operations in oceans, especially in the deep sea
[4],
[5]. Constructed using specific materials, an underwater robot can withstand low temperatures, high pressures, and other extreme conditions
[6],
[7],
[8]. In addition, because of their various sensors
[9],
[10], underwater robots can gather invaluable data in hydrology
[11], marine geology
[7], and marine biology
[12],
[13]. Besides, remote operation of robots can be realized owing to advanced wireless communication and automatic control technology. Thus, underwater robots can replace humans in missions such as marine exploration [14],
[15],
[16], safety inspections
[17],
[18], and seabed mineral extraction [19], which enhances scientists’ understanding of the marine environment and offers insights into oceanic resources and ecosystems
[20]. Given the advantages of automation and intelligence, underwater robots have become an inevitable choice for marine exploration and operation.
Miniature underwater robots (MURs), usually referring to those with dimensions of less than 20 cm, possess unique advantages over underwater macrorobots in terms of compact size, agile motion, and ease of deployment, especially for tasks in narrow underwater spaces. As shown in
Fig. 1 [6],
[15],
[21],
[22],
[23],
[24],
[25],
[26],
[27],
[28],
[29],
[30],
[31],
[32],
[33], for missions in confined environments, such as the detection of inaccessible rugged seabeds (slits, gullies, and caves) (
Fig. 1(a)) and inspection of pipe systems (
Fig. 1(c)), MURs can move freely in narrow spaces that are inaccessible to large-scale robots or humans
[6],
[24],
[26],
[27],
[28],
[34]. The compact size of MURs also allows them to contact underwater organisms with less impact, thereby nondestructively realizing plant sampling
[24] and object grasping
[25] (
Fig. 1(b)). Furthermore, the miniaturization of MURs implies a smaller turning radius and stronger motion feasibility. Hence, MURs can realize dexterous motion, including obstacle avoidance
[29], turning
[30], and swarming [31] (
Fig. 1(d)) in restricted regions. Moreover, the MUR propulsion systems do not generate substantial turbulence, which avoids scaring marine life. This enables MURs to integrate well into the marine environment and work well in concealed military antisubmarines. Additionally, compared with macrorobots, MURs offer greater convenience in terms of construction costs, transportation, and deployment. Hence, MURs, particularly bionic swimming robots that can integrate well into underwater habitats, are efficient solutions for high-precision tasks in small water spaces. With continuous technological development, it is believed that MURs will make considerable contributions to our understanding of the oceans, the protection of marine ecosystems, and the further exploitation of marine resources.Researchers have developed various types of MURs and applied them to various underwater tasks. Early MURs drew inspiration from fish that generated propulsion by swaying their caudal fin, such as servo-motor-driven robotic fish
[30],
[35]. The advent of smart materials has resulted in many flexible materials that exhibit bidirectional deformation, and are well-suited for actuating the caudal fin, dorsal fin, and body, such as ionic polymer-metal composites (IPMCs)
[36],
[37], shape memory alloys (SMAs)
[38],
[39], and dielectric elastomer actuators (DEAs)
[40]. The continuous development of material technologies has led to the emergence of various types of MURs, including bionic jellyfish
[41] and bionic turtle robots
[42]. Robot movements have also evolved from simple forward motion to three-dimensional (3D) flexible motion and have mimicked various marine animal locomotion modes, such as jetting and paddling. In recent years, with advancements in new materials, sensors, and other technologies, researchers have dedicated their efforts to developing new manufacturing methods and driving structures to improve the miniaturization, flexibility, and intelligence of underwater robots, such as swarm-capable robotic fish
[31], bioinspired octopus robots capable of object manipulation
[43], and deep-sea cruising robots
[6].
Fig. 2 presents a chronological timeline of MUR evolution spanning from 2005 to 2024
[6],
[21],
[29],
[30],
[31],
[32],
[33],
[35],
[36],
[37],
[38],
[40],
[41],
[42],
[44],
[45],
[46],
[47],
[48],
[49],
[50],
[51],
[52],
[53],
[54],
[55],
[56]. It is evident that the progression of MURs has transitioned from a simplistic imitation of fish-like motion patterns to integrated composite movement modes and multifaceted functionality encompassing underwater cruising, manipulation, and sensing.
In this article, we summarize the actuation and locomotion methods employed by existing MURs. In Section 2, the MUR actuation methods are categorized into four types: motors, magnetic field actuators, piezoelectric actuators, and soft materials. In Section 3, we focus on the various MUR locomotion modes, including fish-inspired, jetting, paddling, crawling, flagellar-inspired, and hybrid modes. Finally, in Section 4, we summarize the existing challenges associated with MURs and explore the potential future development trends.
2. Actuation method
Considering the rich variety of MURs, we classified the robots based on four distinct actuation methods: motors, magnetic field actuation, piezoelectric actuators, and soft materials. A comparison of the MURs employing various actuation methods is presented in
Table 1 [6],
[25],
[29],
[30],
[31],
[32],
[33],
[36],
[37],
[38],
[40],
[41],
[42],
[44],
[45],
[46],
[47],
[48],
[49],
[50],
[51],
[57],
[58],
[59],
[60],
[61],
[62],
[63],
[64],
[65],
[66],
[67],
[68]. We used the dimensionless index cost of transport (COT) to represent the robot actuation efficiency. The COT is the energy expenditure as the robot transitions from one location to another and is calculated as
, where
P denotes the power cost during transportation,
m denotes the mass of the robot,
g is the gravitational acceleration, and v is the forward speed
[69],
[70].
2.1. Motors
The use of a motor is the most traditional and primitive MUR actuation method, with the advantages of high speed, significant torque, and high reliability. Over the past few decades, many researchers have used motors in conjunction with different drive structures to achieve the movement of MURs.
Although the motors produce only rotating motion, the combination of motors with various actuating mechanisms enables multiple locomotion modes. Early motor-driven MURs primarily used a servo motor to actuate the caudal fin (
Fig. 3(a)) for propulsion
[30],
[35],
[57],
[58]. In Ref.
[71], a robotic fish with a direct current (DC) motor-actuated propeller and a servomotor-actuated caudal fin was developed to realize 3D motion. Recently, researchers developed diverse actuating mechanisms using motors to achieve other locomotion modes. For instance, In Ref. [72], a robotic octopus with servo-motor-driven arms was developed (
Fig. 3(b)). The tentacles of the robotic octopus can move up and down, actuated by a servo motor to the paddle, and produce a propulsion force. The paddling leg of a water-beetle-inspired robot (
Fig. 3(c)) was actuated through the connection of gears and springs
[59] with a maximum moving speed of 75.4 mm·s
−1. Similarly, the jetting motion of shellfish is mimicked by a DC motor in RoboScallop
[44], which also achieves a high motion speed of 2 BL·s
−1 (BL: body lengths) (
Fig. 3(d)). Wang et al.
[32] used an eccentric motor in combination with a rigid–flexible hybrid (RFH) driving device to generate vibration motion, thereby enabling the robot to achieve amphibious movement on both land and water (
Fig. 3(e)). The speed of movement on land and water surfaces can reach up to 815 and 171 mm·s−1, respectively, which is the highest underwater body length speed (2.3 BL·s
−1) in existing MURs.
Although motor-based actuation is a mature and efficient driving method, its working characteristics and structure limit its application in MURs, particularly in soft MURs. First, motors are only capable of generating angle changes, necessitating the design of special mechanical structures to translate the motor motion into robotic motion. This considerably increases the complexity of mechanical structures and enhances the rigidity of the robot. Second, many motors and the circuit boards used for driving and controlling them are not waterproof. Therefore, a waterproof design needs to be a focal point when designing a robot [73],
[74], especially for applications in deep-sea environments that endure high pressures. In addition, concerns, such as motor noise and torque dissipation, must be addressed when adopting motor driving.
2.2. Magnetic field actuation
Magnetic field actuation leverages the principle of electromagnetism to transform electrical energy into mechanical movement
[75]. It uses the power coil of a wire or permanent magnet to generate a magnetic field that interacts with a magnetic component to produce robot motion. Magnetic field actuation has several merits, making it suitable for propelling MURs. First, it offers precise control over generated motions, such as vibrations, allowing for accurate positioning and operation. Second, the magnetic field actuation responds rapidly to the voltage signal [76], providing excellent dynamic performance of the robot. The magnetic field actuation can then be separated from the robot, thereby enabling remote actuation and wireless control.
The typical structure of magnetic field actuation is illustrated in
Fig. 4 [28],
[31],
[52],
[77].
Figs. 4(a) and (b) show a robot with a magnetic composite actuated by a Helmholtz coil system [77]. The electromagnetic coil generates a magnetic field and causes the bending of the composite. The tentacles then deform and paddle for motion. Figs. 4(c) and (d) illustrate a robot with an embedded magnet actuated by electromagnetic coils [28]. The coils generate a rotating magnetic field, and the robot rotates under the magnetic force. Fig. 4(e) displays the actuation by the onboard magnetic coil. When a voltage is applied to the electromagnetic coil, it generates a magnetic field that causes the permanent magnet to flip, driving the fin to oscillate and producing propulsion that moves the robot forward. The oscillation direction and frequency of the fin can be controlled by adjusting the direction and frequency of the voltage applied to the electromagnetic coil. Berlinger et al. [31],
[52] successively developed two MURs that utilize this actuation, as illustrated in
Figs. 4(f) and
(g). The robot’s caudal, pectoral, and dorsal fins are all magnetic-field-driven, with the caudal fin responsible for propulsion and the pectoral and dorsal fins contributing to the 3D motion. Benefiting from the low energy consumption and small size of this actuator, a robot measuring only 130 mm can achieve a forward speed of 1.15 BL·s
−1 and diving speed of 75 mm·s
−1 with a battery for run time of up to 2 h.Although magnetic field actuation has proven to be advantageous for robotic fish in terms of its compact size and high energy efficiency, its motion patterns are relatively limited. For onboard magnetic field actuation, owing to the fixed and unidirectional characteristics of the generated magnetic field, the motion modes of driving mechanisms, such as fins, are confined to fixed and unidirectional motion. The addition of supplemental driving structures is necessary to achieve multi-degree-of-freedom motion. However, this is not ideal for generating motion in highly flexible underwater robots because they require more degrees of freedom to accomplish complex movements. In remote magnetic field actuation, the intensity of the magnetic field is low when the robot is far from the coils. Therefore, the movement of this type of robot is restricted to long voyages.
2.3. Piezoelectric actuation
Among all actuating elements, piezoelectric actuators stand out for many reasons, including their compact and simple structure, high power density, fast response, high resolution, and lack of electromagnetic interference [78],
[79],
[80],
[81],
[82],
[83]. These unique advantages make piezo actuators suitable for MUR development and have attracted the attention of many scholars
[84],
[85],
[86],
[87].
A piezoelectric actuator converts electrical energy into mechanical energy owing to the inverse piezoelectric effect [88]. The typical structure of the bimorph piezoelectric actuator is shown in
Fig. 5(a). The bimorph piezoelectric actuator consists of two piezoelectric layers and a sandwiched resistive layer. The actuator bends when voltage is applied to the piezoelectric layer. By combining this with the transmission mechanism (
Fig. 5(b)), the actuator can realize paddling of the flaps
[46]. Another type of piezoelectric motor (Fig. 5(c)) was developed to actuate an MUR
[63]. The piezoelectric motor consists of a piezoelectric wafer as the stator and a rotor. When an increasing signal is applied to the piezoelectric wafer, a twist motion occurs between the neighboring wafers, and the upper wafer rotates at a small angle. Because of the friction force, the rotor rotates together. During the signal decrease, the wafer moves back sharply and the rotor remains steady due to the inertial force. Thus, a continuous rotational motion of the rotor is achieved. This actuator weighed only 3.6 g with a size of 20 mm × 20 mm and a steady speed of 2200 r∙min−1. Actuated by this motor, the propeller produced a thrust force of 4.6 mN to propel the robot, whose velocity can reach 80 mm·s
−1.
In addition to piezoelectric motors, piezoelectric vibrators are another efficient method for actuating MURs. A typical piezoelectric actuator is shown in
Figs. 5(d) and
(e) [29]. The actuator consists of two shells and a piezoelectric vibrator. When a signal is applied to the vibrator, a high-frequency vibration motor produces flow fields. By adjusting the frequency and waveform of the driving signal, the actuator realizes suction and ejection motions to propel the robot.
Fig. 5(e) presents an antihydropressure miniature cross-shaped underwater robot (CSUR) with multi-locomotion modes using piezoelectric pulsed-jet actuators (PJAs). The CSUR consists of four piezoelectric PJAs that mimic the suction and ejection motions of jellyfish. Thus, the CSUR can achieve 3D motions, including going straight, turning, rotating, sinking, and hovering, with the maximum linear velocity reaching up to 47.8 mm·s−1 (0.44 BL·s
−1).
MURs with piezoelectric actuators have several advantages, such as simple structure, compact size, good waterproofing feasibility, and deep-sea working properties. However, a high driving voltage (several hundred volts) [24],
[89] requires unique circuits and equipment and increases the difficulty in designing cableless MURs, limiting their usage in remote and deep-sea applications. The tethered driving and limited deformation of the piezoelectric actuators are also shortcomings. In addition, the strain produced by piezoelectric actuators is not large compared with that produced by soft materials.
2.4. Soft materials
In contrast to the aforementioned rigid actuation methods, soft materials have unique advantages in terms of flexibility, elasticity, and mechanical properties, making them popular research topics for driving MURs
[90],
[91]. By using flexible and stretchable elastomeric materials
[92],
[93] such as silicone rubber, MURs can deform and absorb energy from collisions, while maintaining good performance in dynamic deformation, high conformal capability, and high-pressure environments [94],
[95]. Currently, various new types of smart flexible materials have been used to drive MURs, such as SMAs
[38],
[39],
[53],
[54],
[55],
[56], DEAs
[40],
[64],
[65],
[66], IPMCs
[36],
[37],
[49],
[67], and pneumatic actuation
[22],
[50]. In this section, we introduce the working principles of various soft materials and summarize their corresponding actuation methods.
2.4.1. SMAs
SMAs are widely used in robot actuation because of their excellent flexibility, light weight, small size, variability in stiffness, and ability in sensing and actuation
[96],
[97],
[98],
[99]. An SMA is a special type of metallic material that exhibits a shape-memory effect, in which the material returns to its initial shape and size when heated beyond a specific transformation temperature
[100],
[101],
[102]. The working principle of the SMA is illustrated in
Fig. 6(a). The SMA was initially in a twinned martensitic state. Subsequently, a load was applied to the SMA until it exhibited the desired deformation (detwinned martensitic state). Then, the loading was removed, and the SMA was heated until it completed the transformation to the austenitic state. At this point, the SMA returned to its original shape. Finally, the SMA was cooled and transformed into a twinned martensitic state. By training the SMA using the above cycle several times, the SMA processes the two-way shape memory effect, which behaves as the original shape when heated and the deformed shape when cooled. Thus, the SMA is ideal for generating stretching or shrinking actions for MURs through cooling and heating cycles.Over decades of study, researchers have developed various MURs actuated by SMAs. Among these, the robotic fish predominated because the bidirectional deformation of the SMAs was suitable for mimicking the beating motion of the fish’s caudal fin. For example, Wang et al.
[38] proposed a robotic fish with a biomimetic fin actuated using an SMA wire (Fig. 6(b)). The tail fin contains SMA wires on both sides. When pulse signals are applied to the SMA wires on one side of the fin, the SMA wire bends owing to the shape-memory effect, and the fin deforms. Consequently, the SMA regains its original shape, and the fin returns to its original position owing to the storage of elastic energy when the signal is removed. Thus, the fin can produce a continuous bending motion and propel the robot forward using a pulse signal. In addition to a robotic fish, SMAs can also be used to mimic the paddling motion of starfish. Jin et al. [103] developed a starfish robot actuated using SMA wires, as shown in
Fig. 6(c). It comprises five smart modular structures, and each structure consists of four SMA wires connected by a printed circuit board (PCB) as the skeleton, with a polyvinyl chloride (PVC) plate embedded in it. When a heating signal is applied to this structure, the SMA wire bends, effectively storing elastic energy in the PVC plate and allowing bending of the structure. As the voltage drops, the PVC plate returns to its initial condition, and a recovery motion of the structure occurs. Thus, the motion frequency of the smart modular structure can be changed by adjusting the regulatory heating signal.The numerous advantages of SMAs make them one of the most suitable soft materials for robotic actuation. Furthermore, researchers have focused on combining sensing and actuation with a simple SMA structure
[104]. However, the development of the SMA actuators presents significant challenges. The deformation motion of the SMA actuator is highly influenced by temperature, which constrains its practical application, for instance, in severe conditions such as deep sea or South Pole environments. In addition, because transformations involving voltage signals and temperature occur during deformation, modeling the procedure and controlling the deformation accurately become highly inexact and complex
[105],
[106]. In the future, the development of precisely controlled strategies for SMA-actuated driving mechanisms will be a trend toward the realization of MURs with multiple degrees of freedom.
2.4.2. DEA
DEAs remain among the most common soft material alternatives because of their notable electrical deformation, high energy density, rapid response ability, and high mechanical energy density [107],
[108],
[109],
[110]. The DEA has a fundamental structure that features three layers: two stretchable electrode layers and one elastomer membrane, as depicted in Fig. 7(a). Typically, the elastomer membrane is sandwiched between two stretchable electrodes. When a voltage is applied, electrostatic pressure results from the electro-field interaction, squeezing the elastomeric membrane and resulting in shrinkage in thickness and stretching in the membrane plane directions [111],
[112]. By adjusting the structural configuration and adding a fixed part, the DEA exhibits specific unidirectional stretching and intra-shaped bent movements. DEAs possess unique and smart polymer material features and are capable of achieving multidirectional and multimodal deformations that are successively employed for driving MURs. As an important form of underwater robot, robotic fish also use DEAs as a common driving method. Shintake et al.
[64] used a double-layer DEA structure to mimic the movement of a caudal fin, thereby propelling a bionic robotic fish forward. Christianson et al.
[40] used a double-layer multistage DEA to mimic the movement of a leptocephalus. By applying voltage to different segments of the two-layer DEA actuator, it is possible to achieve morphological changes in the robot imitating the leptocephalus and thus simulate the movement of the leptocephalus. Li et al. [6] developed a biomimetic snailfish robot using a DEA to mimic the beating movement of snailfish fins (Fig. 7(b)), thereby enabling the robot to move in the Mariana Trench. In addition to robotic fish, various bioinspired underwater robots have been studied. Tang et al.
[48] used a DEA to imitate the leg movements of a frog, as shown in
Fig. 7(c). The robot consists of two bionic frog legs, each of which is composed of two DEAs to mimic the movement of the frog legs and webbed feet. When voltage is applied to the frog-leg joint DEA, a controllable bending movement of the leg joint can be achieved, and the wide area of the webbed foot can change when the voltage is applied to the webbed-foot DEA. Therefore, by adjusting the voltage and frequency values applied to the two DEAs, a frog-like gait can be achieved with the legs creating propulsion by swimming and propelling the robot forward. Under the condition of a driving voltage of 5 KV and a frequency of 0.25 Hz, the swimming speed of the bionic underwater robot can reach 19 mm·s−1. Godaba et al.
[65] developed a bionic jellyfish robot as shown in
Fig. 7(d). This robot consists of a bell-shaped shell, a DEA, and an air cavity. When a voltage is applied to the DEA, it bends and deforms, causing the air cavity to expand, expelling water from the robot’s interior and increasing buoyancy, allowing the robot to float and move upward. The upward and downward movements of the robot can be controlled by adjusting the voltage applied to the DEA, and the moving speed reaches up to 90 mm·s−1.
Despite the considerable utility of DEAs in driving MURs, their effectiveness is typically contingent on high-voltage stimulation in the kilovolt range
[113], necessitating supplementary circuitry design. Such circuits can limit the miniaturization and scope of DEA-powered underwater robots.
2.4.3. IPMC
IPMCs exhibit unique characteristics, such as low weight, high flexibility, low activation voltage, and significant bending strain [114],
[115]. An IPMC is composed of two layers of noble metal and an intervening ionic polymer membrane. Its working principle is shown in
Fig. 8(a). The central ionic polymer membrane is permeated with randomly distributed hydrated cations. Upon the application of voltage across the IPMC, the hydrated cations and water molecules migrate toward the cathode. This migration leads to an anisotropic concentration distribution of cations and water molecules, causing expansion in the cathode region and bending deformation [116]. In addition, when an external force is exerted on the IPMC, ion migration occurs because of the strain gradient coupled with the potential difference between the two layers. This capability allows the IPMC to detect deformations, thereby enabling it to function as an effective sensing component. In recent decades, researchers have developed various MURs that are actuated using IPMCs. Owing to the ability of the IPMC for bidirectional motion, it is exceptionally suitable for mimicking the movement of fish fins. Guo et al.
[67] developed a fish-like, S-shaped swimming miniature robot propelled by an IPMC actuator (
Fig. 8(b)). The IPMC serves as the caudal fin and bends rapidly to propel the robot. Fine-tuning the driving frequency and amplitude enabled precise control of the forward swimming speed of the robot. Beyond lateral swinging, the IPMC is adept at actuating the caudal fin in an up-and-down motion, as demonstrated by the dolphin-like robot
[37] shown in
Fig. 8(c). Shen et al.
[49] designed a biomimetic underwater robot (Fig. 8(d)) that emulated the motion of a cuttlefish using six pairs of IPMC actuators for propulsion. Through experimental data and analysis, a sinusoidal wave voltage is applied to each IPMC, thereby facilitating the achievement of the desired fin waveform and moving speed of 53.3 mm·s
−1.
However, the IPMC actuation is subject to a phenomenon known as relaxation, which causes significant instability during propulsion
[117],
[118]. When a DC voltage is applied to the IPMC, the density of the hydrated cations and water molecules is distributed nonuniformly, and small leakage of water molecules occurs. This leakage increases over time, causing the deformation to become unstable, and the IPMC gradually reverts to its original shape, in a process called relaxation
[119]. Although altering the solvent in the middle ionic polymer membrane can mitigate this back-relaxation
[120],
[121], it cannot be eliminated. In addition, the output force generated by the IPMC actuators is relatively low
[122], restricting the performance of miniature robots, especially in high-velocity fluid environments.
Although the use of flexible smart materials to drive underwater robots has the advantages of simple structure and small size, limited motion modes (mostly biomimetic fish), poor motion controllability, and accuracy hinder their applications. Improving the deformation precision of smart materials and agility of robots remains a major challenge.
2.5. Others
In addition to the four innovative actuation methods mentioned above, scholars have explored alternative actuating mechanisms for MURs, including pneumatic and hydromechanical systems. These systems function by altering the shape of the internal cavity upon inflation, thereby generating propulsion. This can be observed in applications such as the pneumatic actuation of bioinspired frog legs
[23], biomimetic robotic rays
[123], and hydraulic robotic jellyfish
[22]. However, pneumatic/hydraulic actuation often requires supplementary inflation devices, which pose challenges in achieving fully wireless MURs. Concurrently, the extent of deformation and stress achievable by these pneumatic cavities is limited, and the associated actuation devices tend to be comparatively bulky, hindering efforts to create miniaturized systems.
In addition, optical and thermal field actuations are promising for underwater robots owing to their remote actuation and selective control. When light is applied to a robot, it absorbs optical energy and converts it to thermal energy. The heat then leads to the recoverable reconfiguration of carbon, liquid crystal elastomers, or other heat-responsive materials [124]. Because of the photothermal effect, deformation occurs on the film, producing an actuation force. By changing the reconfiguration of heat-responsive materials, a robot can realize forward [125], rotating
[126],
[127], and vertical motions
[128].
The moving speed (unit: BL·s
−1) and body length of the existing MURs are summarized in
Fig. 9. Currently, the maximum speed (2.3 BL·s
−1) is achieved by an amphibious robot with eccentric motors
[32], followed by a RoboScallop (2 BL·s
−1) with a DC motor
[44]. Other motor-driven robots also exhibit considerable motion speed, from 0.35 to 0.79 BL·s
−1 [30],
[35],
[57],
[58],
[60]. In addition, the MURs actuated by electromagnet show remarkable maneuverability with speed from 0.3 to 1.15 BL·s−1 [31],
[45],
[52],
[61],
[72]. For smart materials, the performance is less impressive (SMAs
[38],
[39],
[41],
[42] from 0.14 to 0.77 BL·s
−1, DEAs
[33],
[64],
[65],
[66] from 0.23 to 0.69 BL·s
−1, IPMCs
[36],
[37],
[67] from 0.12 to 0.51 BL·s
−1).
Currently, although existing actuation methods have demonstrated impressive driving performance, they still encounter challenges in areas such as stress, strain, and power supply and fail to fully satisfy the demands of multifunctional MURs for flexible movement, low power consumption, and extended endurance. In the future, researchers must focus on developing new intelligent materials and actuation mechanisms to improve the functionality and intelligence capabilities of robots; for instance, the development of reprogrammable materials that can bend, twist, and stretch could result in enhanced flexibility, faster response, and greater stress. Additionally, the fusion of actuation mechanisms with sensor modules is a promising research topic. Furthermore, advancing toward reduced power consumption while achieving a self-sustaining energy supply remains a key objective for boosting robot endurance and operational capacity.
3. Locomotion
MURs are widely used in underwater environmental surveys and biological monitoring. The unique characteristics of these tasks impose stringent requirements on MUR mobility, including rapid speed, minimal environmental impact, and robust maneuverability. In recent years, inspired by aquatic organisms [10], scientists have developed various materials and structures to meet the requirements of specific applications. This section provides an overview of the current state of MURs and their locomotion modes, and explores future development trends.
According to the modes of locomotion of aquatic organisms, the locomotion of MURs is divided into six types: fish-inspired, jetting, paddling, crawling, flagellar-inspired, and hybrid modes, as shown in
Fig. 10. The performances of the MURs with different locomotion modes are summarized in
Table 2.
3.1. Fish-inspired mode
Over billions of years of evolution, fish have developed bodies that have adapted exquisitely for underwater navigation. Consequently, numerous MURs have adopted fish as biomimetic models to replicate their structural designs and kinetic patterns. Based on these motion paradigms, biomimetic fish-inspired MURs are categorized into two distinct types: BCF mode and MPF mode.
3.1.1. BCF mode
The BCF mode represents the standard movement pattern employed by biomimetic fish-like MURs. This mode imitates fish locomotion, predominantly using the BCF for propulsion and maneuvering in aquatic environments [129]. Depending on the various driving components and alterations in the body shape during movement, the BCF mode can be further subdivided into undulatory and oscillatory types. These encompass anguilliform, subcarangiform, carangiform, and thunniform patterns [130], as illustrated in
Fig. 11. The blue line in
Fig. 11 represents the amplitude of the swing of the robots with different BCF modes. For the anguilliform, the robot flaps almost its entire body to produce a traveling wave. However, for thunniform, the robot uses only the tail to produce waves, and the body is relatively steady. The anguilliform type primarily mimics the movement of an eel, with robots adopting this pattern displaying an overall undulating motion that resembles an “S” shape, effectively propelling the body forward [131]. This new material has natural flexibility and can generate a large driving force, making it suitable for powering biomimetic robots. Christianson et al. [40] developed a frameless DEA for propelling biomimetic eel robots. This novel approach differs from traditional sandwich DEAs by utilizing an internally filled liquid chamber as one electrode and the surrounding environmental liquid as the second electrode, thereby streamlining the design of soft-drive diving devices. This robot achieved a maximum swimming speed of 1.9 mm·s−1 and maintained an average transmittance of 94%, offering a significant stealth capability. Biomimetic eel-like robots require high joint flexibility, body flexibility, and driving torque. This presents significant challenges for the miniaturization and implementation of underwater biomimetic robots. Meanwhile, the autonomous motion of the anguilliform-type robot requires wireless and onboard control, and the integration of a rigid control board and a soft swing body is a problem. With ongoing advancements in flexible materials, it is anticipated that various anguilliform-type MURs will be developed in the future. Robots employing subcarangiform and carangiform swimming styles generate propulsion through body oscillations. The key distinction lies in the body segments used for oscillation; subcarangiform robots oscillate the front half of their bodies, whereas carangiform robots utilize the rear third for propulsion
[132]. Compared with the locomotion pattern of biomimetic eel robots, those using subcarangiform and carangiform robots exhibit smaller body-swing amplitudes. However, by increasing the oscillation frequency, they still achieved rapid underwater swimming.
Subcarangiform-type robots bend at least half of their bodies and coordinate with tail-fin oscillations to generate propulsion. This requires a high degree of body flexibility. Currently, these robots predominantly utilize innovative materials for propulsion, including SMAs and DEAs. Shintake et al.
[64] employed a dual-layer DEA structure to construct the rear of a robotic fish. By controlling the driving voltage, independent activation of the two DEA layers was achieved, allowing left and right swinging of the body and tail fin. However, this design is only a preliminary verification of the use of a DEA for robotic fish propulsion and does not achieve wireless control or sensing capabilities. SMAs, known for their high flexibility and bidirectional deformation properties, are particularly suitable for driving subcarangiform-type underwater robots. Chen et al. [39] as well as Muralidharan and Palani
[53] successfully utilized SMA actuators to enable the rapid movement of MURs, achieving speeds of 65.2 and 24.5 mm·s
−1, respectively. Compared with flexible materials, the traditional actuation mode does not have an advantage in miniaturized design. Although Romano et al.
[45] used a DC motor and magnets to achieve biomimetic oscillation of the subcarangiform type, it still required an external power source and lacked onboard control.Compared with robots utilizing subcarangiform movements, carangiform-type robots require oscillation over a smaller portion of their body, simplifying the driving mechanism to just the rear half of the body. Most current carangiform-type MURs employ servo motors as the actuators. By positioning a servo motor near the tail, these robots can oscillate the rear third of their body along the tail fin, producing the propulsion necessary for forward movement. Epps et al.
[58] designed a carangiform-type miniature robot measuring 148 mm in length that achieved flexible body bending and tail fin oscillation for propulsion using servo motors and mechanical structures. However, this robot lacks onboard control and power supply units, which precludes autonomous wireless operations. Kopman and Porfiri
[30] later developed a wireless underwater robot using a similar servo motor mechanism, equipped with a control board, power supply, and communication modules, thereby enabling autonomous operation and remote control. With the miniaturization of electronic components, MURs are now capable of integrating more advanced features. For instance, Zhao et al. [57] outfitted a robotic fish with a camera, allowing for the real-time acquisition of underwater environmental data and opening new possibilities for detection and exploration tasks. In terms of new flexible materials, Wang et al.
[38] used SMAs as actuators to achieve the 2D movement of a miniature robot fish, demonstrating the potential of innovative materials for enhancing robotic performance and capabilities.
Robots employing thunniform movement primarily generate propulsion through the left and right swinging of the caudal fin while maintaining a largely stationary body
[35],
[133]. Compared with the above-mentioned three BCF modes, the thunniform mode only requires tail oscillation, resulting in a simpler motion pattern and a more concise structural design. Consequently, this was one of the earliest types of biomimetic robotic fish to garner research interest. Early thunniform-type robots were primarily actuated by servo motors for tail fin oscillation, as demonstrated in Ref.
[35]. In addition, new flexible materials such as IPMCs
[36] have proven to be effective for driving oscillating caudal fins. Magnetic field actuation has emerged as a popular and efficient method for driving thunniform-type robotic fishes in recent years. This approach involves equipping the caudal fin with electromagnetic coils and permanent magnets. By applying voltage, the direction and strength of the magnetic field are altered, changing the position of the permanent magnet and enabling left and right caudal fin oscillations. Phamduy et al. [62] designed an electromagnetically driven robotic fish with a length of only 6.6 cm to verify the feasibility of this driving method. Yan et al.
[134] achieved a wireless robotic fish based on an electromagnetic drive. Chen et al. [135] developed a miniature robotic fish and proposed a magnetically actuated pulse-width modulation method to achieve high maneuverability. Berlinger et al. [31] developed a miniature robotic fish equipped with a camera using magnetic propulsion. This technology enables control over the angles and frequencies of the caudal, pectoral, and dorsal fins, facilitating 3D underwater movement (with a forward speed of 1.15 BL·s
−1 and diving speed of 75 mm·s
−1) and collective behavior. These advancements have significant potential for application in environmental monitoring and exploration, particularly in coral reefs and coastal environments.
Although robots in the BCF mode have advantages in long-range high-speed cruising and acceleration performance, they have poor maneuverability. Robots using the BCF mode, particularly the thunniform mode, typically have a swing tail with fewer degrees of freedom. Hence, it is difficult to achieve pitching and tumbling motions. In addition, this type of robot processes a larger turning radius and cannot realize in situ turning because the swing amplitude of the tail/body is limited. To address this challenge, it is necessary to integrate additional actuation mechanisms that mimic pectoral fins. However, this improvement requires complex driving units and advanced control systems, which increase the difficulty of manufacturing and maintenance. In addition, the actuating and control module of a robot using the BCF mode is concentrated; therefore, a high-pressure working environment, such as the deep sea, is a challenge. Currently, robots in the BCF mode have realized 3D motion, swarm motion, and lightweight target tracking. In the future, robots will be equipped with more sensors to monitor, inspect, and sample aquatic environments.
3.1.2. MPF mode
The MPF mode generates propulsion through paired pectoral, dorsal, and pelvic fins, or a combination of dorsal and pelvic fin undulations or oscillations
[136]. Compared to the BCF mode, the MPF mode has a higher propulsion efficiency at low speeds and superior maneuverability, allowing for capabilities such as hovering or on-the-spot turning. Typical body movement types for the MPF mode include rajiform, diofontiform, amiiform, moblastiform, balistiform, tetraodontidae, and labriform [137]. Similar to the BCF mode, the MPF mode can be divided into the undulatory and oscillatory types, with the first five categories employing undulatory propulsion, and the latter two relying on pectoral and dorsal fin oscillations.
Currently, MURs using the MPF mode primarily mimic the rajiform type, which emulates ray movement. Robotic fish using this mode generate propulsion through pectoral fin undulation, resulting in rapid movement and high propulsion efficiency. Li et al.
[6] and Li et al
[33] developed two MPF-type underwater soft robots using dielectric elastomer (DE) membranes. The actuator consisted of two layers of DE membranes sandwiched between the hydrogels. When a voltage is applied to the DE membranes, the entire actuator bends, causing undulation of the pectoral fins on both sides, thereby generating propulsion. Robotic fish can reach a movement speed of 0.69 BL·s−1 with a long endurance of up to 3 h
[33]. Building on this initial design, the team further optimized the robot structure to withstand the high pressures of the deep sea, enabling it to dive into the depths of the Mariana Trench
[6].
The MPF movement pattern for miniature bioinspired robots has been largely confined to the rajiform pattern, with a few other patterns being extensively utilized. Most MPF patterns involve propulsion driven by the pectoral or dorsal fins. However, the limited surface area of these fins limits the amount of propulsive force that can be generated. This configuration offers high agility and is efficient for short-distance, low-speed swimming but poses challenges in achieving long-range and high-speed cruising capabilities. Thus, robotic fish in the MPF mode have been limited to migratory swimming. More importantly, a robot using the MPF mode requires the cooperation of multiple actuation units and a precise control system, thereby increasing the complexity of the structure and manufacturing difficulty. Meanwhile, because the precise control of each unit and prediction of the motion are extremely difficult, the control precision of the robot in the MPF mode is not very high.
3.2. Jetting mode
Jetting is a common locomotion mode in aquatic environments in nature and is utilized by organisms such as jellyfish and octopuses
[138]. For example, an octopus has internal chambers for rapid movement. It fills its body cavity with water and forcefully expels it through a muscular funnel, known as a siphon. The octopus is propelled in the opposite direction by forcefully expelling water in a specific direction, enabling swift motion. The agility and speed of jet propulsion are enhanced by the ability to adjust the shape of the siphon, which allows control over the direction and strength of the jet. In addition, some mollusks generate water jets for propulsion by rapidly and periodically opening and closing their shells. Jetting propulsion offers the advantages of high speed and strong maneuverability, making it the primary bioinspired locomotion pattern for MURs. Currently, MURs using jetting propulsion often mimic jellyfish by employing new flexible materials, such as SMAs [41] (
Fig. 12(a)) or DEAs
[65] (
Fig. 12(b)), to manufacture and control chambers for water-jetting propulsion. Some researchers
[44] imitated the jetting mechanism of bivalve mollusks by attaching a DC motor to the crank mechanism. This setup controlled the opening and closing of the shell of the robot, facilitating the rapid closure and production of high-speed water jets for propulsion, as illustrated in Fig. 12(c).
Although the jetting mode can generate a powerful force, the control precision of the robot using the jetting mode is lower than that using the BCF or MPF modes. A higher propulsive force leads to a higher moving speed; therefore, a small change in the direction of the jetting mechanism will cause a large deflection in the trajectory. Notably, the substantial energy required to produce high-speed water jets, particularly when they serve as the main mode of locomotion, presents a significant challenge to the energy supply and wireless maintenance. Additionally, the turbulence caused by the high-speed water flow during jet propulsion significantly disrupts the surrounding environment, disturbs neighboring animals, and diminishes the stealth capability of the robot. Consequently, jet propulsion is deemed more appropriate for specific scenarios, such as rapid evasion or capture, where these drawbacks are less concerning. Moreover, the current movement patterns available for jet-propelled robots are somewhat restricted and typically limited to single-directional actions, such as surfacing, diving, and lateral jetting. Robots with complex multidirectional escape maneuvers require accurate control of cavity deformation and cooperation with other actuating mechanisms.
3.3. Paddling mode
Paddling is a prevalent mode of locomotion among aquatic organisms, typically generating thrust through a variety of appendages, such as flippers and limbs. For example, jellyfish and octopuses use their tentacles for propelling, whereas frogs, sea turtles, and water beetles use their legs or limbs for paddling. Compared with other locomotion modes, the paddling mode often involves several appendages, thus enhancing maneuverability and adaptability to complex underwater environments. In addition, the paddling motion is minimal, which produces less noise and impact on the surrounding environment.
Given the notable advantages of paddling propulsion, significant progress has been made in the research on underwater robots that employ this locomotion, particularly those inspired by jellyfish and octopuses. These bioinspired robots typically utilize flexible materials, such as rubber or silicone, for their limb structures to mimic the coordinated movement of multiple pliable appendages found in creatures, such as jellyfish and octopuses. For instance, studies [55],
[56] successfully used rubber to construct bioinspired tentacles for jellyfish and octopuses, incorporating SMAs to achieve both flexibility and biomimetic pulse and recovery motions with a moving speed of 0.69 and 0.5 BL·s
−1, respectively. In addition to SMAs, pneumatic systems (Fig. 13(a))
[22],
[50] and IPMCs (
Fig. 13(b))
[49] have been widely applied in paddling propulsion.
Furthermore, many aquatic organisms such as frogs, sea turtles, and water beetles utilize paddling propulsion. Similarly, bioinspired robots designed to replicate these organisms often employ flexible materials, such as polydimethylsiloxane and polyethylene terephthalate, to manufacture the outer shells of the driving structures (appendages). Internally, innovative flexible driving materials are embedded to achieve bending, twisting, and consequently, the paddling motion of the appendages. For example, in Ref. [42], polydimethylsiloxane was used to create bioinspired turtle flippers, with an internal SMA serving as the actuating mechanism to replicate the flapping motion of a sea turtle (Fig. 13(c)). These bioinspired turtle flippers achieved a moving speed of 0.14 BL·s
−1. In Ref.
[47], a photopolymer material was used to manufacture a bioinspired frog-leg driving structure (Fig. 13(d)) with an SMA embedded internally to control joint movements, thus mimicking the surface gliding of a frog at a speed of 18 mm·s
−1. In addition, some MURs inspired by water beetles
[59] and crabs
[66] have employed paddling propulsion for surface and underwater gliding.
Compared with jetting propulsion, paddling propulsion has speed limitations, which can pose challenges in meeting the requirements of capturing or escaping movements. Additionally, they face difficulties in adapting to environments characterized by rapid water flow, such as turbulent conditions. Therefore, aquatic organisms, such as octopuses, often employ a combination of motion patterns that incorporate both paddling and jetting. They use paddling for slow cruising, and resort to jetting when rapid and agile movements are required. For manufacturing underwater robots that use paddling propulsion, the development of highly flexible and freely moving appendages poses significant challenges in terms of materials, actuation, and precise control. Currently, most appendages are limited to planar bending, which restricts their ability to perform out-of-plane, multiangle, and multicurvature bending and twisting. This limits the maneuverability and flexibility of robots. In addition, the paddling mode requires the continuous movement of the tentacles or appendages, which results in high energy consumption. These moving appendages are vulnerable to entanglement and damage by aquatic organisms, limiting the application of robots in complex underwater environments.
3.4. Crawling propulsion
Crawling is one of the main modes of movement for aquatic animals such as octopuses and crabs. In addition, gastropods, such as snails, generate propagation movements through the deformation of their pleopods and utilize secreted mucus to enhance the interaction between the body and the contact surface, thereby facilitating crawling. Crawling offers distinct advantages over swimming. First, during crawling, the tentacles, legs, and pleopods make contact with or adhere to the ground, allowing the animal to move or maintain its posture through friction or adhesion. Second, crawling movements involve deformation, shortening, and stretching of muscles to move the body, enhancing controllability in both direction and posture. With the help of ground friction, crawling can also prevent the loss of balance owing to water currents or other environmental factors. A particularly noteworthy feature is that the tentacles used in crawling can be used not only for movement but also for capturing and manipulating objects. This dual functionality makes them especially well-suited for navigating confined spaces and showcasing a high degree of intelligence.
Considering the advantages of crawling movements, Cianchetti et al.
[43] developed a bionic octopus-crawling robot with eight tentacles. Of these, two tentacles are designated for manipulation, and the remaining six are dedicated to crawling. The manipulation tentacles were constructed from intertwined polyethylene threads and SMA, and driven by a combination of motors and SMA. This configuration enables arbitrary elongation, contraction, bending, and stiffness adjustments. The six crawling tentacles are composed of silicone cones with axially embedded steel cables controlled by motors. When the arms are firmly attached to the ground, the motors contract the steel cables radially, elongate the arms, and propel the body forward. The suction cups were released at the end of each pushing phase and the cycle began again. This robot achieves a crawling speed of 5 cm·s−1 (0.3 BL·s
−1) and is able to capture target objects. In a separate study, Ishida et al.
[51] developed a soft robot that can sense local flow conditions and deform a portion of its body to maintain advantageous hydrodynamic profiles (Fig. 14(a)). The robot consists of four hydraulic legs and a morphing body; pressurizing the hydraulic legs enables leg bending and facilitates underwater crawling.
For biomimetic gastropod-like robots, their research typically involves the design of the pleopod structure and the mechanism of interaction between the pleopod and the mucous environment. Xin et al.
[139] and Chan et al.
[140] designed two snail-inspired pneumatic miniature robots for gastrointestinal tract inspection. These robots mimic snail pleopod movements by controlling the oscillation of air chambers, and have validated their mobility on gel interfaces of different concentrations. In Ref.
[141], a liquid crystal elastomer is used to induce deformation, imitating the undulating motion of a snail pleopod. This approach enables a robot to move on surfaces with various characteristics, such as horizontal, vertical, smooth, and rough surfaces. Recently, a bionic snail robot (Fig. 14(b)) capable of wet adhesive locomotion was developed
[68]. This robot uses vertical fluctuations to change its friction periodically to enable crawling in longitudinal wave patterns. With a combination of poroelastic foams as internal soft constraints, the robot demonstrated retrograde wave motion on dry substrates and direct wave motion on cornstarch suspensions. Remarkably, this robot has a loading capacity 2.84 times its own weight.The crawling mode relies on contact between the robot and the bottom substrate. Hence, robots that use the crawling mode typically exhibit slow movement speeds and limited maneuverability. Meanwhile, rough substrates and obstacles, such as screes and grids, require robust and controllable robots. Combining crawling and swimming motion modes is a promising solution to the above challenges. This combined mode not only allows for maintaining stability and balance in high-flow environments, thereby facilitating exploration in unknown environments, but also enhances agility and environmental adaptability. In addition, the crawling mode works on both the underwater substrate and the ground. When a robot uses crawling mode to transmit water to the ground, the influences of buoyancy, ground roughness, and gait change should be considered.
3.5. Flagellar-inspired mode
In addition to the bioinspired macroscopic aquatic organism propulsion methods mentioned above, there is a special category of helical propulsion inspired by the movement of microorganisms such as sperm and bacteria. These microorganisms typically have a simple structure comprising a head, an intermediate connecting segment, and a tail with flagella and cilia for propulsion [142]. The head stores information, the intermediate segment provides energy, and the tail generates bending deformations through dynein motors. By selectively activating different dynein motors, flagellar/ciliary undulation or helical motion can be achieved, resulting in thrust generation and propulsion of cell movement [143]. Macroscopic aquatic organisms usually inhabit environments with medium to high Reynolds numbers, where traditional methods, such as oscillation and paddling, generate sufficient thrust. In contrast, microorganisms are small (below the millimeter scale) and move at slower speeds in low-Reynolds-number environments, where viscous forces dominate. Helical propulsion, characterized by its asymmetric motion, breaks the symmetry of deformation in low-Reynolds-number environments and is a more efficient mode of movement [144]. Therefore, this method is more suitable for MURs operating in high-viscosity environments such as muddy water.
An example of a miniature robot emulating flagellar motion is presented in Ref.
[60]. The robot featured four helical tails that generated thrust (
Fig. 14(c)), achieving a propulsion speed of 22 mm·s
−1. In addition, the imitation of flagellar motion patterns has been extensively applied to magnetically controlled robots. These robots, typically in millimeters
[145] or even smaller
[146],
[147],
[148], utilize electromagnetic coils to create a rotating magnetic field for propulsion. Coating a robot surface with specific proteins or other chemical materials enables precise movement and drug delivery, offering promise for future applications in precision medicine [149].
It should be noted that the flagellar-inspired mode is efficient in low-Reynolds-number environments such as mudflats, tidal flats, and vessels. When the robot exhibits helical propulsion in rivers, its propulsive efficiency is low and its performance is not as good as that of the robotic fish. In addition, current helical robots can only realize planar motion, and vertical motion has not yet been studied. The planar turning motion is achieved by the deflection between the propulsive force from the helices and the axis of the robot, which transfers a part of the propulsive force to the lateral force. However, the ascending motion requires a large deflection in the vertical direction against gravity and fluid forces, of which the existing helical mechanism is incapable. To solve this problem, the robot should be equipped with more actuators, such as SMAs, pneumatic actuators, or servo motors, to facilitate the transmission of the moving direction from planar to vertical. Although adding another helical actuator in the vertical direction can solve this problem, considerable vertical speed requires large helixes, which increase the size and are easily entangled by aquatic organisms.
3.6. Hybrid mode
Compared to a robot with single locomotion, a robot with hybrid locomotion is superior in terms of flexibility, environmental adaptability, and mobility. Amphibious robots, designed to adapt to both aquatic and terrestrial environments, often adopt a hybrid mode. The main research challenge for amphibious robots is the integration of amphibious motion structures, specifically in the complex design and manufacturing of amphibious motion mechanisms. To address this problem, Chen et al.
[46] developed a quadruped miniature amphibious robot capable of both terrestrial and surface swimming (
Fig. 14(d)). It is propelled by eight piezoelectric motors distributed across its four legs, each featuring two degrees of freedom for rotation, facilitating fore-and-aft and diving movements. In the terrestrial environment, the robot uses four legs for walking, whereas in the aquatic environment, the robot uses the flap on each leg to paddle in water for propulsion. Additionally, electrowetting technology is employed to eliminate the influence of surface tension at the air–liquid interface. Wang et al. [32] developed a miniature amphibious robot based on a hybrid vibration module. This robot has two distinct RFH modules equipped with soft feet and flexible fins on rigid legs to transmit vibrations to the environment using eccentric motors. In terrestrial environments, it operates with soft feet. In aquatic environments, vibrating flexible fins are used to spray water, thereby producing a superior driving effect. This robot achieves a maximum speed of 815 mm·s−1 on land and 171 mm·s
−1 in water, and can communicate via Bluetooth.
The hybrid mode involves at least two locomotion modes that provide an excellent ability to move between land and water. However, because actuation mechanisms are developed for multiple motions, the mechanism design and control strategy are complex, thereby increasing the manufacturing difficulty. In addition, each locomotion mode in the hybrid mode has a preferred working environment or condition. Therefore, installing sensors in a robot to perceive environmental information and select the appropriate motion mode is a major challenge. This involves sensor technology, multisource information fusion, optimization, and decision-making. If a robot has this perception and autonomous decision-making ability, its motion performance and intelligence can be significantly improved.
Currently, miniature amphibious robots still adopt relatively traditional methods, such as piezoelectric motors and DC motors. However, with ongoing advancements in flexible materials, it is believed that in the future, multi-degree-of-freedom amphibious motion mechanisms with higher flexibility, more degrees of freedom, and more flexible modes of movement will undergo substantial development. This evolution is intended to enhance the overall practicality and utility of miniature amphibious robots.
Currently, most MUR locomotion modes only imitate a single mode of aquatic organisms, such as octopus jetting and crawling. However, there is still a lack of agility and maneuverability in terms of motion. At the same time, when robots operate in underwater environments, there is a high demand for their resistance to motion interference, particularly in the face of complex factors such as ocean currents. Currently, few scholars have studied stable motion methods for robots in complex environments. Additionally, as an important means of exploring the ocean, the functionality of MURs needs to be strengthened. In the future, scholars can achieve functions similar to those of aquatic organisms, such as tracking and capturing, through the development of driving-operation-related methods.
4. Conclusion and perspective
In recent decades, many researchers have devoted significant efforts to studying MURs. This has greatly promoted their development and led to many landmark achievements such as centimeter-level wireless control, movement speeds of up to hundreds of millimeters per second, underwater 3D motion capabilities, robot swarms, and underwater operation robots. This article presents a comprehensive overview of the latest advancements in the actuation and locomotion modes of existing MURs and provides a perspective for researchers aiming to gain a thorough understanding of MURs’ research. In the future, the development of MURs will provide innovative solutions for promoting the investigation of the natural marine environment, facilitating the sustainable utilization of marine resources, addressing marine pollution concerns, and safeguarding marine biodiversity.
Although significant progress has been made in research on the actuation methods and locomotion modes of MURs, many challenges and prospects remain to be addressed (
Fig. 15). Some of the key issues and perspectives are as follows:
Actuating mechanisms: Current actuating mechanisms mostly imitate single driving units found in aquatic organisms, offering limited degrees of freedom and posing challenges in achieving complex integrated motions, such as bending, twisting, and stretching. Once the actuating mechanism is designed and manufactured, its locomotion mode is fixed, thereby limiting its movement flexibility.
To address these challenges, a programmable and reconfigurable design approach for the actuation mechanism or material should be explored. This would allow the material to be reprogrammed or the actuating mechanism to be reconfigured based on specific requirements under varying conditions. For example, origami technology can be integrated into actuating mechanisms, thereby significantly enhancing movement flexibility and environmental adaptability.
The concept of modular robots is a promising solution. Each module can be customized to respond differently to various drive signals, and the modules can be assembled using different strategies. Consequently, the robot shape becomes adaptable, its movement becomes flexible, and it can execute a range of locomotion modes, thereby enhancing its agility and robustness.
In addition, enhancing the biological resemblance of underwater biomimetic robots, increasing their degrees of freedom, and aligning their shape and motion with those of underwater creatures are efficient methods for improving the performance of the actuating mechanism. Moreover, optimizing the actuating structure, such as by reducing resistance, streamlining the robot, and minimizing its weight and volume, can collectively improve the robot’s movement efficiency. Furthermore, enhancing the energy density and energy transmission efficiency of soft materials used in propulsion systems is another key aspect in enhancing actuating performance.
Locomotion and control: Currently, most locomotion modes employed by MURs merely replicate the single-motion patterns observed in aquatic organisms, lacking the agility and maneuverability required for effective navigation. Meanwhile, few scholars have studied the stable motion of robots in complex environments such as ocean currents. In addition, stress-induced behaviors observed in aquatic organisms, such as escape and avoidance mechanisms, are yet to be thoroughly studied and integrated into MURs. Furthermore, the development of autonomous operational modes for robots is critical for practical applications.
Improvements in sensing capabilities are important for enhancing autonomy and intelligence. This involves integrating high-precision sensors, such as global positioning systems (GPS), inertial equipment, ultrasound, active vision, and artificial sidelines, to enable multimodal data fusion for a comprehensive understanding of the environment. By monitoring changes in the external flow field in real time, an underwater robot can use vortices to improve energy efficiency and thus improve motion performance. This sensing ability, combined with advanced control algorithms, such as deep reinforcement learning, also allows accurate movement of mechanisms and autonomous decision-making for switching locomotion modes or adopting hybrid motion modes according to environmental conditions. For example, tracking and capturing objects rely heavily on an intelligent cooperation strategy with multiple degrees of freedom. The robot should first adopt a fast locomotion mode, such as jetting, to approach the target and then use paddling, swimming, or other slight modes to track it. During tracking, several appendages cooperate to capture or sample the objects.
Cross-domain adaptability: Currently, the application scenarios for underwater robots are mostly limited to surface or underwater environments, with limited research on cross-domain environments. However, cross-domain environments such as mudflats, sludge, ice, and snow are inevitable scenarios for exploring marine environments and practicing resource protection and utilization. Moreover, significant viscosity changes and the presence of suspended particles in cross-medium environments pose great challenges to efficient propulsion mechanisms, strong adaptability, and stable control of underwater robots. The dynamic mechanical analysis of robots in these environments and the investigation of contact on cross-domain surfaces will aid in designing actuation mechanisms and developing control algorithms for cross-domain transmission.
Expanding practical applications: Currently, research on MURs mainly focuses on achieving underwater motion and is mostly conducted in laboratory environments. Comparatively limited research has been conducted on practical, real-world applications. The ocean environment is far more intricate and challenging than a controlled laboratory setting, making the enhancement of robot maneuverability and autonomous operation in practical working conditions a significant challenge. Furthermore, the collaborative operation of multiple robots will play a pivotal role in future marine explorations. The development of swarm robotics algorithms and coordinated operational control methods is an emerging and crucial trend in this field.
CRediT authorship contribution statement
Panbing Wang: Writing – review & editing, Writing – original draft, Funding acquisition, Data curation, Conceptualization. Xinyu Liu: Writing – review & editing, Writing – original draft, Data curation. Aiguo Song: Writing – review & editing, Writing – original draft, Validation, Supervision.
Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgments
This work was supported by the Natural Science Foundation of Jiangsu Province, China (BK20220813), and the Fundamental Research Funds for the Central Universities (2242023K40014).
PDF
(19376KB)
