key: cord-0441350-jma553zi
authors: Hou, Ningzhe
title: The Design and Simulation of Biomimetic Fish Robot for Aquatic Creature Study
date: 2021-10-13
journal: nan
DOI: nan
sha: 3940cf2b74ab60650a8db05819d131552c2b4dd9
doc_id: 441350
cord_uid: jma553zi

In the application of underwater creature study, comparing with propeller-powered ROVs and servo motor actuated robotic fish, novel biomimetic fish robot designs with soft actuation structure could interact with aquatic creatures closely and record authentic habitats and behaviours. This final project report presents the detailed design process of a hydraulic soft actuator powered robotic fish for aquatic creature study capable of swimming along the 3D trajectory. The robotic fish is designed based on the analysis of the pro and cons of existing designs. Except for the mechanical and electronic designs and manufacturing method of crucial components, a simplified open-loop control algorithm was designed to check the functionality of the application board and microcontroller in the Proteus simulation environment. As the key component of the robotic fish, Finite Element Method (FEM) simulations were conducted to visualise the soft actuator's deformation under different pressure to validate the design. Computational Fluid Dynamics (CFD) simulations were also conducted to improve the hydrodynamic efficiency of the shape of robotic fish. Although physical manufacturing is impossible due to the pandemic, the simulations show overall good performance in terms of control, actuation, and hydrodynamic efficiency.

.1. Table of simplified motion control -----------------------------------------------------22   Table 3 

In the application of underwater creature study, comparing with propeller-powered ROVs and servo motor actuated robotic fish, novel biomimetic fish robot designs with soft actuation structure could interact with aquatic creatures closely and record authentic habitats and behaviours.

This final project report presents the detailed design process of a hydraulic soft actuator powered robotic fish for aquatic creature study capable of swimming along the 3D trajectory. The robotic fish is designed based on the analysis of the pro and cons of existing designs.

Except for the mechanical and electronic designs and manufacturing method of crucial Computational Fluid Dynamics (CFD) simulations were also conducted to improve the hydrodynamic efficiency of the shape of robotic fish.

Although physical manufacturing is impossible due to the pandemic, the simulations show overall good performance in terms of control, actuation, and hydrodynamic efficiency.

Although it has been a special and challenging year due to the outbreak of the pandemic, I have received many supports throughout the final project period. 

Remotely Operated Underwater Vehicle (ROV) is widely used for underwater life study nowadays due to their high capacity to carry sampling and operating equipment and high mobility. However, the main functionality, even for large work-class ROVs, is taking pictures of the worksite. [1] In the case of underwater creature study, the ROVs with loud noise generated by propeller and large size would disturb the objects of research and make them instinctively escape from the ROVs, which is non-ideal for the study on the natural habitats of the underwater creatures as well as the protection of wildlife. A biomimetic robotic fish with a camera could be a perfect substitution for the ROVs in applying underwater creature study, as they could be integrated into the natural environment without disturbing the objects of study and observing authentic habitats and behaviours.

With centuries of evolution, aquatic creatures have developed efficient swimming mechanisms.

The swimming efficiency of fish could reach up to 90 % efficiency, while most conventional human-made propellers could only reach 40 to 50 % efficiency. [2] The researchers created various low-noise and high-efficient Autonomous Underwater Vehicles by studying the fishes' swimming mechanism.

In the design of biomimetic fish robots, most of the designs still use servo motors and multi-link structure as the tail fin and lateral fin actuator. The robot fish with such design could not wholly imitate fishes' behaviour due to the limitation on the degrees of freedom, and the control algorithm is also complicated. The control of traditional robotic fish actuators with oscillatory hinge joints normally needs to establish a general kinematical model and develop control methods based on the model. [2] Moreover, the complexity will significantly go up when more hinge joints are designed. The overall system power consumption will significantly increase when more servo motors are used to increase the degree of freedom of the tail fin actuator.

Soft robotics also has excellent potential in the application of biomimetic designs since the actuation structure and compliant materials are widely used in soft robotics design, and they share similarities with the organic structures on the creatures. The research and application of soft robotics have just come to light in the recent decade. Various soft robotics designs allow robots to integrate into the natural environment and safely facilitate our daily lives.

With all the potentials and advantages elaborated in the previous section, the research and development of the robotic fish with soft actuation structure have just started in the recent decade and is still in the primary stage. Most of the existing laboratory-level soft actuator-based robot (SARs) fish designs could only operate at a certain depth instead of moving along the 3D trajectory. Meanwhile, there are currently no commercial level products.

For the applications of scientific aquatic creature habitats and behaviours observation, an agile robot capable of swimming along a 3D trajectory and closely interact with aquatic creatures with long operating time is necessary. A biomimetic soft robotic fish is capable of achieving such functionality better than servo motor actuated robotic fish and propeller-powered ROVs.

The motivation of the project is to investigate the possible design and performance of softactuator-powered robotic fish for the next-generation underwater autonomous robot for aquatic creature study. The significance of the research does not solely come from its potential to provide a new form of interaction with underwater creatures, it could inspire future robotic designs that combine biological structure and robotics to achieve extraordinary functionality. It could also be helpful for the research of hydrodynamics of fish with artificial lateral lines for underwater navigation research and swarm robotics behaviour research.

[3]

This project aims to build a robotic fish capable of swimming along the 3D trajectory and fully replacing the servo motor actuators on the traditional biomimetic fish robot with soft actuation structures. Compared with existing similar designs, the project aims to resolve low turning control precision and limitation of operating time due to gas cartridge volume. Hopefully, the research can increase the dexterity of the robotic fish while eliminating the restrictions on depth control and operation time.

Since the project is highly practical and experimental, in-lab manufacturing, experiments, and testing are crucial for delivering the satisfactory outcome of the project. As the UK was in lockdown throughout the period of the project, although the biggest efforts have been made to test the performance of the robotic fish physically, the manufacturing of robotic fish is impossible before the presentation of the project due to the restriction of access to the school facilities and government policy. After discussion with the project supervisor, it was agreed that the physical manufacturing and testing were replaced by virtual simulations on the subsystems, and both the final report and presentation will base on the design and simulation result.

The physical soft robotic fish for real-world testing has been replaced by CAD modelling and necessary hyper-elastic Finite Element Method simulation of the performance of the soft actuator in COMSOL; the testing on the program and performance of the controlling circuit board was replaced by the Proteus simulation. The simulations should provide solid verification of the design for future physical manufacturing.

Literature Review

Mustafa Ay et al. [4] developed a robotic fish that actuated by a two-link tail mechanism. Each link is driven by a servo motor. The elevator up-down motion is achieved by using a servo motor to slide the battery along the medial axis with a lead screw mechanism. In the physical testing, the robotic fish can reach a circular turning motion of 0.2792 m. For the two pitch angles choose to test the elevator up-down motion, the robotic fish could achieve effective depth control with the lead screw Centre of Gravity (CoG) adjusting mechanism. Furthermore, the author mentioned that all parts need to be covered with epoxy resin to prevent potential water leakage. The difficulty of waterproofing goes up for the multi-joint actuation structure.

Andrew D. Marchese et al. [5] have developed an Autonomous soft robotic fish capable of escape manoeuvres using fluidic elastomer actuators, a CO2 cartridge, and flow control valves are used to enable escape manoeuvres by releasing a large volume of CO2 gas ranging from a baseline flow of 5 L/m to a maximum flow of 50 L/m into the pneumatic chamber. However, there are several drawbacks to Andrew's design. Firstly, the swimming time is restricted by the gas volume of the CO2 cartridge since the pressurised gas is released into the water after inflating the pneumatic actuator. Secondly, the design of CO2 cartridge and flow control valves could only pressurise one side of the pneumatic actuator instead of pressurising one side while vacuuming another, which may decrease the efficiency of the swimming. Furthermore, most importantly, the design does not provide control on pitch angle for 3D trajectory swimming.

Robert K. Katzschmann et al. [6] have developed a robotic fish driven by a soft hydraulic actuator, the Buoyancy Control Unit, and Dive Planes to enable the fish to dive at the depths of 0-18 m. The fish can achieve an average swimming speed of 21.7 cm/s. A custom-designed unidirectional acoustic communication modem accomplishes the real-time control of the robot, and the distance of successful signal transmission is around 21 m from the robotics fish. Hence, a human diver needs to follow the fish. There is also a fisheye camera placed at the front to record real-time video from the view of fish. The tuning control was achieved by the dive planes at a limited depth.

Manual adjustment on the weight by the diver is needed when the fish goes out of the allowed range in order to return to the original depth.

Tiefeng Li et al. [7] developed a fast-moving biomimetic menta ray robotic fish with dielectric elastomers soft actuators. The design took advantage of the surrounding open water as electric ground and boost the battery voltage to 10kV to drive the dielectric elastomers soft actuator. The fish could achieve 0.024 W input power and 10.25% power efficiency, which is comparable to the actual fish with similar size.

Due to the limitation on the maximum page count, the table of comparison of the designs is attached in Appendix 6.1 to provide a more structured view of the pro and cons of existing designs.

A variety of pneumatic actuation structures, manufacturing and control methods were developed to fit into different needs for actuation. The FDM 3D printing manufacturing method for dualchannel bellows-type actuators with a bottom-up approach was developed by Hong Kai Yap et al. [8] ; The actuator can be directly printed with NinjaFlex filament with Shore hardness of 85A, comparing with other designs, the actuator can be wholly printed with FDM 3D printer, which is more accessible than expensive SLA 3D printers. The manufacturing process is simplified comparing with mould casting as well. Although the actuator can withstand a maximum of 400 kPa pressure before failure, the actuation pressure needed is higher than silicon actuators with similar size since the silicone rubbers generally have the Shore hardness below 45A, compared to 85A filament. The shape bellows-type actuators are non-ideal to be placed on robotic fish in terms of hydrodynamic efficiency since the bumpy faces will generate higher drag than streamline body.

Raphael Deimel et al. [9] developed a silicone-rubber-made soft pneumatic actuator called "PneuFlex" that can be manufactured with a two-part mould and reinforcement helix; the manufacturing process is significantly simplified by the novel actuating structure. Except for the FDM 3D printing, Direct in writing (DIW), Selective sintering (SLS), Inkjet printing, and stereolithography technology could be used to manufacture various soft robotic systems materials with different properties.

In terms of the sensing and control of robotic fish actuation structures, Yu-Hsiang Lin et al. [10] embedded a soft sensing mechanism with liquid Gallium-Indium alloy metal in flexible microchannels (eGain) in the soft fishtailing actuators, the morphing of the microchannels will cause the resistance change in the liquid metal, and the strain can be calculated based on the resistance change. Such a flexible sensing mechanism enables closed-loop control robotic swimming based on the sensor measurements.

Comparing with other open-loop controlled soft actuators, the precision of the swimming trajectory can be significantly improved.

The following parts are included in the rest of the report:

In the second section, detailed mechanical designs, and the manufacturing method of some of the key mechanical components will be first introduced. The components include the soft actuators, gear pump, waterproof shell, pectoral fins, and balance control unit. The electronic components design, including water sensor, microcontroller, application board, IMU, and battery pack, are introduced after the mechanical designs.

The implementation and results of three subsystem simulations are included in the third section.

The simulation on the application board functionality and simplified control algorithm is elaborated in the first part. And then, the method to perform hyperelastic FEM simulation on the soft actuator performance and simulation results are introduced. At last, the CFD simulation on the hydrodynamic efficiency of the robotic fish design is covered.

Finally, the summary of execution is addressed in the fourth section of the report. The highlights and flows of the design and simulations are analysed in this part. Based on the analysis, future work that can be done to further improve the design and the plans to physically manufacture the robotic fish is further discussed at the end of the report.

The assembly of the robotic fish is shown in Figure 2 .1 below. The soft robotic fish includes ten main components: Soft actuator, gear pump, waterproof shell, balance control unit, artificial pectoral fins, inertial measurement unit (IMU), battery pack, application board, water sensor, and microcontroller. The design and manufacturing of each component will be elaborated and discussed in this section. 

The soft actuator is the most crucial component in the soft robotic fish design. A deep dive into the available designs was conducted. There are three main design methods: hydraulic, pneumatic, and dielectric elastomer actuated.

The soft pneumatic actuator is ideal for fast turning, but it is extremely difficult to swim along a 3D trajectory since the buoyancy distribution will be affected when the pressurised gas is released to the actuator; the gas volume in the cartridge will limit the operation time, and the replacement of the cartridge is problematic. Dielectric elastomer actuator is a novel design that generates extremely low noise during the operation, although a boost converter is needed to boost the battery voltage to more than 10 kV in and the bending of the actuator is not symmetrical. Thus, it is not suitable for achieving fishtailing behaviour.

The hydraulic actuator was chosen to be the tailing actuator of the fish. Since the liquid could be cycled between the left and right actuator and will not affect the buoyancy distribution, the limitation of the gas cartridge could be eliminated, and the gear pump is relatively simple to operate. The soft actuator functions by pressurising and depressurising the soft actuators with liquid by a gear pump, the fluidic channels within the actuator will be expanded and compressed, and the constrain layer will be bent as it is inextensible in length. By cycling the fluid between left and right actuators, they are pressurised and depressurised separately; the constrain layer will be bent in both directions. Through such a mechanism, the fishtailing behaviour can be imitated. [6] Ecoflex 30 Silicon rubber will be used to fabricate the soft fluid actuator by mould casting. The mould design is shown in Figure 2 .3; the design consists of 3 separate parts: top mould, bottom, and plug for generating the water in/outlet hole. All three parts will be manufacturing by a Stereolithography 3D printer so that the top and bottom can be precisely assembled to each other. Because removing the support from the 3D printed part will significantly affect the casting part's precision and smoothness, the parts' design has avoided the occurrence of undercut geometry so there will not be any support generated during the 3D printing. Liquid silicon will be injected from the top of the mould, and after which the mould will be placed in a vacuum container to remove the bubbles; if the liquid silicon level goes down, repeat the previous step several times until the level no longer drops. After curing in the oven, the silicon-made soft actuator can be fabricated. 2 Spectra Symbol Flex Sensors [11] are placed in the middle of the constraining layer to provide control feedback by measuring the bending angle of the actuator. The sensors will produce 10k

Ohm resistance when the constraining layer is flat, and the resistance will increase to up to 110k

Ohms as the bending angle increases. Each sensor could only provide reliable measurement in 1 direction, so two sensors will be placed reversely to measure the bending angle in both directions. The sensor manufacturer does not provide an accurate resistance-curvature relationship in the datasheet. The sensors will have to be tested physically to validate the accuracy and range of measurement.

The decision was made to use a gear pump to cycle the liquid between left and right the soft actuator. The crucial reason for choosing a gear pump is output is directly proportional to the motor speed, which means the gear pump could deliver smooth, pulse-free, and invertible flow comparing with piston pumps and peristaltic pumps. The rate of flow and pressure generated by gear pumps is generally higher comparing with centrifugal pumps and diaphragm pumps with similar power ratings. The compact volume could also fit into the body of the fish.

After conducting research on available gear pumps online, DHE 385 micro gear pump [12] is found to be the optimal option to drive the fishtailing actuator; the gear pump uses an SRC-385MP-2165-54DV motor with 12 V rated voltage and 0.8 A maximum rated load current, it can produce 2L/minute liquid flow and achieve maximum 2.45 bar pressure. However, the water inlet and outlet of the pump are placed on the side of the gear pump (shown in Figure 2 .5), which is nonidea for the connection with fishtailing actuator. 

The shell protects all the electronic components from contact with water during the operation. It consists of 5 main parts: pelvic fins, dorsal fin, rigid supporting frame, shell, and electrical ports.

The main functions of the pelvic fins and dorsal fin are to stabilise the fish during swimming and prevent the fish from rolling during the turning. The electronic components are assembled on the 3D printed rigid frame as shown in Figure 2 .7, except for the water sensor and electrical ports are assembled on the semi-soft shell, which is printed with 90A Shore hardness resin. The shell is supported by the rigid supporting frame. During the assembly, after all the electronic components are assembled and constrained, the supporting structure and shell can be easily assembled to each other, after which applying waterproof glue, the shell is properly sealed. The transparent window at the front of the shell enables the fisheye camera to be placed on the fish. The balance control unit consists of a metal driven by a lead screw and a two-phase hybrid micro stepper motor since stepper motors could provide simple and precise position tracking, the step angle of the stepper motor can be calculated as:

The relationship between the rotational speed of the stepper motor and the linear speed of the metal can be driven as:

where 0.002 is the lead of the screw and N is the rotational speed of the screw, and v is the linear speed of the metal.

It controls the pitch angle by sliding the metal along the lead screw, resulting in horizontal adjustment on the centre-of-gravity of the robotic fish.

FR-IR12 reflective water sensing module [14] is used to detect if the robotic fish is underwater or has reached the water's surface. It is placed on the top of the robotic fish nearby the dorsal fin. Comparing with traditional mechanical water sensors, FR-IR12 detects the water through light reflection. The sensor has a prism-shaped probe to reflect the light from an internal luminous diode to the receiver. If the probe is placed in the air, the prism will reflect most of the light to the receiver and outputs a logical high. When the probe is placed in the water, most of the light will scatter in the water; in this case, the receiver will pick up significantly fewer light signals, and the sensor will output logical low.

The compact and reliable sensor design is an excellent choice to be placed on the robotic fish not only because of the space problem, the waterproofing of the robotic fish will also be significantly simplified comparing with traditional mechanical water sensors.

The available space inside the robotic fish is extremely limited. Ensuring that the controller and motor drive board are as compact as possible while the IOs are sufficient for the system is crucial in the design.

Arduino Micro is chosen to be the controller of the robotic fish. It is designed based on Atmel The application board mainly consists of 4 parts: 2 H-bridge modules with optical isolation, 12V to 5V step-down voltage converter, current monitor circuit, and connection ports and headers. The wiring diagram of the application board is also shown in the Appendix 6.2.

Polygon Pour connection is used to reduce the mutual capacitance and inductance to improve the signal integrity. Headers for the motors and sensors are placed at the bottom of the board. Most of the heat will be generated at L298P dual H-bridge modules [16] and the L7805 power regulator [17] . They are placed at the top so that the wiring will not affect heatsink assembly and performance. Since the thermal conductivity of water is higher than air, thermal management should not be problematic. The input voltage from the battery is 12V, to supply 5V to the Arduino Microcontroller and servo motors, an L7805 positive voltage regulator IC is used to provide a constant 5V power supply to

Arduino and servo motors. The IC can deliver a maximum of 1.5A current with a proper heatsink [17] , which is more than the total current rating of the Arduino and servo motors.

An ASC712 Hall Effect-based linear current sensor [19] is used to monitor the real-time current outputs from the battery. The relationship between sensed current and output analogue voltage is shown in Figure 2 The build-in 10-bit Analogue-Digital Converter will divide the 0-5V analogue input voltage into 1024 steps. Thus the numerical relationship at 25 degree Celsius between the digital input and sensing current can be further derived as:

= 40.92 × + 511.

The schematic diagram of the voltage regulator and current sensing circuit is shown in Figure 2 .14. 

Since the robotic fish is designed for swimming along a 3D trajectory, an Inertial Measurement Unit (IMU) is necessary for tracking the real-time position of the robotic fish. MPU 6050 [20] integrates a 3-axis accelerometer and a 3-axis gyroscope on a single IC, and it is also called 6 degrees of freedom motion tracking device. MPU 6050 also has a temperature sensor and a Digital where vector a is the numerical acceleration in the direction, s is the chosen scale range, d is the axis acceleration data acquired from MPU 6050 module, and 32768 is the maximum possible reading from the 16-bit ADC module onboard. Since the arccosine function can only return a positive number, the roll angle needs to be negative if y is positive, and the pitch angle needs to be negative if x is negative.

It is difficult to calculate absolute yaw angle without the reference of earth magnetic field measurement, but the change in yaw angle, which is the angular velocity can be easily acquired, and the measurement of angular velocity is sufficient for the application. A complementary filter can be implemented to combine the roll and pitch angle calculated by the accelerometer measurements and direct measurement from the gyroscope. Gyroscope could normally provide accrete measurement since it is not affected by linear motion, and thus it normally has a higher weight when merged. Still, the angle calculated by the accelerometer can offset the error caused by drifting in the gyroscope from a long-term perspective.

Although there is a DMP on MPU 6050 module, the output signal is still highly likely to be noisy.

Kalman filter is ideal for eliminating the noise in the readings. It can be implemented directly through the Kalman filter library in Arduino IDE.

The [21] . In order to maximise the operating time of the fish, Panasonic NCR18650B 3400 mAh battery cell will be used to manufacture the battery pack.

Since the space within the fish body is very limited, three cells are placed in a triangle shape to minimise occupied volume, PVC heat shrink packaging could prevent the battery from short-circuit caused by water leakage, the 3D rendering image and example image provided by the battery pack manufacturer is shown in Figure 2 .17 below. Panasonic NCR18650B battery has a built-in protection circuit to prevent the battery from overcharge, over-discharge, and short circuit. Such a feature makes the battery safer to be used in the robotic fish since a short circuit caused by water leakage is highly likely to happen during the testing. The circuit can effectively protect the controller board and other electronic components.

Because of the outbreak of the coronavirus, at the current stage, physical testing and manufacturing of the subsystem are challenging to achieve; thus, two methods will be used to conduct testing on the subsystem development: Hyper-elastic FEM simulation on the soft actuator based on COMSOL Multiphysics and circuit control simulation based on Proteus Design Suite.

Proteus Design Suite is chosen to simulate the circuit and program's performance due to its compatibility with the Arduino microcontroller.

The circuit is simplified to the version in Figure 3 .1, and it includes 2 L298P ICs, an ATmega32U4 microcontroller, a stepper motor, a DC motor, two servo motors, and five switches as controller

input. An additional oscilloscope and probes are also added to the system to observe and verify the control and output signal. Since the DC motor is used to drive the gear pump to cycle water between left and right soft actuator, and the angular speed is proportional to the input voltage for the DC motors, the motion of the DC motor can be expressed by a function of input voltage and time, for the Straight Motion in Table 3 .1, the expression can be derived as:

where the is the desired maximum voltage input to motor and is the fishtailing frequency in rad/s. Assuming is the maximum voltage 12 V and the frequency is 1 Hz, Figure 3 .2 shows how the straight swimming can be achieved by ℎ ( ) function with the aid of FEM simulation results, which will be discussed in section 3.4. where the 1 and 2 is the desired maximum voltage inputs in positive and negative directions to motor and is the fishtailing frequency in rad/s. According to the design, a positive voltage will result in depressurisation in the left actuator and pressurisation in the right actuator and vice versa, and thus for , 1 should always larger than 2 in order to turn left. From to the same principle for ℎ , 1 should always smaller than 2 in order to turn right. Assuming for ( ), 1 is 12 V and 2 is 4 V, and for ℎ ( ), 1 is 4 V and 2 is 12 V. The frequency for both of the function is 1 Hz. The graph shows how the left and right turn can be achieved. The five pushbuttons in the simulation interface can be clicked to control the operating mode to switch between 5 modes in Table 3 .1.

Although the half-stepping control method could increase the accuracy of stepper motor control, the full-stepping method is used to implement stepper motor control since for lead screw structure, one revolution will only result in 2 mm linear motion, half-stepping seems redundant in such application.

After compiling the code, importing the hex program file to the microcontroller, and run the simulation, the motor rotation status will be displayed in the simulation. The oscilloscope will also display the input and output signal to further validate the circuit design.

When the robotic is swimming straight, which means ℎ has been applied to the DC motor on the gear pump. The input signal that the microcontroller applied to the L298N module is shown in the Figure 3 .4 below. The signal that coloured green is the PWM signal measured by the probe in enable pin on the L298P module, and the signal coloured in blue and red are the signals which control the output voltage polarity. It can be observed that the duty ratio of the input PWM signal changes according to the sinusoidal relationship, and the polarity changes every half cycle. The waveforms meet the design principle. The output voltage applied to the DC motor is shown in the Figure 3 .5 below. 

The prototyping of soft fluid actuators is time-consuming and complicated. As described in section 2.2.1, the mould needs to be manufactured by a stereolithography 3D printer to achieve high precision and avoid rough surfaces. After the printing, the mould needs to be washed with alcohol to remove the liquid resin and cured under UV light to increase the strength. The liquid silicon injection and removing bubbles will also take a significant amount of time. Generally, the manufacturing of a new version of soft fluid actuators from design will take at least a week, and the morphing of soft fluid actuators is difficult to directly predicted without testing, and thus simulation is effective to increase the efficiency of design and prototyping.

As a widely used method to analyse mathematical physics models, the Finite Element Method is useful to analyse the soft actuator's morphing under different pressure to validate and improve the design.

Simple linear shear-strain relationships can be used for the analysis of solid mechanics materials that withstand small strains. However, the analysis of soft robotics structures typically involves the highly flexible elastomers, in this case, silicone rubber. The deformation of these materials is usually larger than normal solid materials and is non-linear. Hyper-elastic solid mechanics models are needed to analyse the morphing of the soft actuator under different pressures. [22] Two Parameters Mooney-Rivlin Model is chosen to simulate the soft actuator since it is suitable for moderate deformation. The model equation is given by:

where C is the right Cauchy-Green deformation tensor, and W is strain energy density. The functions of the pressure applied to the left and right actuator with time are plotted in MATLAB and shown in Figure 3 .16 below. can be seen from the equations that the frequency is much lower than the actual fishtailing frequency. The reason is that if the frequency increases, the simulation calculation could not converge since higher frequency oscillation increases the instability in the system. It is possible to achieve a higher frequency in reality.

The time-dependent simulation results are shown in 

Although biomimetic streamlined shape design could significantly increase the swimming efficiency of the robotic fish in the fluid, the design of the robotic fish still faces the constrain of the volume of electronic components, and the electrical sensors and ports need to be placed on the shell, like the pyramid-shaped water sensor and electronic ports to charge the battery and upload the program. Since the object underwater could face 800 times larger drag comparing with the same object in the air [24] , assuming the same velocity, an analysis of the hydrodynamic shape of the robotic fish is important to improve the swimming efficiency.

The pectoral fins connecting with the servo motor is designed to respond to the change in the turbulence and provide dynamic lift force as well as collaborate with the fishtailing actuator to turn the fish. The hydrodynamic simulation could validate the function and performance of the pectoral fins.

Computational Fluid Dynamics (CFD) is a useful tool to analyse the hydrodynamic shape and structure of the design. Solidworks Flow Simulation add-in allows the user with little basis in CFD simulation to easily set up simulation environments by following the steps in the Wizard. The simulation parameters are shown in Table 3 .3 below. After the simulation is finished, the 3D flow trajectories can be visualised in the simulation result. 

Three models were simulated in the Solidworks Flow Simulation. Each has different pectoral fins angles (0 degrees, 30 degrees, and -30 degrees). The 3D flow trajectories of the simulation results are shown in Figure 3 .18, 3.19, and 3.20 below. The results also show that the change in the angle between pectoral fins and turbulence could change the direction of flow trajectories to control the dive depth and change direction. It is also worth paying attention that the simulation shows the maximum pressure generated near the electronic ports cover, which means the design of the electronic ports cover is hydrodynamically inefficient. Thus, further improvement can be made to the design of electronic ports to increase swimming efficiency. Although the simulation successfully proves the design concept, the results show that applying different numerical models could result in quite different morphing under the same condition.

Thus, physical testing to determine the best fit numerical model is necessary. The hydrodynamic efficiency is checked by conducting CFD simulation, and the overall result shows excellent hydrodynamic efficiency, a potential improvement that can be made is also discovered through CFD simulation.

Due to the restriction of the project time and lab access, there are a couple of simulations and physical testing that can be done in the future to improve the design.

The manufacturer of Spectra Symbol Flex Sensor, the flexible curvature sensors placed in between the left and right actuators, does not provide a detailed and reliable resistance-curvature relationship, and according to the feedback from the people that purchased the sensor, the output signal of the sensor is noisy. Physical testing needs to be conducted to determine the resistance- The current design does not consider the remote wired or wireless control of the robotic fish. The radio signal could travel a long distance through the air but attenuate rapidly in water.

Communication underwater is normally done through the acoustic signal. An acoustic communication modem can be added to the robotic fish to enable remote control of the robotic fish. However, the distance for valid acoustic communication is very short. Another solution is adding a camera at the reserved space at the front of the robotic fish, autonomous swimming can be achieved through image recognition algorithms, but more sensors like pressure sensor to measure the depth of the fish are needed to achieve autonomous swimming. 

As required by the department, this statement is attached in the Appendix to describe the impact of COVID-19 on the project and actions taken to minimise the project's negative impact.

Since this is a bespoke project designed in August, back then, the daily Covid-19 cases had been under control, and the lockdown restriction had been lifted across the UK. In the original planning, two parts were included in the project: virtual mechanical and electronic design and physical manufacturing.

However, the daily COVID-19 cases exponentially went up, and the whole country was in lockdown again since early October 2020.

Initially, with the help of the project supervisor, Dr Alexandru Stancu, several attempts were made to realise the physical manufacturing. He helped me with prototyping some of the 3D printed parts, and I reached out to a Swiss silicon rubber 3D printing start-up company for the manufacturing of soft actuator. But since the later school announcement that the campus facilities were going to remain closed, which means even the assembly of the components at the lab was impossible, the decision was made to replace all the physical manufacturing and testing with software simulation in early January.

Luckily, the results of the three simulations were satisfactory and sufficient to validate the performance of the robotic fish. In addition, Peking University Intelligent Biomimetic Design Lab are willing to help with the physical manufacturing of the robotic fish during the summer holiday of 2021.

The ROV Manual: A User Guide for Observation Class Remotely Operated Vehicles

Parameter Optimization of Simplified Propulsive Model for Biomimetic Robot Fish

Three-dimensional multi-source localisation of underwater objects using convolutional neural networks for artificial lateral lines

Mechatronic Design and Manufacturing of the Intelligent Robotic Fish for Bio-Inspired Swimming Modes

Autonomous Soft Robotic Fish Capable of Escape Maneuvers Using Fluidic Elastomer Actuators

Exploration of underwater life with an acoustically-controlled soft robotic fish

Fast-moving soft electronic fish

High-Force Soft Printable Pneumatics for Soft Robotic Applications

A compliant hand based on a novel pneumatic actuator

Modeling and Control of a Soft Robotic Fish with Integrated Soft Sensing

FLEX SENSOR" FS datasheet

Micro gear pump" 385 datasheet, n.d

Digital mini servo" KM1850MD datasheet, n.d

Single-point photoelectric level sensor" FS-IR12 datasheet

DUAL FULL-BRIDGE DRIVER" L298P datasheet

Positive voltage regulator ICs" L7805 datasheet

OCTAL BUFFERS/DRIVERS WITH 3-STATE OUTPUTS" SN74AHC244DW datasheet

Fully Integrated, Hall Effect-Based Linear Current Sensor with 2.1 kVRMS Voltage Isolation and a Low-Resistance Current Conductor" ACS712ELCTR-05B-T datasheet, n.d

MPU-6000 and MPU-6050 Product Specification Revision 3.4" MPU-6050 datasheet

EVERYTHING YOU NEED TO KNOW ABOUT THE 18650 BATTERY

Finite Element Modeling of Soft Fluidic Actuators: Overview and Recent Developments

Mechanical responses of Ecoflex silicone rubber: Compressible and incompressible behaviors

Shape of Aquatic Animals and Their Swimming Efficiency

Full Schematic Diagram of Application Board