In Japan, the number of casualties due to an industrial accident in 2009 was 114,154 people. Currently, a comprehensive system to prevent such an accident is basically done manually. Furthermore, it has not been automated. Therefore, in this study, we develop an anomaly detection method for the prevention of industrial accident using machine learning technology. In order to carry out anomaly detection reliably, we use a skeleton map of a person as training data from skeleton information extracted by OpenPose. Here, Variation Autoencoder (VAE) is applied as a deep learning model. We confirm that detection results with high accuracy are produced compared with the conventional method.

In human-robot collaborative operation applications, maintaining protective separation distance during robot operation is a safety requirement based on the associated safety standards ISO 10218 and ISO/TS 15066. In the study, we implement our proposal of potential runaway space in task space which was originally proposed by our research group on a vertical articulated robot and evaluate the distance from the robot end effector to the envelope of an additional potential runaway space. The purpose of this study is to consider the safety of workers when a robot runaway, which has not been discussed so far. As a result, the runaway distance of the robot varies depending on the direction, and the additional term to protective separation distance is sometimes too large to ignore. By using the results, optimal placement of robot and worker considered runaway of robot was possible, and the importance of adding runaway elements to protective separation distance was also shown.

In our laboratory, we have been researching and developing an assembly task partner robot PaDY that delivers parts/tools to a human worker for assembly line. In ISO/TS 15066 on collaborative robots, limits are imposed on the relative velocity of a human worker and the robot. In this paper, we apply this ISO/TS to our PaDY. First, based on the ISO/TS, the maximum relative velocity limit between the human and the robot worker is calculated from the human worker’s body part which may collide with the robot and the mass of the robot. Second, we introduce the motion planning method for PaDY that satisfies the relative velocity limitation using the movement of the robot and the trajectory of the worker measured in advance. Finally, we evaluate the effectiveness of the proposed method using simulation.

The purpose of our research is to reproduce human-to-human physical contact by robot because human feel happy through touching. We made an artificial hand with flesh and elastic joints aiming to reproduce the characteristics of human body. We examined PAM (Pneumatic Artificial Muscle) as actuator applicable to this hand. We considered that contact force sensation, which is one of the human somatic sensation, is important. We designed a control method of the contact force of the fingertip by PAM (in this paper, think about the fingertip force against the plane of a hard plate). The hand has nonlinear component. It is difficult to control analytically. By the experiment, we made a table of the inner target pressure of PAM that apllies desired contact force to the plane (we considered the angular relationship between the plane and the back of the hand), and we verified the table.

In this research, we aim to develop an estimation system of minute sound source position so that it is buried in ambient noise. A system integrating sound source estimation system based on time difference detection model engineering applied human auditory mechanism and micro sound detector using stochastic resonance was created and its effectiveness was verified by experiments.

Swallow is an important part of the dietary process. However, the number of the patients with swallowing disorder is increasing. The process of swallowing is a very fast reflex action and therefore the mechanism has not yet to be solved. Instruments for observing swallow has been developing, for example, the Cine MRI, X-ray video, and endoscope. They all supply the fundamental information to analyze the actions that occurs inside the body. By using these analysis results of Cine MRI and X-ray video in terms of swallowing motion, we have been developing the swallowing robot which can mimic the human swallowing motion.

Estimating the exercise torque of the muscles around the ankle joint is effective in determining the treatment policy of contracture and spasticity. As a method for estimating this exercise torque, a method using an EMG is known. Up to now, a method for estimating the exercise torque with high accuracy by optimizing the integral interval and the proportional gain of the integrated EMG has been proposed. In this study, we propose an algorithm to estimate the exercise torque around the ankle joint by using an EMG and an ankle angle. Here, it is assumed that the exercise torque is generated by the group of plantar flexor such as the gastrocnemius and soleus muscles. Experimental results show that the estimation accuracy is improved by using not only an EMG but also an ankle angle.

A major goal of this study is to design the nervous system model for quadrupedal locomotion. We apply a two-level central pattern generator (CPG) to our nervous system model, which generates locomotion rhythm and reproduces leg trajectories. And this nervous system model has an afferent sensory feedback, which uses leg loading, to CPGs. A cat hind leg model, which has six muscle models to drive three joints on each leg, is controlled using our nervous system model. When the locomotion speed of this model was forcibly increased by an external force, this model autonomously adapted to its speed and kept walking despite not changing any nervous system parameters. Since similar phenomena have been reported in animal experiments, our results suggest that our nervous system is an appropriate model.

An insects ability to discern information from scents is critical to their self-preservation efforts. Most insects establish this sensory awareness through their antennae. When searching for the source of a scent, they flick their antennae in response to environmental factors. This pattern can be replicated using robotics by adjusting sampling ranges accordingly when conducting odour searching experiments. For this research, odour sampling devices were designed that adjust sampling ranges automatically and analyse the effect of sampling ranges on robotic odour searching performance. Experiments were conducted with a custom-built robot equipped with odour sampling devices. Testing concluded that odour searching success rates were highest when sampling ranges were adjusted according to behaviour patterns as opposed to an invariable sampling range. This indicates that sampling ranges adjusted accordingly are more valuable in odour searching.

We have been developing robots closer to living snakes in order to understand snakes's locomotion mechanisms with the constructive method. However, because the passive wheels are used for the abdominal mechanism, the continuity and edge shape of the snake's abdomen are not reproduced. Therefore, the purpose of this research is to develop a snake-like robot equipped with such an abdominal function of a snake on the premise that it is driven not only on a plane but also in a three-dimensional environment. The experiments show the fabricated mechanism mimics the continuity and edge shape of a snake abdomen. This mechanism withstands three-dimensional bending and can interpolate the discrete robot segments with a curved line. The robot can push concave portions and convex portions of the environment by this mechanism.

In this study, we constructed an mobile treadmill for measuriing adaptive behavior of an insect. We focused on the odor source localization problem and constructed the mobile treadmill. We used a male silkworm moth as an experimental insect which searches a female by walking. We evaluated the performance of the mobile treadmill by performing an odor source localization experiment. As a result of the search experiment, the silkworm moth on the mobile treadmill was able to reach the odor source with 100% success rate. Therefore we succeeded in constructing the insect behavior measurement system and driving system of the mobile treadmill.

The purpose of this research is to develop a fine-grained and whole-body model of rodents for detailed motion analysis as well as dynamic simulation. As the first step, we developed a whole-body skeletal model of rat based on X-ray computed tomography (CT) data with approximately 60 μm resolution. We defined joints based on anatomical landmarks and developed a whole-body skeletal model of 168 DOF. By using motion data of a rat, we estimated sequences of joint angles of the skeletal model by inverse kinematics (IK). Furthermore, we examined the consistency of motion analyses between our results and those of previous studies.

Researches on various bird type flight robots have attracted attentions. However, there are few researches on takeoff and landing. Birds have excellent legs and sense of balance that can stay even on unstable tree branches and electric wires, and they can stay on branches energy-efficiently. Mechanically imitating the landing function of these birds is very important for realizing the autonomous bird-like robots. In this research, we develop grasping mechanism imitating the structure of legs of birds.

Recently, the self-driving system for automobiles and the driving support system have been actively developed. In the shipping industry, the research related to autonomous and remotely controlled ships has been carried out to reduce seafarer's work load, support the seafarer’s shortage, decrease the operational cost and improve the working environment. However, many autonomous ship studies do not consider the recognition problem of other ships and obstacles using cameras during the night navigation.

In this paper, the recognition problem of other ships by the camera during the night navigation is discussed. The existing object detection library "YOLO v3" has been tested. Three processes of gamma correction, HSV conversion, and gray scale conversion were performed on the image for the object detection, and the recognition rates of four images including the original image were compared.

Recently, researches related to autonomous ships and decision support system for seafarers are actively carried out. In order to realize autonomous ships and decision support systems, the obstacle detection is one of important technologies. In this paper, YOLOv3 which is one of obstacle detection software using AI has been examined to detect other ships during the navigation. The pre-trained weight file provided by YOLO is applied. VGA camera, HD video camera and 360 degrees camera were tested to compare the obstacles detection performance.

Among underwater resource surveys, the quadrat method is difficult to investigate deep inside rocks. In this study, multiple ROVs cooperative control system was constructed to achieve intrusion into the shade and investigation by synergistic movements of independent arm part and base part. Especially, we examined the autonomous attitude control performance of the base part considering the change of posture of the base part passively generated by the movement of the arm part as disturbance. As a result, the base part maintained autonomously the arbitrary two-dimensional plane on which the arm part operates. Therefore, we confirmed that the cooperative control constructed in this study is purposeful for task demands.

Although resonant type contactless power feeding devices are suitable for long-term operation of AUVs, there is a problem that the resonance can’t be maintained when the AUV is disturbed by tidal flow or wave. This paper proposes to introduce Q value control to improve efficiency and robustness against alignment and load balance.

The measurement method using sound waves (e.g., SONAR) is general method of location measuring in the underwater. However, the accuracy of this method is not sufficient to measure the location of seabed mining vehicle. In addition, the optical positioning method can’t measure accurate position of seabed mining vehicle because the surroundings of the seabed mining vehicle are muddy during mining of resources. Thus, we propose a new positioning system that avoid muddy area. In this system, the measuring device is floated by float right above the seabed mining vehicle, thereby enabling the measurement avoiding the muddy area. The measuring device is installed in the tracker having a thruster. And the tracker keeps the measurement accuracy by following the seabed mining vehicle. In this paper, we describe details of new positioning system and tracker.

Underwater exploration robots are utilized for exploring underwater resources and surveying ecology of aquatic organisms. Biomimetic robots can conduct surveys without affecting aquatic organisms. A Manta robot with a propulsive mechanism using pectoral fins, one of which imitated an actual Manta’s pectoral fin, has already been developed up to now. In this Manta robot, the self-position is estimated using optical flow. However, the Manta Robot generated a pitch motion during swimming and affects the posture of a camera. To more accurately estimate the self-position of the manta robot, it is aimed at constructing a model to control the posture of a gimbal mechanism for a mounted camera accurately, in the framework of two-degree-of-freedom design.

Valuable seabed mineral resources such as rare earth elements are deposited 2–3 meters into the deep sea floor, and sampling and analyzing the seafloor is necessary to clarify how they are generated and distributed. The current general sampling method based on vertical drilling effectively acquires the geological profile in the vertical direction. However, it is inefficient for investigating deposits widely distributed in the horizontal direction because the sampling range is limited to the diameter of the pipe. We are developing a seafloor robotic explorer that can excavate horizontally and collect samples of rare earth elements, which may enable wide-area exploration as multiple robots could be simultaneously implemented to autonomously search beneath the seabed. The excavation depth of our previously reported underwater drilling robot SEAVO II was limited to 430 mm; the discharging outlet became buried in the ground, making it difficult to discharge the drilled soil and ultimately limiting the excavation depth. In this work, we introduce a discharging mechanism that implements water jetting to improve the excavation depth.

Under water robot is very usefull for getting visual information under the water. When the robot uses vartilcal thruster with water jet, it jummed the sediment on the buttom groud of the lake. It distubes observation of under water on dision for visual. It is good idea to use buoyancy adjustment deive for moving vartial positon under the water. We cosider the mechanism suite for small water robot.

In Japan’s Exclusive Economic Zone, there are vast reserves of valuable resources undersea. Demands for Ocean resources exploration are increasing, and robotics can be applied to the exploration. However, it is frequently difficult to realize complicated work by an underwater robot with a general metal robot hand. To solve this problem, we will focus on soft materials for robotic hands in this study. To design and develop soft robotic hands that can be used in the deep sea, we developed a water pressure proof measurement system for soft materials and conducted a basic experiment.

The hovering type underwater robot has a disadvantage that the number of thrusters is larger than that of a cruise type / glider type underwater robot. Therefore, instead of increasing the number of thrusters, we are developing an underwater robot that can deflect the thrust direction of the thruster and propel it in all directions.

Underwater robots capable of performing seafloor exploration are highly useful to discover resources and unknown creatures. However, traditional underwater robots equipped with screw propellers may be harmful to surrounding environment and ecosystem. This motivates us to develop a robot that incorporates the shape and function of aquatic animal, in this study, a stingray. We have developed a stingray robot consisted of soft material and cartilage that mimics the structure of natural counterpart. We show that the use of cartilage can significantly improve the swimming speed.

The preferred propulsion mechanism for most commercial underwater robots is the screw propeller. However, it has a negative effect as it can hit surrounding creatures and plants. Another way of propelling is to mimic fishes, which can move very efficiently and quietly. In this work, we verified how to derive the propulsion speed using the resistance received at propulsion.

Nowadays, underwater vehicles have been studied and developed for seabed search. It is necessary for the underwater vehicle to search for a long time in deep sea in order to explore efficiently. We have developed a dual-eye-based docking method for underwater battery recharging to enable the Autonomous Underwater Vehicles (AUVs) that can operate for extended period of time. In addition, the 3D pose estimation for real time docking performance was improved by using active/lighting 3D marker against turbidity under dark environment. In real sea, we have studied that the robot could not perform docking continuously due to the influence of changing the ocean current direction occasionally. As a solution, we have proposed and constructed the rotary docking station which can control the docking direction along with ocean current direction. In this paper, we verify the functionality of the rotary station and the results of docking experiment using it.

Autonomous Underwater Vehicles (AUVs) have been developed for seabed resources exploration. For that purpose of AUVs, it is necessary that an they are able to work for a long time to accomplish these works efficiently. Therefore, to extend active time of the AUV, researches have been conducted on autonomous returning and docking system of the AUV with a power supply station under the surface of the sea. We designed and constructed a dual-eye-based docking system for underwater battery recharging. In this system, the AUV recognizes the light-emission 3D target object installed in the station. However, a recognition performance of this system is affected by turbidity of seawater in the actual sea. To verify the influence of turbidity, evaluation experiments on recognition performance were conducted. In this thesis, we analyzed the recognition performance with different turbidities and recognition distances in dark and turbid environment.

In recent years the leg typed rovers has attracted attention as a exploring rover with high running performance. However when the leg typed robot walks on a rough terrain with slope, problems are found. This problems are breaking rough terrain and slipping the leg typed robot. In this study, we focus on this problems and propose solution. We focus on vibrations and attach the vibrating parts to the legs to give vibration to the ground. In this method, two effects are expected; increasing the shear strength of the ground. Increase of sinkage of the legs. First, in order to confirm effectiveness of the solution, we confirm changes in ground when vibrating the ground. Next, multi-legged robot mounted a vibration generator is made. Effect of proposed solution are confirmed using multi-legged robot. Experiments of walking on rough terrain with slope are carried out using a multi-legged robot. In experimental results, the effectiveness of the proposed solution was confirmed.

Recently, many multilegged robots that can move dynamically like animals are developed. In order to realize jumping motion by a multilegged robot, because the behavior depends on the maximum power, high-speed rotation and high torque actuators are required. Therefore, a leg mechanism which can obtain high output power using spring elements and vibration for supporting actuator is proposed. An experimental apparatus that consists of a single leg mechanism equipped with a torsion spring around its joint axis and a vibrator mechanism utilizing inertial force by rotating eccentric weight is designed and developed. The resonance angular frequency of the system is estimated and the behavior of the mechanism against the vibration is shown.

In this paper, we introduce a mechanism called bi-articular muscle mechanism to a walking robot and aim at shock suppression control for landing impact by floor. The bipartite muscle mechanism is capable of outputting a large force in the direction connecting the toe and the waist of the robot, and there is already a prior study on vertical shock suppression control using this. However, this technique has not been optimized for bipedal walking. Therefore, in this paper, we focused on suppression of impact during landing of bipedal walking and study control of suppressing shock caused by floor reaction force by using bi-articular muscle mechanism.

In recent years, the search robots and rescue robots have been investigated which have wheels, crawlers and legs. However, those mechanisms have some difficulties such as (1) easily turn over, (2) possibly break down during falling, and (3) being crushed under fallen debris. This study proposes a new design of a soft mobile robot which can (1) move even after fall down, (2) absorb pressures applied on its body, and (3) move to all directions. Specifically, an Icosahedron Tensegrity is used for the unit of the body which is able to flexibly change its shape according to external forces. Legs having appropriate directions and sizes are attached at the end points of the compression members. It is shown that the legs push the body to a desired direction by changing distances between adjacent compression members with motors. During moving, the robot is responsive to external forces by passively deforming its body. To demonstrate the performance of the proposed mechanism, a prototype of the robot has been developed. The prototype has its own battery and controller so that it can move around by itself. The results of the experiments show that the robot is able to move forward by using the legs. It is also shown that the stride is more powerful and longer than the Tensegrity type robots which use only the gravitational force to move.

Many studies on human’s musculoskeletal structures and neural systems have been conducted in order to derive the principle of human walking. This research report describes a musculoskeletal bipedal robot that we have developed for verifying some reflex-based neural network walking controllers. The developed bipedal robot has sufficient range of joint motion for realizing reflex-based walking. We conducted a preliminary walking experiment on a treadmill and demonstrated that the biped robot can walk for 8 steps with a phase-based walking controller.

This paper deals with control of our four-legged robot with a pressure sensor on each foot. By using a pressure sensor, we measured the force on the toe, which we can utilize to control the angular velocity of the toe trajectory. We use FSR as a pressure sensor due to its light weight and thin dimension, although its accuracy is not good and it has only one dimension. We conducted experiments and succeeded in discrimination of gait. Then, we implemented the control method mentioned above and found that the gait changed to the bound gait from the walk or the trot gait. We think that this result caused by some mechanical conditions in the experiment.

We are aiming to realize an efficient running motion of a quadruped robot with compliant legs by the periodic input control method. The periodic input control is a method that can control the energy of a system operating periodically. First, energy control systems by the periodic input control is designed to each four elastic legs. The robot can hopping at a certain height by the function of these systems Next, it is designed synchronization pattern control system that synchronizing energy control systems with arbitrary phase differences. The robot can run while hopping at a certain height by adjusting the phase differences between the legs. Finally, it is considered that the improvement of running speed by adding a kicking motion. The effectiveness of designed control system is verified by numerical simulation.

In order to investigate gait generation and adaptation mechanism of animals, we propose a not mathematical but physical model connecting body dynamics and sensor feedback. In our previous study, we showed that CPG (leg controller) using leg loading/unloading for the leg phase transition can integrate rhythmic motion control and posture control. In this study, we intend to make a model of split-belt gait adaptation in decerebrate cats by cerebellum using a quadruped robot Kotetsu. As a result of experiments, we showed that inter-leg coordination mechanisms (ACM, CCM) and self-inhibition mechanism (SIM) play an important role in split-belt adaptation. While comparing gaits between decerebrate cats and Kotetsu, we found that our model is still not sufficient in spite of a significant similarity. Spec. and movies of Kotetsu can be seen at http://www.robotlocomotion.kit.ac.jp/kotetsu/index-j.html.

In this paper, we propose a method of manufacturing a tactile sensor using 3D printer and pressure sensitive conductive rubber. Furthermore, we propose to incorporate reflex control to avoid stumbling into Follow-the-Contact-Point(FCP) gait control. With this method, the robot performs reflex movement and prevent stumbling what is impossible with original FCP gait control. We verify whether the walking ability of the robot improves by introducing the sensor and reflex control on the real robot.

This paper describes the development of a hexapod robot which can perform both walking and rolling locomotion with less actuators. For this robot, we have designed a mechanism to preserve the both capabilities, utilizing arc outer shell, without the necessity for additional resources beyond those required by walking. The effectiveness of the design has been verified through experiments, which has shown the developed robot can move 0.04 meters in per second with the walking and 0.25 meters in per second with the rolling. The experimental results demonstrate the potential of our developed hexapod robot to be able to adapt to different environments.

Ventilation ducts has become important in urban high airtight houses. However, due to the long-term use of ducts, dust accumulates inside. Therefore, cleaning inside the duct is required. In the conventional cleaning method, there is a method of pushing the cleaning tool inside the duct. In this method, the deep portion of the duct cannot be cleaned by the friction between the cleaning tool and the inner wall of the duct. Also, cleaning time is limited. Therefore, the authors propose a peristaltic motion type robot for duct cleaning using a high contraction unit as a method of cleaning up to the deepest part of a duct in a limited time.

In this research, we are developing a robot platform that runs in a drainage aisle to autonomously. This robot aim to clean of drainage ditches in hygiene control of buildings handling food. Previous research we are developed a platform and angle estimation algorithm for using two infrared distance sensors. However, these infrared sensors has many noise, therefore, angle estimation has many errors.

In this paper, we report that implementation results using a noise reduction and experienced an autonomous moving test.

Recently, Japan is one of the aging-society countries due to low birth rates. Aging-population leads to labor shortages in the country in which cause the labor market under pressure. Thus, various robotics researches have been carried out to overcome labor shortage issues. One of the building construction tasks is to measure the illuminance of newly constructed buildings efficiently. Therefore, researchers developed mobile robot systems for navigating the building and measuring environmental illuminance autonomously. Most of the existing robot systems are solely depending on laser range finders for obstacles detection and navigation. However, objects such as ladders and cables are neglected by the existing robot systems which may cause a collision during the navigation. In this paper, we propose a multi-sensor system that integrates laser range finder data and depth camera images for objects detection and avoidance. A series of experiments are conducted to validate the effectiveness of the proposed method.

This paper presents a new method to advance and to carry wired sensors like a camera or washing hose inside narrow and curving pipelines, which is aimed at inspecting and cleaning inside the pipelines. To fulfill this request, two improvements were added to the conventional robot "adapting to different-diameter piping" and "making direction steering with a branch pipe" possible. In order to ascertain the effectiveness of the proposed method, we conducted a propulsion experiment in a pipe imitating the actual environment with an improved robot, realized stable forward and backward movement and direction steering while checking the camera image at the head.

Periodic cleaning is necessary for the duct which performs ventilation inside the house. However, existing duct cleaning tools are difficult to clean thoroughly. Therefore, a duct cleaning robot is required. In the previous study, we focused on a peristaltic motion type robot and developed a peristaltic motion type duct cleaning robot for 75 A duct. We conducted a cleaning experiment. The usefulness of a peristaltic motion type robot was confirmed from a speed of 6.1 mm / s and a cleaning rate of 99%. Therefore, in this paper, we first develop robots that are compatible with 50 A ducts that are popular in Japan. However, it is difficult to achieve the target speed simply by miniaturizing the conventional robot. Therefore, we propose speed improvement by the control method focusing on the dynamic characteristics of the unit. Finally, we conduct a driving experiment and confirm its usefulness.

A mobile mechanism by wheels is efficient on flat road surface but not for step climbing. A crawler is good for step climbing but inefficient on the flat road surface. On the other hand, drone can move in air, but it is limited on the payload and moving distances. Therefore, in this research, a new mobile mechanism combining wheels with drone is proposed and its autonomous control is researched to realize efficient movement on both flat and step surface and overcome the limited payload.

Fence following system for a cleaning robot of fattening piggery is developed. Environmental conditions of the fattening piggery include a narrow and long corridor, various type of fence, and misty by water for washing. This paper proposes the fence following system by using SOKUIKI sensor (LiDAR) and describes some experimental results in a real fattening piggery.

Regimented society performing maintenance such as buildings and high-rise apartment are carrying out staircase cleaning. As the conventional staircase cleaning becomes cleaning while going up and down, worker is burdensome and risk falling. Therefore, this research develops a cleaning robot that performs staircase cleaning. This paper examined cleaning method and down method of the stair cleaning robot to be developed. Moreover, prototype was made based on the cleaning method and down method examined, and the operation verification was carried out. As a result, it was possible to demonstrate superiority of step clearance ability and cleaning ability of the stair cleaning robot which was prototyped.

Recently, it is highly expected that robots work instead of human in a dangerous site such as disaster area. Sufficient working ability is required for such robots as well as moving ability. Thus, we have presented a method for autonomous gripping and carrying of a polyhedral shaped object by a mobile robot with twin arms based on plane detection in our earlier study. However, in this method, it is necessary to give the positions to observe the object in advance. Therefore, we aim to present a method for calculating observation positions, and to increase working ability of the robot. We apply the method to a quadruped tracked robot and verify its effectiveness in an experiment for autonomous gripping and carrying of a box shaped object.

This paper describes a spiral-shaped soft-swimmer with propulsion force change by geometry change. In nature, microorganism has spiral-shaped flagella. The softness of flagella is important to realize intellectual swimming such as velocity control. Spiral-shaped soft-swimmers were continuously formed by using a bevel-tip capillary. Soft-swimmers were rotating by applying rotational magnetic fields that were generated by two axial electromagnetic coils. We confirmed that propulsion force of soft-swimmers with different pitches by tracking the motion of surrounding microbeads (f_mag = 5 Hz, 20 mT). Type 1 (p = 418 μm) was no propulsion force, and propulsion force of Type 2 (p = 1638 μm) was larger than Type 3 (p = 2640 μm). This result indicates that the relevant pitch size is necessary for efficient propulsion. We believe that our micro-swimmer could pave the way to various microscale biochemical applications such as autonomous soft-robots and soft micro-probes for intricate environment.

A biocompatible soft micromachine is desirable to manipulate small biological objects like cells. This paper evaluated thermal responses of thermal responsive microgel aggregate to utilize as an actuator. The microgel aggregate was fabricated by gathering both microgels charged positively and negatively based on relative diamagnetic assembly. The aggregate shrank when it was heated up to 40 °C and expanded when it was returned to room temperature. The area shrinkage ratio between 25 to 40 °C was about 9 %, which is comparable to conventional actuators. The result shows that the aggregate of thermal responsive microgels is applicable for an actuator of soft micromachine.

Recently, muscle tissue and cells have received much attention as novel actuators due to their superior characteristics as actuators. Many research groups fabricated tissue engineered muscles with C2C12 myoblast cells in vitro. In spite of their struggles, the muscle tissues reported so far can generate force less than 1/10 of native muscle tissues in a living body. This study develops an automated culture system which can automatically apply electrical stimulation and replace medium and evaluates contractile force of muscle tissues. As a result, electric stimulations appropriate for muscle cells increased contractile force of tissue engineered muscle. This result suggests that the developed system is able to improve tissue engineered muscle.

We have proposed cell manipulation method on a microfluidic chip based on a vibration-induced flow. We achieved various two-dimensional cell manipulations in previous studies. To realize three-dimensional rotation of cells, 3-DOF actuator was required in previous study. However, vibration-induced flows in vertical (Z) direction is observed when we applied a vibration in XY direction by using 2-DOF actuator. Thus, these vertical flow can be applied for three-dimensional cell manipulations. In this study, we propose three-dimensional rotation method of cells by using X or Y vibrations with 2-DOF actuator.