In recent years, many of the piping systems used in Japan are more than 50 years old, and the risk of rupture due to aging is increasing. That is why, Research on mobile robot which can move inside pipe has proceeded. However, general mobile robot can only be used in large-diameter pipe. In this study, micro wheeled mobile robot for Small-diameter Pipe-inside test is invented. Main parts of the robot are made by stereolithography 3D printer. Also, a micro ultrasonic motor whose torque per unit volume and rotational speed is high is used as the actuator. The highest rotational speed which is really measured was about 3000 rpm at 40 V_p−p. In this study, a geared motor combining a micro ultrasonic motor and a planetary gear system will be fabricated, and the mobility performance of the robot equipped with it will be evaluated.

Our previous study has developed a flexible device that can switch suction and air blowing with input air flow control. This device generates a flow in the opposite direction simply by changing the input flow rate. The smaller the device is, the more it can be mounted in a limited space. In this study, we miniaturized the flexible device. The dimension of the miniaturized device is 25 mm in length, 15 mm in width, and 3 mm in thinness. The elongated flow paths were formed in one piece using a 3D printer, instead of using a mold to form them.

This paper proposes a soft and silent robotic finger driven by the Twisted and Coiled Polymeric Fiber Actuator (TCPFA), which is one of the flexible and silent polymeric actuators. In order to overcome the narrow movable area and slow responsiveness, which are the typical drawback of the TCPFA, the proposed finger configures that a pair of TCPA is arranged to imitate intrinsic and extrinsic muscles of a human finger, and an oil-cooling system is introduced. Several fundamental experiments are conducted and through these results, the increase in the range of motion and the improvement in responsiveness are verified.

The purpose of this study is to develop the ankle wearable device that constructed with artificial muscles and the pneumatic friction clutch to train leg motion directly. The developed clutch is constructed with the air bellows, the friction member and the wire. Pull-out of the wire is constrained by the friction force of the friction member increased owing to compressed foce from the bellows . As the artificial muscle can be kept at initial length, this friction clutch reduces displacement dependence of an artificial muscle. In this paper, the structure of tthe pneumatic friction clutch is described and then the influence of the device on gait are confirmed experimentally.

The purpose of this study is to develop a soft robot hand that can grasp an object, which has various shape or characteristic. By combining bending motion of a soft actuator and vacuum suction using a vacuum ejector, the developed finger can both grasp objects and suck object surfaces. In addition, as a vacuum ejector is introduced in order to generate negative pressure, both bending and suction movements can be realized by using only a positive pressure source. In this paper, we describe the control method for bending and suction pressure in order to keep simultaneously the desired value and then discuss the method for switching grasp/suction movements according to an object.

The researches on autonomous operation by imitation learning have been reported nowadays, but they have not accomplished yet reducing estimation error in untrained situations. It makes imitation learning challenging for use in practical situations. Then, there is room for the expectation that convergence to the command value after multiple trials by controlling the difference between the trial result and the command value in the case of a task that command value is given quantitatively in every trial. We verified by conducting an experiment if it is possible to measure a substance even if it was untrained. As a result, it was confirmed that the trained model could measure an untrained substance within 12.36 grams of the standard deviation of the target weight.

We propose a new method for solving the learning problem in the door opening task based on Bayesian optimization. The method is evaluated in a simulation of a door with rotating handle and an omnidirectional mobile robot with two 6-DoF manipulators. In this method, the objective function of Bayesian optimization runs a simulation where the robot attempts to open the door by moving the grip in a direction indicated by a five-dimension problem space. The value of the objective function is the time until the robot encounters forces or torques above a certain threshold. The results indicate that learning the sequence of movements necessary to open a door requires no more than 15 trials for each movement.

This study compares the disturbance suppression performance of discrete-time unknown input estimators with current observer and prediction observer. First, a UIE with a current observer and a UIE with a prediction observer are designed. Next, simulations are performed by applying a unit-step disturbance to the designed UIEs. Finally, the response waveforms and magnitudes of the UIEs are used to compare their disturbance suppression performance.

Recently, soft actor-critic (SAC) is employed for robot control. Although its ability to maximize policy entropy is expected to achieve robustness to noise and perturbation in robot control, the priority of maximizing the policy entropy is automated based on equality constraint to lower bound. Therefore, sufficient robustness is no longer expected. To resolve this issue, this paper proposes a new automation method on SAC using a slack variable for handling inequality constraint. As a result, the modified SAC achieved the higher robustness. In addition, as a side effect, it successfully suppressed its outputs.

In recent years, on-chip soft actuators have attracted large attention because they can perform low invasive cell manipulations. In our laboratory, we use thermoresponsive hydrogel as an on-chip soft actuators for cell manipulation. However, it is difficult to design the actuator because the volume expansion rate of gel changes depending on the constraint area and its shape. Previously, we evaluated volume expansion rates of gel with various constraint areas. However, we didn't evaluate the expansion rates with various shapes. In this paper, we evaluate the volume expansion rates of gel actuators with various shapes. In the future, we will apply this result for actuator design.

A method to examine red blood cell (RBC) deformability has been proposed by passing RBCs through a slit-shaped channel that is narrower than the RBCs to observe their velocity changes. However, the deviation of the evaluated values is significant even if the RBCs are deformed in the same way as they observed in microscopic bottom view. There is a possibility that the RBCs may have different shapes when observed from the cross-sectional side of the channel. To verify this possibility, it is necessary to observe from the cross-sectional side of the channel. However, the conventional method of fabricating microchannel molds generates a large number of creases that disturb observation, on the sides of the microchannel. In this study, we propose a new method of fabricating molds while maintaining the fluidity of SU-8 as a way to reduce creases on the sides of PDMS microchannels. Experimental results show that we confirmed that our new method has a less blurred features compared to conventional method.

In this paper, we proposed a system for simultaneous measurement of multiple single cells in response to high-speed solution stimuli. Conventional methods have not analyzed cellular responses to millisecond-order solution stimuli at the single-cell level and have been limited to evaluating ensemble-averaged characteristics. The measurement system proposed in this study establishes a highly accurate single-cell response measurement by using a deep convolutional neural network to automatically analyze image information obtained by high-speed vision and fluorescence imaging, in addition to millisecond-order solution stimulation using air-liquid two-phase flow control. As a result, analysis that considers cell heterogeneity, which was not visible in the ensemble average characteristics of a cell population, was achieved.

We evaluated propulsion efficiency of micro-gel robot with two different propulsion method. We made flagellar robot and peristaltic robot that we have proposed, and compared an energy efficiency of them in order to consider which principle is more suitable for propulsion in micro-scale. These two types of robot were made of temperature-responsive hydrogel(PNIPAAm), and we actuated them by scanning light to heat the hydrogels. We fabricated these two types of robots with various characteristic length L and drove them by scanning light with a galvano scanning system. We evaluated the scanning speed of light which can drive robots with the fastest speed. Furthermore, we evaluated a propulsion efficiency (body/min) according to L and succeeded in get the results that suggests the flagellar robot is more suitable for propulsion in micro-scale.

Arabidopsis is a model plant for studying the mechanism of plant growth and development. Many studies on Arabidopsis requires the analysis of the root tip. Previously, biologists always cut the root tip and put them in enzyme solution to isolate the protoplast and analyze the RNA or protein of the cells. But this method has a disadvantage. The location information of protoplasts is lost, and more advanced gene expression analysis cannot be performed. In this research, we propose a single cell level protoplast pickup system to isolate cells from the root tip under dynamic environment. Specifically, the system integrates an optical inverted microscope, a camera for image analysis, a micromanipulator and piezo pump for single protoplast pickup and a chip for root immobilization and observation. Using this system we have achieved automated 3D pipette tip calibration and automated single protoplast pickup.

In micromanipulation, stable and low-invasive manipulation of the target object is an important technology. Conventionally, various types of microgrippers with different operating principles such as piezoelectric elements and shape memory alloys have been developed for manipulating biological materials such as microstructures and cells. However, since the operability depends on the shape and rigidity of the target object, it is necessary to use grippers that have suitable shape and rigidity for the target object. Therefore, there is a growing expectation for a novel microgripper that is less affected by the shape and rigidity of the target object. In this paper, we propose the miniaturization of a variable stiffness microgripper using photochromic materials that undergo a solid-liquid phase transition by light, and an increase in the rigidity change rate by changing the coating material. We report achieving a tip size of about 100 μm by adsorbing photochromic materials molded into fine particles onto the tip of a glass capillary, and achieving a rigidity change rate of about 100 times by using a hydrogel as the coating material.

We propose a method for fixing microorganisms using heat-sensitive gels. The purpose of this study is to selectively control the gelation range since the heat-sensitive gel hardens over a wide range in the previous study. First, we considered changing the arrangement of nichrome wires. Then, we used nichrome wires with smaller wire diameters. Then, the use of hysteresis characteristics was considered to take less time to gel and to maintain a certain gelation range. Finally, we succeeded in controlling the gelation range of the thermal gel.

To evaluate the pathology of liver organoids, measurements of mechanical characteristics such as viscoelasticity and local internal environmental conditions such as dissolved oxygen and pH are required. In this study, we studied on fabrication of a probe-type sensor to measure local dissolved oxygen concentrations. Conventional glass microprobe sensors are highly brittle, and it is difficult to measure multiple parameters with a single probe, then we have designed a structure in which a three-dimensional printed body is formed directly on top of a substrate with electrodes fabricated by photolithography. Here, a macro model was created and evaluated for oxygen sensor. Knowledge of controlling the thickness of the membrane, which is an important component of the sensor is obtained through experiments.

Optoporation is a technique for introducing drugs into cells and is expected to be a safe and efficient method because it is a physical technique. However, it has the problem of low throughput due to the irradiation of a single pulse laser to a single irradiation point and the manual switching of filters. These problems have made the search for irradiation conditions inefficient. In this study, we solved these problems by automating filter switching and using continuous laser irradiation. At the same time, we searched optimal laser power by evaluating the molecular incorporation rate and cell viability after laser irradiation.

We have developed a MEMS tactile/slip sensor that combines MEMS technology and vision sensor technology. This sensor is a flexible and high-aspect-ratio microwall array made of polydimethylsiloxane (PDMS). The contact force can be obtained by image processing the amount of deformation when this microwall array contacts the object. In addition, by performing image processing on the deformation of the microwall array when sliding in the horizontal direction after contact with the object, slipping can be detected. The relationship between the contact force of the sensor and the deformation of the sensor was clarified not only by image analysis but also by finite element analysis.

This paper reports fabricating a micro suction cup array using inclined-rotated UV lithography with liquid-immersion. Conventionally, micro suction cups have been fabricated with inclined lithography. However, such methods had limited inclination angles of the suction cup caused by the difference in refractive index between materials. The proposed fabrication method with liquid immersion exposure allows structures to be made at a higher inclination angle. The device created a patterned mold for a micro suction cup array on a glass wafer. PDMS was poured into the patterned mold to fabricate the micro suction cup array. The results revealed nearly double the inclination angle 51° compared to conventional methods. In addition, the suction force of the device was evaluated.

This paper reports a biaxial force sensor which is entirely fabricated in a single step with a fused deposition molding (FDM) 3D printer. The force sensor consists of two strain gauges and holder of the gauges. The strain gauges are printed with an electrically conductive poly-lactic acid (ECPLA) and show a resistance change to strain. 3D printers specialize in forming three-dimensional parts, resulting two strain gauges of complex shapes simultaneously without any specific jigs. Moreover, it is easy to change the details of the sensor from CAD software. As the result, the fabricated force sensor measured biaxial force with high independence.

In our previous work，Au was used as the resistance temperature detector for tactile sensors that can be detect both force and cold-warm sensation．However，it was difficult to detect even slight differences in the cold-warm sensation that can be perceived by humans．In this study，the VO2 film with a large temperature coefficient of resistance (>20000 ppm/℃) close to the human skin temperature is newly used as the resistance temperature detector for the tactile sensor．The developed sensor is capable of detecting differences in the cold-warm sensation not only due to material variation，but also due to slight differences in micro surface roughness．

In this paper, we present a prototype mechanism for tip growth in liquid using clustered alginate gel tube membranes. By clustering a single mechanism introduced in a previous paper, it becomes possible to control the direction of tip growth. Here, we focus on the interfacial buffer as a component of the mechanism, explain the theory, and present the results of the trial.

Recently, the fields in which cell culture contributes have been expanding such as drug discovery and regenerative medicine. Therefore, the labor required for cell culture must be reduced. In the general cell detachment method using proteolytic enzymes, centrifugation is used to remove proteolytic enzyme after cell detachment. And many manual operations are required to perform centrifugation. In this research, we proposed a cell detachment method that does not require centrifugation by shortening the time for proteolytic enzyme treatment of cells, collecting proteolytic enzyme before the cells are completely detached, and then detaching the cells by forced vibration. We conducted experiments to determine the time for proteolytic enzyme treatment and the vibration conditions of the forced vibration, and as a result, we succeeded in detaching 89% of the cells from the culture surface. This result suggests that the proposed cell detachment method effectively detaches cells from the culture surface.

Insects have a variety of sensors with micro-sensory organs such as compound eyes and antennas, and advanced walking ability with micro actuator mechanisms. If controlled by a control module, they can be used as small robots called insect cyborgs with high trampling performance, and are expected to be utilized in disaster site exploration and environmental monitoring. For use in disaster site exploration and environmental monitoring, we attempted to create swarm behavior by physically linking multiple cockroaches. In this study, we demonstrated that one of the linked cockroaches was induced to behave by electrical stimulation and the other linked cockroaches showed follow-up behavior, and we succeeded in controlling movement by electrically stimulating multiple cockroaches.

We have aimed to alleviate pain in diabetic patients by developing minimally invasive microneedles. In this study, a microneedle with a liquid suction mechanism was developed to extract a target volume of 0.3 μl of blood. Our approach mimics the capillary force and water-absorbing behavior of water strider legs, and we fabricated a plate to imitate their microstructure. The flow rate of water and blood was measured to evaluate the performance of the microneedles. The desired blood absorption volume was achieved by applying hydrophilic treatment to the microstructure. Microneedle with a liquid suction mechanism was developed based on the principle of improving capillary force by adding microstructures that mimic the legs of a crucial carp. The effects of microstructure shape and density on suction speed were evaluated using water and blood as suction targets.

For robots that play an active role in our living space, it is better to use legged robots that can move to places where wheeled robots cannot. However, in general, leg robots have a poor energy efficiency. Therefore, we consider a quadruped robot with a self-weight support mechanism to improve it. This mechanism constrains only the vertical displacement of the toe. On a terrain with small unevenness, the mechanism maintains the vertical displacement of the body. However, there is concern that the vertical height at the base of the legs may actually be decreased. In this paper, we use a prototype of the self-weight support mechanism to clarify the model of its height decrement.

Periodic input control method can control an energy of a system by modulating the amplitude of the periodic input using a deviation from a target energy.In this paper,we apply the periodic input control to dynamic walking control of a biped robot with a torso.The designed control system controls the energy around the roll axis of the robot by periodically moving the torso,and generates a walking pattern using a periodic solution of the VDP equation.Simulation results that stable dynamic walking can be achieved on a plane and a slope by controlling the torso are showed.

Human dynamic gait has a kicking phase, a swinging phase, and a landing phase, and a gait control with switching control is necessary to reproduce it. This paper proposes a walking control method for a biped walking robot with bi-articular muscles that uses position control during the swing phase and final state control during the kicking and landing phases. The control of the kicking and landing phases is achieved by simultaneously controlling the torque by the carrier motor and the position by the sun motor and ring motor of the bi-articular muscle mechanism. The results showed that the robot walked continuously with an error average of 6.25×10−4rad at the end of the kicking phase and an error average of 4.5 × 10−3rad at the end of the landing phase.

The aim of this study is to develop a efficient walking robot and walking support system. In this paper, we experimentally demonstrate a level walking of 2D passive walker by towing forces of weights. Moreover, we demonstrate transfer efficiency of the level walking by energy loss coefficient.

This study aims to verify the effect of feedback of the body tilt angle around the roll axis on a quadruped robot equipped with an autonomous gait generation method. In a previous study, an autonomous gait transition in response to speed was achieved by feeding back the body tilt angle around the pitch axis to a central pattern generator (CPG). Furthermore, it was shown that the method also contributes to traversing uneven terrain. However, these experiments were limited to experiments in the sagittal plane and did not include feedback of the body tilt angle around the roll axis. Therefore, we conduct an experiment in which the body tilt angle around the roll axis is fed back to the CPG. As a result, the phase difference between the right and left legs is improved in the galloping, and it is possible to traverse uneven terrain with pacing.

We developed a gait control system for a quadruped robot using policies trained by deep reinforcement learning. The system switches between multiple policies depending on the situation and handles the problems of fall recovery, staircase adaptation, and power efficiency improvement. We evaluated the gait control system and a navigation system using it through simulation and real-world experiments.

In this study, we developed a walking robot that replicates the structure of a cat’s hindlimb to investigate the contribution of the musculotendinous structure on walking locomotion. Specifically, we focus on the emergence of leg motion resulting from interlocking mechanism generated by articulated muscles and the interaction between the ground and the robot’s limbs. The experimental results show that the robot generated a leg trajectory and walking motion with stance phases and swing phases, even though only two motors in the hip joint were oscillated. These results suggest that the motion generation strategies taking advantage of the passive kinematics of the musculotendinous structure are effective in generating walking locomotion in the cat-like robot.

The authors proposed a method called“ CEPHEID(Ceiling Embedded PHoto Echo ID)“ , which learning model that can obtain three-dimensional information that takes into account the height to thecan identify illumination equipment by flickering pattern. This method enables self-position estimationwith illumination light, but it ’s limited to two-dimensional information. So we aim to create a deep illumination. In this paper, we describe the training results of a model that estimates the distance to a single illumination.

Research is underway to combine position estimation methods. This paper focuses on the conventional method using RTK-GNSS/IMU and LiDAR, and proposes a method to continue highly accurate position estimation by correcting position and attitude angles. In the proposed method, the position estimation by RTK-GNSS and LiDAR is first pre-measured to obtain the error amount. Then, we propose a method to correct the position error by applying the method to combined positioning. Although LiDAR position estimation requires highly accurate search initial values, the accuracy of LiDAR position estimation is degraded due to errors in the two position estimations. Therefore, we propose a method to compensate for the attitude angle error by using a conventional method using LiDAR and IMU and reflecting it in the search initial value of LiDAR's position estimation.

In recent years, autonomousu driving has been attracting attention, and is being actively researched by various companies and institutions. In particular, map-matching using 3D point cloud maps and LiDAR (Light Detection and Ranging) has attracted attention. However, this localization method is easily affected by the characteristics of the driving route, and various environmental factors such as vehicle congestion, vegetation, and weather can cause localization to fail. Therefore, it is necessary to spend a great deal of time evaluating and verifying the accuracy of localization in various environments through experiments before implementing automatic driving. However, this is costly and time-consuming, and there are many safety concerns.

In this study, we aimed to verify the robustness of localization in advance by reproducing the environmental changes on a simulator.

This paper presents an acquisition method of autonomous navigation models for mobile robots by deep reinforcement learning in the game engine Unity. In this study, the mobile robots are required to reach goals in virtual urban environments while avoiding pedestrians using only a monocular camera and positional information, without using any 3D range data from laser sensors. The navigation models are trained by performing reinforcement learning based on Proximal Policy Optimization and input images to predict the pedestrians’ movements including last five frames. As a result of curriculum learning, the robot was able to reach goals at a rate of more than 80% in both of the navigations on the sidewalk and in crosswalks.

Autonomous navigation systems of mobile robots using deep reinforcement learning with camera images have been studied. However, there are differences among images obtained during training in simulation and obtained in actual environments. The differences make it difficult for robots to be navigated in the real world using action models trained in the simulation. To reduce the differences, semantic segmentation is applied to the navigation system. The semantic segmentation can simplify camera images such as floors or walls. The experimental results show that the proposed method can obtain action models for mobile robots that can be autonomously navigated to a specified destinations with segmentation images as input. A ROS-based navigation system to apply the obtained models to the actual robot has been built and the experiments in the real world have been conducted.

We propose a novel navigation method applying an offline reinforcement learning method based on Implicit Policy Constraint to mobile robot navigation in environments with pedestrians. The proposed method utilizes a flow-based generative model for the behavior policy, and the latent policy is trained using a method based on Advantage-Weighted Regression. The proposed method is evaluated in a simulation environment.

Recent years, mobile robots are expected to provide security and transport in multi floor buildings such as condominiums and office buildings. To move between floors, robots need to use an elevator. From the viewpoint of labor shortage, robots need to be able to move to their destinations without human intervention. For this, robot itself needs to know the floor it is on. LiDAR is the most common sensor used in localization. However, with LiDAR, every floor could be similar data. So, need to use another sensor. Since WiFi signal strength varies with the floor of the building, WiFi receiving sensor is required for floor estimation. As a validation of the floor estimation we compare the likelihood of each floor in the entire area of the WiFi map using only WiFi information. As a result, we were able to estimate the floor where the robot is located in the building.

Zhou et al. proposed a mutual collision avoidance algorithm based on “Buffered Voronoi Cells (BVC)” for quadrotors. This approach allows the control input to be easily determined using only the relative position of the quadrotors. However, when it was applied to quadrotors in an environment with stationary obstacles, “deadlock” occurs sometimes in which the quadrotors could not reach the goal. In addition, it was assumed that each quadrotor could use the relative position of obstacles anytime, and it was not discussed how to get the information in the conventional method. In this paper, we propose methods to avoid deadlock and generate Voronoi seeds from the local information obtained by laser sensors mounted on the quadrotor. Then, we verified the operation of the proposed methods by simulations. As a result, we confirmed the quadrotors could move to the specified target position without colliding based on sensor data.

Universal Map (UMap) is a localization system consisting of a server and a client that matches images obtained from a camera with DB images with coordinate information obtained from a 3D surface model. The problem with UMap is that it is difficult to send correct coordinate information in areas like corridors where fittings such as doors, windows, and lighting are lined up repeatedly. In order to prevent such mismatching, we calculate the mutual similarity between DB images, quantify the indoor repetitive structure, and verify the position of the camera and the generation area of the DB images that do not cause mismatching.

We modify Monte Carlo localizaiton as it can eliminate noisy measurements from a LiDAR derived from unknown obstacles. This modified method makes each particle have its observation direction with a range smaller than that of the LiDAR. In the calculations of Bayes theorem and resampling, the directions of the observation are changed as particles do not observe unrecorded obstacles on the map for self-localization. We have examined this method in a simulator. The result suggests that a restriction of the observation range makes MCL stable at the expense of precision.

This paper introduces a novel software, opt4imu, for motion estimation from Inertial Measurement Units (IMUs). The software currently supports four types of optimization algorithms for error reduction, which fuse loop closures and zero velocity and gravity direction, and so on. Potential applications include sports racket/bat swing capture and gait estimation from foot-mounted IMUs, as well as mobile robot localization from external sensor pods. Our experiments demonstrate a significant reduction in processing time, achieving results in less than one-tenth of the time compared to previous MATLAB implementations. The software will be available under open source license at https://github.com/kiyoshiiriemon/opt4imu.

Quadruped robots are expected to contribute to a variety of applications such as inspection, search and rescue, work support, and transportation due to their mobility and stability. Among them, hydraulic-driven robots are differentiated from electromagnetic-driven quadruped robots in terms of their large payload and ability to carry and tow heavy objects by taking advantage of the high F/M ratio of hydraulic-driven actuators. However, to handle hydraulics, various hydraulic components must be included, often making downsizing, weight reduction, and actuators a challenging issue. The purpose of this study is to develop a compact and lightweight electrically driven hydraulic power unit for the Tough Runner, a hydraulic quadruped robot that aims to walk long distances independently.

Long-distance guidance of dogs requires a feeding method that don’t restrict the dog's range of movement. We have developed a dog-mounted feeder capable of feeding dogs on the move. However, it was unclear if the dogs-mounted feeders were adaptable to multiple dogs. Here we demonstrate multiple dogs feeding experiments and light-guided training by using the canines-mounted feeder. we confirmed that the dog-mounted feeder could feed with high accuracy, with a 93.0% success rate for stationary dogs and a 92.7% success rate for training for dog’s moving direction. Our result shows that the dog-mounted feeder can contribute to the long-distance training of dog behavior guidance.

Passive dynamic locomotion arises without the use of actuators or sensors, but only through interactions with the environment. Previous studies have shown that one-legged passive dynamic hopping (1-PDH), which maintains a constant torso posture, has stable limit cycles. However, the stability of 1-PDH with a tail has not been clarified. In this study, the effect of the tail attached to the hip joint of 1-PDH model on the stability was investigated by numerical simulation. The results revealed that the passive hopping with a tail with appropriate physical parameters improves stability and forward speed.

The walking system of a quadruped robot must be able to respond quickly to changes in posture and the environment around the machine. In this study, a multi-process walking system was designed by configuring a program for each leg based on the originally existing single-process walking system in order to improve the immediacy of the walking system. In addition, this study used shared memory for inter-process communication to align the timing of each leg’s drive. Experiments were then conducted to compare the communication speeds using the single-process walking system and the multi-process walking system, and the results were presented and discussed.

In this paper, we show a rimless wheel robot with elastic telescopic legs has the ability to overcome high steps through 3D numerical simulation and actual robot. Previous studies showed that a rimless wheel robot with elastic telescopic legs has high adaptability to steps in numerical simulations. However, as far as we can ascertain, no actual rimless wheel robots have been developed for walking environments with high steps. To develop the rimless wheel robot with the ability to overcome high steps, we create an actual rimless wheel robot. We then confirm the optimal elasticity of the elastic telescopic legs.

This paper describes the study on a legged mobile robot foot sole mechanism for use in unknown rough terrain movement. The mechanism has a structure consisting of three toes, and includes frictional locking and unlocking mechanisms by parallel links. The mechanism levels the bottom of the legged mobile robot’s foot by using the three toes when on unknown rough terrain. To stabilize the foot sole of the legged mobile robot on a level surface, the frictional locking and unlocking mechanisms connected to the three toes are used for the foot sole mechanism. Stable walking control of the legged mobile robot is simplified by the foot sole mechanism we developed.