We propose an improved design method for Unknown-Input Estimator (UIE) using Artificial Bee Colony (ABC) Algorithm. The cost function used in the previous method had a problem that the evaluation criterion was indirect. In this paper, we propose a new cost function using the closed-loop transfer function from the disturbance to the control target output.

Since surfaces have many variables and are integrable, the control theory for robots representing arbitrary surfaces is complicated. The kinematic theory is simplified in continuum manipulators by considering a manipulator and a model with piecewise constant curvature. This paper proposes a robot and a model with piecewise constant mean curvature for surface robots and explains its forward kinematics. The robot consists of a checkerboard pattern of sections with specified mean curvature and sections with zero mean curvature. The shape of the former is known, and that of the latter can be obtained by solving a simultaneous equations to minimize the area.

This paper proposes an anti-sway control method for overhead cranes. The proposed method sends position commands to the trolley under magnitude constraints on the velocity and the acceleration. Experimental results show that our method suppress swing angle of payload at various velocities, disturbances, and initial swing angle.

In order to achieve high positional accuracy with force control, To use admittance control is one of the possible candidates. However, there is a trade-off between positional accuracy and responsiveness and stability due to contact, and especially the viscosity of the virtual object contributes significantly to the stability. In this paper, we propose to expand the stability region of the above force control using series admittance impedance control through experiments.

With the current focus on renewable energy, there has been a move to build offshore power generation facilities such as this one. However, due to the lack of breakwaters in the vicinity of offshore power generation facilities, the rocking transmitted from ships is not reduced, making it dangerous to transfer. However, most of these systems are designed for large ships, and the development of a small system has not yet been completed. Optimal trajectory planning based on sway prediction using deep learning is performed for a small ship transfer system at sea, and simulation results show its effectiveness for sway beyond the range of motion.

In order to realize the deployment of a humanoid robot to a disaster site by parachute descent, it is necessary to generate a landing impact absorbing motion. This paper develops a dynamic simulator to demonstrate a motion of a one-legged robot with a Series Elastic Actuator (SEA) during the landing. First, the model parameters, such as spring and damping coefficients of the SEA, are identified with an optimization method by using the results of the impact experiment on the assumption of the linear spring-damper model of the SEA. Then, the dynamics simulation of the robot during the landing is carried out to evaluate the validity of the developed SEA model.

Imitation learning has been getting results in robot control that requires dexterous and complex control. However, it is difficult to learn behaviors that diverge from the same appearance in the training data by imitation learning. In the real world, state transitions branch from the same appearance, and in such an environment it is necessary to search for the appropriate actions by trial and error. So, we developed a model that can flexibly change its strategy to an appropriate action when it realize failure by introducing unsupervised action classification and failure recognition into autoencoder. In our experiments, we evaluated the performance in a door-opening task where there are multiple ways to open a door from the same appearance and the correct way to open it is unknown in advance. As a result, we showed that our model is able to modify failure by switching its behavior when it deviates from the training data.

Path planning is important for fully autonomous flight by a UAV. Since facility inspections using UAVs require flight in narrow spaces, we need to plan the path that can move in all directions to escape from dead end. Path planning in all directions inevitably increases the computational cost. Since UAVs cannot change their speed rapidly, it is desirable to be able to consider the effect of speed when selecting a path. We adopt a method of pre-computing a group of paths using the Lemason’s equation, because the method can take both direction of the UAV and its speed into consideration. We use BS-MPC for path following, which has both computational efficiency and high control performance. We report the investigation that these can be used to fly at what kind of environment.

In this research, we study on posture control of a planar one-link torque unit manipulator which has a revolute joint with viscous friction. The torque unit is a module that consists of an electric motor and a reaction wheel, and it can be set to an arbitrary position of the link as an actuator to control the rotating link with free joint. However, the system has an uncontrollable mode. There arises a problem such that angular momentum remains in the reaction wheel of the torque unit after controlling to a desired posture. In this report, we consider using a reaction wheel with air resistance. Then we show that the link can be controlled to a desired posture with avoiding the problem mentioned above even though the system has an uncontrollable mode.

This paper describes action planning for a dual-arm mobile manipulator aiming to move a chair to an instructed position. The initial posture of the mobile manipulator for task execution is determined by the evaluation based on the size of free space. This process includes the selection whether to use both arms or one of them. Next, QP optimization is used to calculate the suitable joint angles of the mobile manipulator to approach the target posture while satisfying the constraint conditions. In the proof simulation, the feasibility of the proposed method was examined by performing tidying tasks in several different conference room layouts.

Surface rubbing motion has been studied in engineering. Surface rubbing motion requires high accuracy. But, the tracking performance of robots is degraded by disturbances caused by the increase in motion speed. In this paper, multiple regression analysis is performed from the position, velocity, and distance from the fulcrum of the robot’s end-effector to compensate for the disturbance in the high-speed surface rubbing motion. In our experiments, we created a model that combines the terms of position, velocity, and distance, respectively, with playing a glass harp as a high-speed surface rubbing motion, and confirmed the compensation effect when applied to a robot. The proposed method showed improving the performance by multiple regression analysis with velocity.

Redundant manipulators can be defined as robots that possess more than the required degrees of freedom to accomplish a specific task. This feature allows redundant manipulators to achieve its target trajectory while changing its configuration according to the surrounding environment. When introducing robots in the workplace, a process called "teaching" is necessary to make them learn what to do. However, it takes time and technological skills to program their jobs. To increase productivity in frequently changing environments, we propose a method that automatically generates a trajectory to reach the target position and posture according to the work environment. The effectiveness of the proposed method was demonstrated using a cylindrical obstacle to simulate whether the robot can avoid the obstacle and approach the target position and posture.

This paper presents a new concept of control architecture for dual-arm manipulation. Controller switching is essential for robots to achieve complex tasks, and a switching mechanism naturally constructs hierarchical structures in control frameworks. However, the hierarchical structure makes parallel task execution difficult. To address this issue, this paper introduces an arbitrament mechanism based on the classical redundancy resolution method at the top of the control architecture. As a result, the architecture makes flexible decisions to maximize the robot capability. Dynamic simulation shows that the proposed architecture prevents a dual-arm manipulator from getting stuck even if its one arm gets so.

In this study, we consider the suppression of vertical errors in periodic motion of a robot, such as weaving motion by feed-forward control. Because feed forward torque is calculated based on the dynamical model, it requires the identification of dynamic parameters. So far, we have identified the parameters based on the sensitivity of the dynamic/static motion, however there is a problem of stability of the iterative calculation. To overcome this problem, we consider the sensitivity of feed-forward control to improve the stability of the iterative calculation. As a result, we confirm that vertical errors of the weaving motion are reduced by the control using the feed-forward torque of the identified parameters compared to the conventional method and the Least Mean Square.

Robot assisting human to put on and take off clothes, it is important that an end-effector is soft enough not to hurt a person and able to manipulate clothes. In this study, we propose a flexible suction pad that can hold fabrics. The suction pad has flexibility made of only silicone. In addition, in order to achieve a sufficient suction flow and hold a fabric, two spots of negative pressure are generated according to Bernoulli's theory to suck a sufficient flow, and the suction part is arranged so that these negative pressure spots can effectively suck up a fabric. The performance of the proposed suction pad is evaluated by measuring the suction flow rate and sucking-up test for a piece of cloth.

Twisted and Coiled Polymer Fiber (TCPF) actuators can generate relatively large force, however, they have a disadvantage of slow response due to their thermal actuation. This paper proposes a method to drive TCPF at high speed by using the resonance phenomenon of an elastic beam. Experimental results show that a TCPF with a slow thermal response of approximately 0.04 Hz can even excite vibration of approximately 7 Hz by using the resonance mechanism.

Roller-type two-way clutches have high torque loss and cannot rotate at high speed because the rollers slip inside. Therefore, it is not suitable for a device that needs to rotate at high speed. In this paper, we propose a two-way clutch mechanism using planetary gear meshing. In the proposed mechanism, a sun gear is not used, and the distance between the planetary gears is variable. When the input shaft rotates, the planetary gears can rotate at a distance from each other. Thus, the mechanism can rotate with low torque loss and high speed. On the other hand, when the output shaft rotates, the two planetary gears engage, and the shaft locks. We made a prototype of Φ44×22 mm and experimentally confirmed that the mechanism works as expected.

The purpose of this study is to develop a soft robot hand that can grasp any shape or characteristic object due to the expansion of the actuator applied positive pressure and vacuum adsorption due to negative pressure. The actuator made of soft material has advantage to grasp a complex shape or fragile objects and can pick up a flat plate and objects with large radius by vacuum adsorption. In addition, by generating negative pressure from positive pressure source, it is possible to operate this a system with a single pressure source. In this paper, we describe the outline of the soft fingers and then the performance of this finger is verified experimentally.

In order for future musculoskeletal robots to perform complex tasks that require a variety of body movements, the body must have a reasonable degree of freedom and complexity. In order to establish a design methodology for complex musculoskeletal bodies, this paper reports on how to create planar artificial muscles with flexible ends that are easy to fit to the robot surface and easy to manufacture.

In this paper, we reported a recipe for a McKibben-type artificial muscle that can be used in small to medium-sized robotics research to enhance the reach of biomimetic robotics and soft robotics. All the details of the recipe, the necessary purchased items, and video of the making process, are available at the URL given in the paper.

The goal is to achieve high performance by reducing the mass of electromagnetic motors instead of increasing the torque density or the output density of electromagnetic motors. Methods to reduce the weight of electromagnetic motors can be categorized into structural optimization, use of new materials, and modification of the torque generation principle. In this paper, we focus on structural optimization for weight reduction of electromagnetic motors. From many electromagnetic motor types, we select an IPM motor that can achieve high torque and high efficiency. We focus on the structure of the rotor core of IPM motor and design a lightweight structure using topology optimization with a normalized Gaussian network. IEEJ D-model is selected as the reference electromagnetic motor. As a result, although the torque ripple was larger than D-model, the average torque was the same as D-model and the structure of that was lighter than D-model.

Soft actuators made of flexible materials, such as rubber and polymers, have possibilities that conventional solid actuators do not have, such as no unnecessary force is applied and the object is not damaged, and the flexible mechanism allows quick operation. In this paper, we describe a transport mechanism that we believe can be achieved by using conductive threads. We believe that this mechanism can be realized by switching the thread to which high voltage is applied. In the future, we plan to use Ansys to obtain optimal values for the design parameters of size and tension, and to demonstrate that it is possible to actually transport objects.

The purpose of this study is to verify the effect of changing the material of only end on the contraction force characteristics of McKibben type artificial rubber muscle. In previous studies, it was thought that the hysteresis characteristics of artificial muscles were influenced by the material properties of the ends. In this paper, we made artificial muscles with different material properties at the ends. The contraction force characteristics were compared with the characteristics of artificial muscles composed of a single material and the calculated values of the artificial muscle model. From the comparison, the effect of changing the material properties at the ends was verified.

Previously, an underwater ultrasonic motor with suspended rotor has been reported. The rotor (also called opposing block) has tilted surfaces that imitate impeller. On the other hand, rotation of rotor without the tilted surfaces was observed. The rotor has through holes. In this study, effect of hole shape on rotational characteristics was experimentally investigated. As a result, it was confirmed that inverting the rotor induced inverse of rotation direction. It was also confirmed that rotational characteristics varied depending on length and position of holes. To clarity rotation principle, water flow around the rotor was observed. A hypothesis of rotation principle was derived from observed water flow and rotational characteristics of some rotors. Contrarily, rotors with rotational characteristics incompatible with the hypothesis were found in the experiments. Therefore, further investigation of factors that act on the rotational phenomenon will be required.

This paper presents a manipulation method for interacting swarm of objects based on visual-motor integration learning. Here, a mixing manipulation problem is treated, in which a single probe mixes the swarm of objects in a closed environment. First, from images of the swarm of objects, an index of mixing level is obtained employing the gray level cooccurrence matrix method. Then, the probe trajectory to increase the mixing level as high as possible in a time limit is obtained by trial and error using reinforcement learning. The proposed method is validated in numerical experiments. The proposed method has a potential to recognize physical features of objects through their interacting behaviors, and to optimize the probe trajectories based on such features.

In this study, we aim to overcome the problems of model-based control by teaching human behavior using a master-slave system. First, we construct a master-slave control system for an easy control design. Next, we propose a method to generate the robot trajectory from the motion trajectory obtained from the teaching data using the control system. The proposed method can derive spatiotemporally averaged motion trajectories even for multiple instruction data that contain spatiotemporal variations. The effectiveness of the proposed method will be experimentally confirmed.

Efficient usage of crew-time is essential for sustainable human spaceflight activities. Throughout current ISS operations, crew (astronaut) are required to devote their precious time to stow and arrange various goods supplied to the station. In most cases, such cargo is transported and stowed in the flexible bag called Cargo Transfer Bag (CTB) and requires fixation to withstand launch vibration or to be attached during storage, not to float away, in micro-gravitation environment. As cargo-handling tasks are mostly repetitive and standardized, it is prospected to be automated for more efficient future space utilization. However, there are technological challenges in manipulating current CTB IF by robotic manipulator, especially when communication delay is expected. In this paper, preliminary ground testing results of current cargo IF manipulation by dual-arm robot remote control system is presented to outline the challenges in manipulation of CTBs.

In recent years, robots have been introduced to automate work in various industries. On the other hand, the automation of cloth handling is not so advanced. Cloth is an irregularly shaped material that deforms easily when force is applied, making it difficult for a robot to move in response to such deformation. In this study, we aim to realize flexible clothing handling using a robotic hand with a camera at its finger. We propose control methods that utilize video information, conduct verification experiments, and discuss the usefulness of the method. We also do the application experiment, and show the possibility that the task set as the goal can be automated.

In this study, we develop a method to predict the grasping stability by proximity-based local shape sensing from the fingertips of a multi-fingered multi-joint robot during the grasping approach process. In this paper, we first describe the improvement of the detection range of the spherical-shaped proximity sensor and the addition of a transparent coating to enable distance and orientation estimation even if the distance is 0 mm. Next, we create a calibration table for the improved sensor to estimate the nearest point, and show that the distance and azimuth angles of the nearest point can be obtained with high accuracy. In addition, we show that it is possible to predict the grasp stability qualitatively, and discuss the issues related to the local shape estimation.

We propose a method of estimating the state of an object using force information in a picking operation using a manipulator. The method is verified by a real robot system.The force information is collected by attaching a force sensor to the hand of the robot arm. The estimation is performed using a Manifold Particle Filter, which prevents the loss of particles, which is a problem of conventional methods. The object to be estimated is a semi-cylindrical object with a diameter of 60 mm.The force information was obtained from the sinking amount of the contact point which can be obtained by the force sensor.As a result, we were able to estimate the position of the object within 1.2 mm error from three times of force information.

In recent years, the development and research of robots has progressed, and they have been introduced in various fields. As one of the uses, robots are expected to perform work in the places that are too dangerous for humans to work. If the robots malfunctions in such a place, human can’t repair them. Therefore, other robots have to repair them.We targeted the connector removal for repair because the cable is involved in many other components of robot.In this study, repairing robot recognized the connector detachment point using communication between repairing robot and small robot. By communication, information that only small robots have can be used for repairs. Besides, the cable has a terminal unlike the strings, then repairing robot attached cable by taking advantage of its characteristics.

In order to automate the washing process of tableware in a restaurant, a robot needs to know the state of the manipulated object and determine how to approach and grasp it. In this paper, we describe a ROS-based grasp motion planning system to accomplish a task of grasping, transporting, and placing tableware. Our developed robotic hand can grasp various tablewares including heavy iron plates.We execute grasping and transporting stacked iron plates, and placing them side by side in a plate rack, and confirmed the effectiveness of the robot through preliminary tests, in a known environment.

Animals produce a variety of gait patterns depending on their speed of movement. It has been suggested that animals move energetically and efficiently by changing their gait, and many studies on gait patterns have been conducted in various research fields. For example, the bipedal crow generates a walking gait at low speed and shifts to a hopping gait at high speed. The ultimate factors that generate such gait patterns have been actively debated. In this study, we focused on the mechanical property of equal distribution of floor reaction force of legs, based on the idea that there is a possibility of falling if the driving force of legs is not equalized from the viewpoint of body symmetry and dynamics, and manufactured a one-input, two-output underactuated biped robot with a differential gear in the powertrain. Furthermore, we investigated the gait pattern of the robot caused by the interaction with the environment by controlling only the input value, and discussed the transition phenomenon of the gait by mathematical analysis. The results suggest that the transition of the gait pattern can be explained by a bifurcation phenomenon with the magnitude of the input torque as a parameter.

Leg’s stepping off from a step leads a legged robot to falling down. This paper proposes a kind of reflex control against such the situation for a hexapod robot. The robot has a leg which has a distance sensor on the leg-tip and has a posture sensor on the body to detect leg’s stepping-off. When the robot detects stepping-off, it steps out the slipping leg forward. The amount of the stepping-out is determined based on a condition of kinematic and potential energy. The effectiveness of the proposed method is verified via experiments in which a stationary standing robot steps off a step.

Real horses swing their heads while walking. We consider that the neck swing motion can be utilized for increasing the walking speed or stabilization when walking. In this report, we designed and fabricated a head and neck joint for our quadruped robot to investigate the effect of the neck swing motion on walking. We utilized Chebyshev linkage as a neck joint mechanism. We slightly changed the link length ratio from the Chebyshev linkage. We conducted a waking experiment to investigate the effect of neck swing motion on the robot’s walking speed. From the experimental result, we found that the average walking speed of the robot with the head is larger than the robot with without it.

This paper proposes a novel rimless-wheel type legged robot that can be made by adding the constraint of the rotation axis to the tensegrity structure formed by four rigid frames and eight springs. First, we introduce a planar 12-DOF robot model for passive-dynamic walking on a gentle downhill, and develop the equations of motion, constraint conditions and inelastic collision. Second, we numerically show that the generated passive-dynamic gait is not always asymptotically stable due to the variable impact posture, and that it is possible to generate a passive-dynamic gait that is faster than a normal rimless wheel by appropriately setting the spring coefficient.

This paper proposes novel combined rimless wheel models with a tensegrity structure formed by two cross-shaped frames and eight viscoelastic elements. We consider two different models; one is formed by two cross-shaped frames where two rigid frames that can rotate around a common central axis, and the other is formed by two identical cross-shaped frames where two rigid frames are fixed so as to be orthogonal to each other. Both models have 12-DOF, but the number of constraints during motion is different. We develop the mathematical models, and numerically investigate the basic properties of passive-dynamic gaits with respect to the slope angle.

3 actuators concentrated leg mechanism around thigh axis was designed for quadruped dynamic walk. The Dynamixel XM430-W350-T, ROBOTIS Co. Ltd., was selected as the higher torque actuator unit instead of Dynamixel AX-12A. Leg length was determined from the catalogue torque of Dynamixel XM430-W350-T and necessary torque to stand up statically and accelerate dynamically. Authors adopted Raspberry Pi4 as controller and developed an extension board for control. The trot gait walk was experimented.

Since it is not easy to incorporate the deflection of robot components and servo stiffness into the dynamics simulator, it is important to verify the motion control using not only the simulation but also the actual robot. In this paper, we constructed an experimental environment for walking in the sagittal plane of the biped robot CRANE, which has active degrees of freedom in the hip and knee joints and touches the road surface at points. We also report the development of a walking motion generation method based on the linear inverted pendulum mode.

Quadruped animals exhibit high-performance locomotion using the degrees of freedom of the whole body. In this study, we developed a running robot based on the passive dynamical mechanism of a simple model of a quadruped. We developed a legged robot whose two-link hind leg behaves like a spring similar to the SLIP model. The robot experiment showed that the robot with a rigid fore leg can achieve periodic bounding. However, the robot with a flexible fore leg can not achieve periodic bounding by itself, where the body pitch angle change was small. The results suggested that the robot has a stable limit cycle, and the pitching movement of the body plays a significant role in bounding.

Recently conflicts between wildlife and human becomes worldwide problem, but conservation of wild animals is also a current hot topic. We are developing a unique multilegged robot “Goat” series for animal menacing to reduce wildlife damage. For the goat series, we employed simple mechanical leg driving system - inverted Chebyshev link with 3D trot vectoring mechanism, so the durability is relative higher than conventional legged robot. We have already reported the concept model “BigGoat” and its partial mechanism. In this report, we will describe bear-size modified “BigGoat” with the wheel-leg switching mechanism, which enables the robot for practical operation in various fields.

This study aims to achievement stable running by a quadruped robot with an asymmetrical gait. The gait is classified into a symmetrical gait represented by a trot and an asymmetrical gait represent by a gallop. In the development of previous quadruped robot, autonomous gait transition from trotting to galloping has been achieved. It implemented a control method in which the central pattern generator (CPG) outputs were used as the motion rhythm and the desired angle of the PD control was switched based on the output. However, when the gait transitioned to gallop, the running posture was disturbed and the straightness decreased. The reason for this was that the leg trajectories of each leg were the same. As a result of generating the leg trajectory according to the gait based on the CPG, the straightness during galloping was improved.

In the ballroom, a couple of dancers dance with flexible and coordinated movement. This coordination is likely produced by physical interaction through their bodies. Understanding of the underlying mechanism could lead to the development of robots that can exhibit coordinated movements. In the present study, we propose a mathematical model of coordinated movements of dancers based on two inverted pendulums that are physically connected. In the proposed model, each pendulum is controlled so that dynamics of the other pendulum is exploited. We demonstrate through simulations that smooth movements of dancers can be well reproduced.

Due to growing needs for labor-saving, many logistics support robots for transportation are expected to coexist and collaborate with workers in future logistics sites. Those robots should be introduced while avoiding the decrease of transfer throughput due to interference between humans and robots or among robots. In addition, flexibility is required to respond to the ever-changing on-site environment and work progress. In this paper, we propose an algorithm that reduces the interference caused by those logistics support robots that perform tasks based on autonomous control. The proposed algorithm was validated using a simulator and showed results that reduced interference and improved transport efficiency.

Aiming at high stability for off-road transportation, our final goal is to develop an autonomous distributed mobile robotic system for various environments. In this study, we examine the system performance on cylindrical surface with zero Gaussian curvature as the first step of our work. The proposed mobile system consists of the homogeneous multiple units with a micro computer. The mechanical structure of the unit is designed based on the DELTA mechanism and provides the system with the function of legged movement. The operation of every unit would be synchronized over the simple serial communication channels among them. A series of experiments justified the feasibility of the developed system with two types of the gaits proposed.

Snake-like robots are expected in disaster rescue, because they can go through narrow space that we cannot reach. Wild snakes behave various style of locomotion depending on their species or habitat. By providing feedback this feature, we must able to develop more versatile robots. We attempt to develop the robot that can locomote various terrain as the initial stage. In this study, we use model-based reinforcement learning to control snake-like robots. We demonstrate that the robot with torque information has almost same performance trained on single terrain. By adapting fast fourier fransformation (FFT) into robot’s observation, we show the difference of policy in term of frequency.

Our research group is developing a modularized production system. The proposed system can change the size and shape of work space size and shape by combining multiple modularized rails. This study aims to develop a fluid and electric power method for machining robots that move on the rail modules. Fluid and power supply paths are installed in the rail module, and the machining unit receives these supply from the nearest rail module surface. The fluid supply path is constructed at the same time as the rail modules are connected. The pipe switching mechanism is developed in the machining robot to connect to the nearest supply port automatically depending on the position of the machining unit. Trolleys and collectors installed in the rail module and the machining unit respectively realize constant supply power. In the fluid supply experiment using the developed prototypes, a continuous fluid supply to the machining robot was achieved and time of supply intteruption is less than four seconds. Continuous electric power supply is also achieved without large voltage fluctuation.

Tubificinae is a subfamily of long and thin worm-like organisms. It forms a blob by entangling among soft and long bodied individuals. The blob can generate adaptive behaviors through physical interactions (e.g. entangling) of the individuals. For example, it changes its shape into a snake-like shape to move towards a better environment. Clarifying the underlying mechanism will lead to establish novel swarm robot systems that generate adaptive functions through physical interactions among soft and long bodied robots. However, mechanism that enables such blob to generate adaptive functions has not been elucidated. In this paper, we performed biological experiments and proposed a model that can recapitulate several behaviors of them.

Quadrupeds exhibit a variety of running patterns by orchestrating the whole-body degrees of freedom (DOF). This excellent coordination mechanism inherent in quadruped locomotion has attracted much attention so far. However, previous studies have focused on specific animals, and there has been no unified discussion of their mechanisms. In this study, we propose a unified perspective underlying various quadruped running patterns, and a systematic design methodology for morphology-dependent whole-body DOF coordination mechanisms. For this purpose, we focused on the motion of the torso, having the majority of the body mass, and systematically constructed control laws for two body models: one with an active joint for head nodding, and the other with an active joint for body bending. As a result of the simulations, we successfully reproduced the running patterns depending on the body model.

Soft-bodied legged robot shows high adaptability to the environment by utilizing its flexible body mechanics. However, it is difficult to generate adaptive coordination patterns between limbs (i.e., interlimb coordination), and decentralized control could be a solution for this problem. To develop such control mechanism, we focused on wandering spiders that exploit well-balanced coupling between hydraulic and muscle actuations and aimed to construct the decentralized control mechanism that can reproduce inter-limb coordination. For this purpose, we employed a synthetic approach wherein we construct a simple mathematical model based on biological insights. As a first step, here we built a two-dimensional robot model with two legs and proposed a simple local reflexive mechanism based on the interaction between leg flexor and body fluid.