In recent years, robots have attracted attention as a new labor force. Therefore, it is thought that highly robust communication is required between robot and human. In this research, we try to realize the contact communication operation with high affinity between the human and the robot, using manipulators with expressive power equivalent to human beings. In this paper, we discuss the experimental results of the actual control of the position control and the study of the operation method of the scapula.

In this paper we estimate the principal contact between an unknown manipulation-target object and its unknown environment, when the object is moved while keeping contact with the environment. We derived objective functions for two types of principal contacts, by focusing on the constraints caused by the contact. By minimizing the proposed objective functions, the geometric information of a contact (a point of vertex, a line of edge, a plane of face) can be estimated. We verified the two objective functions by an experiment. We were able to estimate the geometric information of the contact.

It is difficult for a robot to operate a heavy object due to the limit of the payload a robot has. In this research, we propose a method for a robot with a small payload to operate a heavy object by attaching rope to it to reduce the load a robot takes. Especially, a robot manipulate a heavy object by pulling the rope and turn over the object. We realize this by dividing the task into four parts, recognizing the point clouds of the rope, planning the motion of pulling the rope, recognizing the point clouds of the object, and planning the motion of turning over the heavy object by pushing it.

In this paper, we propose a method to plan cooking tasks autonomously based on cooking recipes. When a robot plans a cooking task, it is not enough to use information on the actions and objects described in the cooking recipe. In this research, the actions and objects needed which are not described in the recipe but need to consider are clarified by expressing cooking tasks using a network structure. The proposed network can express the relationship between actions and objects and their state changes. Through experimental result of task planning based on a simple cooking recipe, it was confirmed that the robot could actually perform the cooking task.

This paper presents a method for generating robot motions for daily tasks by learning the relationship between daily life objects shapes and human motions. Robots are required to handle multifarious objects in various categories. However, it is difficult for robots to plan motions automatically for performing tasks because many objects, including objects in the same category, have different shapes. In the proposed method, object shapes are estimated from a statistical shape model and motions are generated by learning the relationship between object shapes and human motions via linear regression analysis. We applied the proposed method to a task of scooping beans with spoons. We evaluate the generated motions and experimental results of the tasks, which were performed by a robot, to verify the effectiveness of the proposed method.

This paper proposes a method for picking piled sushi topping. By observing a human motion picking a sushi topping, we propose two picking strategies where one is to insert a finger into the separation among two toppings with shaking the finger, and the other is to insert a finger into the separation between a topping and a table. Along with two segmentation method, i.e., plane segmentation and LCCP segmentation methods, we experimentally verify the effectiveness of the proposed approach.

The rimless wheel is investigated as the most simplified model of passive dynamic walking. Previous researchers lay emphasis on the mechanical design of the passive rimless wheel. Controller design that enables adaptive locomotion to the environment has rarely been studied so far. This paper investigated whether the rimless wheel can realize both adaptive and energy efficient function. To illuminate this uncharted area, we developed a simple-structure rimless wheeled robot and examined classification of slope environments with SVMs (support-vector machine). Experimental results show that our adaptive control approach, in which locomotion velocity is modified online through auto-classification of the slopes, allows the rimless wheeled robot to achieve the energy efficiency.

In this study, the necessity of parameter setting based on the characteristics of the real machine in the gain adjustment method of the quasi-continuous stabilization controller for nonholonomic systems is revealed. convergence characteristics in stationary value control of mobile robot with two independent driving wheels are discussed by experimental methods. the geometric simulation of the product Roomba corresponding to this model, the real machine specification base, and the real machine characteristic base obtained by experiments are used for the parameters to be compared. the effects of applying these gain adjustment methods are confirmed from the response characteristics of the robot. the effectiveness of this gain adjustment method and the need to consider the characteristics of real machine are verified from the results obtained. in this way, we considered the utilization and usefulness of this method for connected vehicles and other nonholonomic systems from the knowledge obtained by the verification.

In recent years, the small wheel typed rovers are focused on as an exploring rover. However, running performance of them are low on the loose ground. The ground is hardly compacted by wheels because weight of these rovers is small. Therefore, drawbar pull is low. Purpose of our research is to improve the running performance of these rovers using effects which are given by vibration to the ground. In this research, how to use vibration to improve running performance is confirmed. The difference of the ground before and after vibrating were confirmed through any experiments such as measurement of shearing strength and sinkage. these experiments were carried out using some shape wheels. Furthermore, some experiments using the testbed to run on the loose ground were carried out using some vibration timing patterns. From theses experimental results, we could get knowledge that how to use vibration to improve running performance for the small wheel typed rovers.

This paper proposes a new mechanism for an inverted pendulum type robotic wheelchair with the ability to climb stairs. This mechanism is developed to improve occupant safety and comfort. First, physical analysis confirmed that the proposed method met the purpose of ensuring the safety of the occupants. Next, the dimensional requirements necessary to develop the mechanism were clarified , and the mechanism was actually designed and manufactured. Finally, the operation was actually confirmed using the manufactured mechanism ,and problem were extracted

In recent years, environment-friendly personal mobility has attracted people’s attention. In this research, we developed a personal mobility that anyone can ride and operate. The usefulness of the proposed personal mobility was verified by comparing it with a wheelchair in this paper. As a result, the operation based on the user’s foot was higher than the operation with the button based controller.

Snow removing, traveling, and working in the snow area is a hard job for the aged person. The aged people require simple and small-sized robots replacing their hard works. We apply a robot equipped with Push-Pull Locomotion (PPL) for solving these problems. PPL robot is unique locomotion and has a good traveling performance on loose soil. The traveling performance on the snow surface, unfortunately, has remained unclear for the PPL. To investigate this performance, this paper focuses on the interaction between an individual wheel and snow surface, which is the key aspect of the PPL traveling scheme, and investigates the sinking phenomenon and resistance force of the wheel. Experimental results revealed that the wheel can easy to sink into the snowfield, and the wheel can obtain the resistance force.

Planetary exploration is being carried out for the expansion of the area for human and the development of science and technology. The planet surface is covered with regolith, which is loose soil, and have many slopes because of craters. Therefore, when rovers cross the slopes, traveling without lateral slip is required to avoid indicating miss direction. In this research, we propose a high traveling performance for rover against slope using four wheels independently by inching locomotion. This study focused on reduction of lateral slip by camber when wheels sank, and returning the path by changing azimuth when wheels were moving.

In this paper, we propose new behaviors of a snake robot to adapt to obstacle environment. By using these behaviors, the snake robot can maintain contact with obstacles. In addition, the robot can switch between maintaining contact and avoiding obstacles depending on the situation of obstacles. With this motion, it is possible to improve the moving performance of the robot in environment where obstacles are scattered. The effectiveness of the proposed movement is confirmed by simulations.

This paper introduces an autonomous mobile robot system based on a monocular camera that travels to destinations without any map building in advance. The mobile robot system selects velocity commands based on action policy trained by deep reinforcement learning. The input state of the trained action policy is depth information acquired from struct2depth using the monocular camera. Also, localization of the robot is performed by online visual SLAM using the monocular camera. This paper describes the system integrating velocity command selection, depth estimation, and localization. The proposed system configuration and preliminary experiments are provided in the paper.

The paper introduces a method of automatic map generation used for learning of action policy in mobile robot navigation. Road contours and waypoint candidates are extracted as map components by image processing from an image of existing electronic map such as Google Map. Road contours are extracted using topological structural analysis of the binarized electronic map image. Also, waypoint candidates used as respawn or destination points in learning system are randomly selected from pixels of road areas. The generated map can be applied to learning simulator based on deep reinforce learning system. The paper describes an abstract of navigation system based on reinforcement learning, a proposed map generation method.

Some objects, like gravels, sands or stones, comes into the crawler during these types of vehicle moves on rough terrain. These objects can be a key factor to hamper their locomotion. However, the causal relationship has been unclear. Therefore, the relationship between the sprocket rotation torque and the tension applied to the crawler has been investigated in this research. Experimental setup that can be changed the tension between two axes connected by crawler, and be measured both the sprocket torque, and belt tension was established to be shown the results of the transition of a tension of the crawler and a torque when foreign substance was caught into the sprocket.

As generalized in the previous report, camber angle control is essential for non-slip running on a curved trajectory in a wheel running system that runs with an axisymmetric rotating body on the ground. In this report, we propose a wheel control strategy to realize non-slip omnidirectional movement only by controlling the camber angle. The configuration of a truncated cone wheel as an example of its implementation and its application to a curved crawler running system using the wheel as a sprocket are discussed..

Recent years, a lot of people are developing innovative micro mobility, for solving the short distance travel in the city with traffic jam. Using Stilt-type Mobility, we can climb steps while driving and take it in the train. Until now, we didn’t research the usability of Stilt-type mobility. When someone drive the stilt-type mobility, it was vibrating so hard, and someone felt so slow for him. In this research, we suggest the new control system and confirm it for improvement the usability. In chapter 2, we suggest new system of movement control and weight control. In chapter 3, we suggest the gain optimization system using reinforcement learning.

The crawler type autonomous mobile robot platform that can traverse over rough terrain was developed successfully. The snow remover “Oh! Snow” on the market was utilized as a base vehicle. T-frog motor driver was applied to make easy a system development. Also, the mechanism around a rotary encoder, and the control circuit box were designed and implemented in the robot platform. In the driving experiment on rough terrain in the golf course, it showed that the crawler type mobile robot platform could traverse over rough terrain stably. The crawler type mobile robot platform is useful for the system development of an autonomous mobile robot which can traverse over rough terrain.

The paper introduces a study on acquisition of action models to avoid crowds by reinforcement learning. There are many dynamic obstacles such as crowds avoid for autonomous mobile robots in real worlds. Since it is dangerous to conduct training of action models in the real environments, simulation environments are used for training. In this paper, training of action models was performed by deep reinforcement learning in simulation environments where there are many pedestrians walking at random or along flows in various densities. Then, trained action models were evaluated in a viewpoint of arrival to the destinations. The results indicated that training in realistic human density is effective for traveling to destinations avoiding pedestrians, compared with humans in high density.

The multi-purpose module system (MMS), a modular robot, is possible to transform into various forms by manually connected. MMS needs an automatic desorption function to improve versatility. This study proposes a movement control method for MMS that can be automatic desorption. MMS is consisting of differential-drive wheeled robot which have a connector and a passive joint. Propagation control method is selected to drive the MMS. This method is suitable for the control of the modular robot because it does not depend on the number of modules. The proposed control method is evaluated by simulation. The simulation results indicated that the subsequent module followed the leading module. From the results, the usefulness of using propagation control for the connected modular robot.

For Self-completion of a 1/3 scaled model of a 4-wheeled and 4-leged personal robot manufactured using a resin-based 3D printer, improvements were made to put the circuit and battery in the torso. We examined a method of tilting the body of the robot to match the stairs gradient, and conducted experiments. The robot's front legs could be raised to the second step of the stairs.

Personal mobility (PM) taking parts in society is a challenge due to two major problems as follows. The one is difficulty in estimating passability precisely and proactively. The other is difficulty in controlling PM in a narrow space. Human operators are essentially not good at dealing with the problems, so we propose a proactive operation support system based on estimating passability index related to the height and width of passages. For height, the system detects steps taller than 5 cm, visualizes them on monitor, and provides emergency stop if collision is estimated. For width, the system extracts labelled clusters of walls and calculates distance for adjacent walls, visuals passages on monitor, and decrease the speed based on the width. The experimental results indicated that the proposed system enhanced the efficiency and safer operation, compared with no-support.

Recently, an inverted pendulum vehicle has attracted attention as a new type vehicle in a pedestrian space. However, there is a few researches on the improvement of safety of an inverted pendulum vehicle. The purpose of this study is to construct a driving assistant system with automatic braking for an inverted pendulum vehicle at sudden braking, and to examine its validity and safety through the experiments.

This paper outlines the research field of Riding Robotics (RiRobo). This research field targets small vehicles (small mobility) such as manned mobility, manned robot, riding robot, personal mobility, smart mobility, vehicle, riding, mobile robot. Small vehicles are fun, convenient and help our lives. This paper introduces research in the field of Riding Robotics, focusing on the presentations made in the organized session over the past two years. Researches on various aspects such as research on the structure, mechanism, control, and movement of vehicles, research on the body of a rider, research on the psychology of a rider, and concept proposal are being conducted. This field is an area where future development is expected.

Doors are one of the barriers for powered wheelchairs in daily life. It is hard for elderly people and people who have hemiplegic upper limb to open a door by themselves. A robotic arm mounted on a powered wheelchair can help opening a door. However, opening a door using a robotic arm requires difficult operations. Particularly, opening a door with a door closer needs difficult operations. Accident that a powered wheelchairs and a robotic arm collide with a door may happen. In this report, we propose a methodology of opening a door with a door closer. And we propose an end effector that can reduce the number of procedures for opening a door more easily.

Recently, simplified electric wheelchairs, that is, wheelchair with a motorized device attached to a manual wheelchair has been gathered attention. Simplified electric wheelchairs are relatively lightweight and can be folded. However, simplified electric wheelchairs have a risk that the wheelchair will fall backward when running over a step. To solve this problem, we developed the 3-wheeled simple electric wheelchair. This wheelchair has two large wheels and an active-caster, a type of powered caster which can move in any directions. With this structure, the active-caster saves as a support, the risk of falling backwards can be reduced when the wheelchair running over steps. However, 3-wheeled simple electric wheelchair have a problem that stability of straight running is poor. In this paper, we will clarify that deflection of 3-wheeled wheelchair traveling direction is caused by human control by experiment. Also, confirm the improvement of running stability by the angular velocity feedback.

Recently, autonomous driving has been developed due to increase of traffic accident, traffic congestion and tailgating. In this study, we mainly focus on shared autonomy, where an autonomous system shares control of a vehicle with a driver. The system in shared autonomy has a role to help the driver avoid dangerous situations without interference in the driver’s control. To this end, the system should know the driver’s skill and attention when driving. This paper, therefore, integrates imitation learning and attribution method for this purpose. The proposed method demonstrated to visualize the visual field that driver especially paid attention.

Vehicles are expected to support people in the compact city or aging society. However, vehicles with conventional wheels can not move in certain directions, such as the lateral direction, which reduces usability, especially in the limited space. In order to solve this problem, an active omni wheel, that can actively move in an arbitrary direction, was developed, but many motors are required to drive the wheel. In this research, we propose a mobile mechanism which can move in both longitudinal and lateral directions by using the active omni wheel. The proposed mechanism realizes multidirectional movement with only one motor by switching the transmission paths.

This paper proposes a new kind of personal vehicle which is driven by four omnidirectional wheels and whose body is dynamically stabilized. Unlike personal vehicles with two or more wheels, the proposed vehicle can move immediately in all directions and the body is always kept upright on the road surface of variety of slopes. The most distinctive mechanical structure is that the chassis and the vehicle body is connected by a single, free universal joint. Since the body is not directly actuated, the vehicle is an unstable, underactuated system and the body attitude needs to be kept stable by appropriate feedback control. The overall mechanical structure, its kinematic and dynamic models, derivation of the controller are described. Early expermiental results request for further refinement of models or design of the controller. Safety issues and automated driving or driver assistance are under development.

Stochastic resonance is a phenomenon, in which a weak signal resonates with noise and improve the signal-noise ratio. It has been frequently observed in the natural world. In this study, inspired by stochastic resonance in the tactile sensation of organisms, we verify the effects of noise enhancing the sensitivity of the biomimetic tactile sensor. In this experiment, noise is applied through a vibration motor attached to a tactile sensor and discriminate ten types of sandpaper by using a support vector machine. As a result, the accuracy was improved more than twice by adding noise. Furthermore, the signal-noise ratio is also improved. It suggests the sensitivity was enhanced by stochastic resonance.

Understanding the tribological phenomena between fingers and soft surfaces such as skin is important for the product’s design. In this study, we investigated the coefficients of friction when a bare finger slides over artificial skin with seven levels of hardness under dry and lubricated surface conditions. Under dry conditions, the coefficient of friction depended on the normal force only for artificial skins that were softer than the finger, whereas it was constant for harder artificial skins. When the surface of artificial skin was lubricated with mica, it depended on the normal force only when artificial skins were softer than a finger, as with under dry conditions. When the surface of artificial skin was lubricated with TiO2, it depended on the normal force even for harder skins.

This study verified that the realism of virtual roughness texture is improved by using vibrotactile and electrostatic friction stimuli in conjunction. These two types of stimuli have complementary features, therefore they are expected to enhance the perceived quality of stimuli. In our previous research, conjunct stimuli improved the realism of virtual surfaces with sinusoidal surface roughness profiles. In this research, we tested with complicated surface roughness, and experimental participants ranked vibrotactile condition, electrostatic friction condition, and conjunct condition in terms of the realism compared to actual roughness textures. The conjunct condition was jugged to be more realistic for complicated surface profiles consisting of multiple spatial wave components.

The gap detection threshold is an index of spatial resolution at the fingertip. This threshold value has been measured only on a flat plate such as an iron plate, and it has hardly been measured on a rough surface such as a rough surface. This paper describes how the gap perception threshold changes when the surface roughness of an object is changed. In the experiment, the perception threshold was measured by the method of constant stimuli by passively tracing nine kinds of width slits opened on three kinds of sandpaper. As a result, there was a tendency that the discrimination threshold decreased by the roughness of the texture, and no significant difference was observed, when the analysis of variance was carried out for the roughness factor of the 75% discrimination threshold.

Research on haptic devices aiming at application to virtual reality (VR) has been conducted. Most of them have to be fixed on a desk, and users can’t move around. To avoid this problem, exoskeleton type haptic device have been developed. However, there are problems such as discomfort caused by the reaction force of the presented haptic force at another part of the body. In addition, limitation of the rage of motion by exoskeleton structure is also problem. Therefore, in this research, we propose a wearable force feed back device using air jet without skeletal structure. In this report, an elastic ball presentation experiment using a prototype was conducted. Results confirmed the ability to render virtual ball with different sizes and stiffness.

This paper describes a tactile sensor employing camera that detect whether the shrimp has the shells. The curved tactile image sensor, which is made based on the reflective membrane method, acquires the tactile information by sliding the sensor along the shrimp’s surface. This method obtains fine surface textures with high spatial resolution. Post image processing was applied to detect the shell. In order to maintain the load applied to the object, a mechanism with a vertical slider and spring was introduced. Through experiments, the proposed sensor could detect pieces of shell remained on the raw shrimp.

Presenting kinesthetic sensation usually requires a large space so as to avoid users colliding with objects or other users. One way to tackle this issue to use kinesthetic illusion, which is a way of presenting kinesthetic sensation without physical motion. However, realizing dynamic motion and fast movement remains difficult. Considering that multiple synergist muscles are usually involved in a movement, stimulating multiple synergist muscles might enhance the illusion. Thus, we investigated whether multi-point vibratory stimulation to multiple synergist muscles enhances induced kinesthetic illusions. We found that stimulating multiple synergist muscles created more vivid illusions. Additionally, we found that our method was effective for inducing steady illusions, but not for rapid illusions. We also calculated the contribution of each proposed stimulation point to the illusion.

We have proposed a tactile image sensor employing event camera to realize tactile sensing with both high spatial and high temporal resolution. The event-based camera detects temporal changes of intensity in each pixel and generates events. Our sensor consists of a black elastomer which has 361 white markers inside and the event-based camera. Event outputs generated when the elastomer deforms and markers move. In this study, we propose force measurement using the event-based tactile image sensor. The normal force and friction force were estimated based on accumulation of event outputs. Also, friction coefficient was estimated.

We propose an artificial skin with a flexible thin film and a tactile function that requires no external power supply. The proposed skin is made from a lamination of an electret, an elastomer and two electrodes. The electret is a material in which electric charges are stored semi-permanently. In this study, an electret with a thickness of 15 μm was made by applying amorphous fluororesin on a 10 μm thick copper electrode. After that, an elastomer containing carbon black was applyed to the opposite electrode with a thickness of 250 μm, and these were overlaid to produce an artificial skin of about 300 μm. These were laminated each other to produce an artificial skin. The results of experiments in which the pressure was applied to the prototype skin showed that the detected, voltage was 0.5V, which was sufficient to used as a sensor for detecting the contact.

In this study, we construct a method to present the movement of an object at a position away from the body in order to provide experiences having higher presence and improve operability. We present a sense of object movement using lightweight and inexpensive vibrators, so we give Phantom Sensation, one of the tactile illusions. In previous studies, it showed a sense of object movement between vibrators, so we propose new algorithm for extending Phantom Sensation. The best algorithm was considered through experiments in which the participants felt a sense of object movement by the presented stimulus after training with each algorithm.

We have proposed a grounded haptic display that can widely present the force sensation using fishing rod. The display incorporates a fishing rod to present the force sensation at the user’s fingertip using a thread being reeled from the tip of the fishing rod. This device requires conversion from the position of the user’s fingertip and magnitude of the force to necessary winding force of the thread and the posture of the fishing rod. In this paper, we developed a control method and evaluated it by considering a case wherein the force to be presented is exerted in the vertical direction.

We propose a tactile display with high intensity using many small vibrators suitable for mounting. At this time, it is necessary to know the details of the relation between the presentation area which changes by the number of vibrators, the actual perception area, and the subjective intensity. In this study, we measure changes in subjective intensity and perceived area during vibration stimulation at multiple points.

Sensory evaluation is generally used to grasp tactile sensation of an object. However, sensory evaluation requires many subjects in order to evaluate the tactile sensation quantitatively, and it is time consuming and costly. Therefore, tactile sensing technology is attracting attentions instead of sensory evaluation. In order to establish tactile sensing technology, it is necessary to estimate tactile sensation of an object from information measured by a tactile sensor. In this research, we developed a tactile sensor and measured vibration information while the sensor traced on an object. Then, in order to estimate tactile sensation from the measured vibration data, we extracted features from the vibration data by machine learning.

This study analyzes the recognition process of examining the texture of fine particles through tactile sensing using human fingers and aims at establishing the structure of texture recognition. Characteristic values of the tactile sensing are studied in order to investigate a relationship between tactile sensing and its verbal response. In this study, the SD method (Semantic Differential method) is used to execute the language estimation (the evaluation value) of fine particles. This paper is intended as an investigation of a feature of KANSEI evaluation in their mid-teens, mid-twenties and around 60 years old using multivariate analysis.

This paper describes a mounting design for extracting electric signals from a Si substrate of a MEMS tactile sensor. In the conventional mounting design of MEMS, bonding wires have been widely used to extracting signals. However, in the case of a tactile sensor where the external force on the order of Newton acts on the device, the bonding wires have been cut very often. Therefore, we proposed a new design using a Flexible Printed Circuit that reduced the load on the contacts on the MEMS and confirmed the effect by measuring external force.

Humans judge their own skin condition through the tactile sensation experienced in sliding the fingers over their skins. However, there has been no quantitative evaluation method of tactile sensation of bare skin with the same accuracy as that of a professional evaluator. Therefore, we developed a sensor that could evaluate the tactile sensation of skin using the vibrations and friction generated as a motor-driven probe lightly slid over the skin’s surface. In order to verify whether the developed sensor could accurately evaluate tactile sensation, we performed a skin texture measurement test and created statistical models that estimate the palpation scores of a professional evaluator. The physical quantity used for the estimated model was only vibration, but the estimation accuracy was close to that of the model constructed from skin condition indices measured by several commercially available sensors. The developed sensor can be used for evaluating the tactile sensation of bare skin.

We have proposed a method for realizing the augmented reality in haptic by using electrical muscle stimulation in combination with actual haptic stimulation. The results of the experiment that compared to existing methods showed the effectiveness of electrical muscle stimulation. We thought it is important to examine the properties of electrical muscle stimulation. Therefore, this study measures the relationship between an input voltage and the presented power. Then we use the results to measure the force discrimination threshold due to electrical muscle stimulation. This paper considers the relationship between the obtained input voltage and presented force, and details the results of a psychophysical experiment performed to measure the discrimination threshold.

This paper presents a wearable fingertip tactile display for rendering of forces by skin deformation to index finger with six-Revolute-Spherical-Spherical(6-RSS) parallel link mechanism. It is composed of a static baseplate and a mobile endplate which are connected by six legs. The baseplate is located on the nail side of the finger and the mobile endplate can move in six-degree-of-freedom(6DoF), with contact-no contact capabilities, by the legs actuated by the motors. At first, we decided the operation range of the endplate on finger pad for rendering of orientation, stiffness, and weight. Secondly, we decided the parameters that include the length of arm and rod, the position where index finger are fixed. Finally, we designed a prototype using the parameters.

In this research, we propose an underground excavation robot that can move in the ground. This robot reproduces the movement of moles using biomimetics. However, there are few detailed documents about the movement of moles. Therefore, it is performed from the analysis of the operation of the mole. In addition, this robot is the same size as the mole to omit the consideration of material properties and viscoelastic analysis. The mole is about 15cm in size, and this robot uses a hydraulic actuator with a low output weight ratio. This report reports on the analysis of the behavior of moles and the design of soft actuators.