We have presented the availability of our slip sensor using pressure conductive rubber for detecting initial slip, however, we have not revealed the principle of high-frequency wave occurrence which takes the detection method. Clarification of the principle is required for optimized design of the slip sensor, especially the property of the pressure conductive rubber and the shape of the detecting part, and for reduction of individual difference of detection characteristic of the slip sensor. This paper discusses the principle through a series of experiments and finally shows that localized fixing and peeling between pressure conductive rubber and electrodes in the slip sensor configuration has influence with the principle.

View full abstract

This paper proposed a flexible tactile sensor using a flexible wiring, which was laminated copper and polymer. A urethane gel of low Young's modulus material was filled in the gap between the electrodes. This sensor can be achieved high sensitivity and wide load range due to thin urethane gel. The sensitivity, resolution, load range in linear sensitivity, and load capacity of fabricated sensor were 0.5 pF/gf, 0.1 gf, 100 gf, and 900 gf, respectively. The waiting time until the electrode returned to an initial position was 10 min. When the sensor was set on a circular pillar (10 mm in diameter), the sensitivity and resolution were 0.45 pF/gf and 1 gf, respectively.

View full abstract

Motion capture system has been used in various fields. At present, optical motion capture is most popular motion capture system. However, optical motion capture is often complex, costly and limited by the infrastructure itself. Here we propose to use stereo camera information and Inertial Measuring Units (IMU) to obtain a precise motion measurements at all time. The combination of the two systems thus overcomes the weaks point of each. On stereo camera information, we use optical flow and stereo processing. IMU data are obtained using coordinate transformation with correlation coefficient. Occlusion is resolved using correlation coefficient. To build a motion capture system using sensor fusion by these methods.

View full abstract

This paper presents the design and implement of a three dimensional measurement system employing our developed 3D location poistioner. To continuously observe borders according to obstacles and walls, a commercial laser distance sensor is mounted on the crank of our location positioner. Specifically, the integrated measurement system is installed into a mobile robot platform to apply it to explanation and surveillance activities in unknown real-world conditions. Our challenge is placed on how to realize a three dimensional measurement scheme. We will describe the implementation and measurement scheme in detail. To demonstrate their effectiveness in our laboratory environment, extensive experiments are shown.

View full abstract

In this paper, we propose an autonomous navigation system. To examine the traversability robustly, the system adopts the following technologies. The system involves a localization with GPS and dead reckoning, and integration method using Divided Difference Filter (DDF). The system also estimates traversability of road regions using a laser range finder (LRF). Lastly, the system detects moving obstacles. The effectiveness of the proposed system is proved through some experiments.

View full abstract

In this paper, we propose three-dimensional coordinate measurement using a stereo camera and a projector. In order to perform stereo matching of an object, a specific pattern is projected on the object's surface. Stereo matching gives advantage in measuring the shape of the object precisely by comparing the similarity between input and template image. To verify the precision of the system configured, the shape of a planar body was measured by using the proposed method. As the result, we were able to measure the shape of the object correctly. We verified the accuracy of the depth measurement which is within error of ±1.00 mm in an optimal situation.

View full abstract

For a lot of people, dominant and non-dominant hands may be related to both degree of skill and mastery of movement. Study on difference between dominant and non-dominant hand movement may lead to elucidation of human being's mechanism. For the hand movement, it has been confirmed that the dominant hand is superior in many cases. However, investigation between them has not performed as much for memorial movement by hand. Therefore, we performed experiments of reproducing lines by participants after teaching using Haptic interfaces, and study its difference for the memorial movement by both dominant and non-dominant hand. The results show that the difference are not confirmed for the memorial movement by hand.

View full abstract

A force feedback device can display a kinesthetic sense to an operator. Previously, several types of force feedback devices have been developed, however, an ungrounded force feedback devices are still not in practical use. Ungrounded force feedback devices do not require any origin of a reaction force outside an operator, which extends the range of working space to the infinite. Therefore, it is technically competent for a mobile interface. The present study develops a prototype of an ungrounded force feedback device using a gyroscopic torque. This prototype uses a Control Moment Gyro (CMG) as an actuator unit that generates a gyroscopic torque. Our prototype, which has four CMGs arranged in a pyramid form, generates torque in arbitrarily directions using the gyroscopic effect. The kinesthetic sense is displayed to an operator by the synthesized torque of the four CMGs.

View full abstract

This paper presents a novel ungrounded force feedback device which makes a user feel a sensation as if a pulling force is displayed by utilizing non-linear characteristics of human perception. The force feedback device is mainly composed of a spring-mass system and a voice coil motor (VCM). A weight, which connects to the VCM through the spring, hits against a wall periodically to generate a one-directional impulsive force. Then, a user grasping the device perceives the pulling force to the designed direction despite a real constant force is not displayed. The device is 140 x 41.5 x 41.5 mm3 with total mass of 230 g. A sensitivity of the pulling force perception were evaluated and presented in this paper.

View full abstract

In attempt to improve haptic interaction and manipulation technique in surface computers, we propose an interactive haptic feedback display that utilizes 'direct touch' on a surface computer without using a mediation device. The proposed haptic display lets users interact with the virtual objects inside a pressure-sensitive screen directly as if the user's finger can penetrate through the surface and sink in to the digital world inside the screen, as well as receive haptic feedback direct to their fingers from the display. Using the proposed display, users were able to perceive touch sensation with haptic feedback when they come in contact with virtual object surfaces and discriminate different stiffness using vibratory and pseudo-haptic feedback respectively. This innovation creates a new breakthrough in virtual interaction with more accessible and realistic haptic interaction approach for surface computing.

View full abstract

Generally, Transformation and Elasticity of the elastic object is reproduced by Finite Element Method. In the methods, the problem in calculation of deformation and stress for elastic object is response delay. This paper presents a haptic rendering method for the elastic object using non-linear prediction of the finger-tip position. Using the non-linear prediction of the finger-tip position, the delay of the response in canceled.

View full abstract

We have studied a caterpiller robot whose forward motion can be obtained simply by applying Q learning method as one of reinforcement learning methods. In these studies, the robot was acting on the basis of objective rewards obtained by sensor. Therefore, this study will focus on the subjective reward. It examines the emergence of the motion using Q-Learning with subjective rewards in order to clarify the difference between the learning results. This examination provides a novel perspective on human robot interaction (HRI) via reinforcement learning.

View full abstract

The dynamic plane covering by swarm robots is a problem to make a trajectory of move multiple robots in a given field so that the trajectory covers the field, which often appears in agricultural works such as agricultural chemicals spraying. We solve the problem by applying the swarm robot concepts in the way that a single robot called a leader moves around in the field and all the other robots called followers try to catch up the leader. The advantage of this solution is that we only need to specify a trajectory of the leader and we expect the scalability of the number of the robots. On the other hand, we have to carefully design the trajectory of the leader and the connectivity network between the followers so that all the robots are always connected and under control. In this paper, we concentrate the gyration swarm and analyze the stability of the case.

View full abstract

The objective of this paper is to develop a differential wheel mobile robot for a loose soil environment. The loose soil works as a resistance force to the robot, and it can be changed by a speed, slip, soil properties and so on. Therefore, motion control of the robot turns to a difficult problem. In this paper, we develop the dynamics model of the mobile robot with the wheel model in the loose soil environment. We investigate the model with numerical experiments.

View full abstract

In this paper, a generalized EKF-employed swarm robot localization system with RSSI-based measurement is developed, and the operating limit of the system is analyzed numerically. Generalization of the swarm robot localization system is performed by parameterizing various aspects of the system and environment, such as noise of sensors, and layout of robots and positional features in the environment (landmarks). With two sets of experiments, small-scale and large-scale, we obtained the fine granularity of statistics about position estimation error per parameter combinations of the system and environment. In conclusion, we developed the system, and the boundary of the operating limit in terms of less than 10cm was explicated.

View full abstract

Multi-robot systems (MRS) can be expected to solve a task which one robot cannot perform. In MRS, reinforcement learning is one of promising approaches for controlling each robot. However, its performance depends a great deal on the segmentation of state and action spaces. To deal with this problem, we have been developing a new technique which segments state and action spaces autonomously, named BRL. In order to improve the learning performance, this paper introduces mechanism of selecting either of two state spaces: one is parametric model useful for exploration and the other is non-parametric model for exploitation. We investigate our proposed technique, by conducting physical experiments for a cooperative carrying task with three autonomous mobile robots.

View full abstract

An Island GA is one of the parallel models of GAs. In an Island GA, each sub-population evolves and is processed in parallel. Therefore, an Island GA is suitable for parallel computing. Now then, GPU (Graphics Processing Units) which has better performance than CPU is attracting large attention in the parallel computing domain. Consequently, GPU is expected to be applied not only for graphics processing but also for general computation applications, and this technology is called GPU computing. This paper proposes an effective implementation method of an Island GA with CUDA based on reduction of the on-chip memory usage in order to apply to real-world problems.

View full abstract

The sheet metal inspectors in automobile industries know that the knitted work gloves help the perception of surface undulation. The knitted work gloves have some touch enhancing mechanisms. We developed the tactile display based on their mechanisms. This tactile display can give larger deformation stimuli and clearer perception of micro surface undulation than our previous display.

View full abstract

This research has developed the User Interface, called Vibration Alert Interface (VAI), which presents information by the difference in the frequencies of vibration. Much of conventional vibration presentation equipment warned the user by on and off change of vibration, however, VAI is to change a frequency of shown vibration, and its degree as well as warning of just danger can be even shown. But actually, it's studied that the vibration sensing ability isn't fixed because of disturbance, for example, the user's state. To reduce these influence, a system with reproducibility is requested, therefore it has been studied variously so far. Accordingly, this paper contrived the method to suppress dispersion of vibration sensation from the different angle, presentation method of vibration. It's stated that the experiment, which investigates reproducibility of vibration sensation, was made using the Relative Vibration Sense Presentation Method, and that this presentation method improved repeatability.

View full abstract

In this paper, we propose a novel method for remote haptic communication. The method uses active touch on a user's own body surface. The user replicates a touch motion of the other one by tracing the touch position and pressure. A prototype enables detecting the touch position on the body surface by using a shirt with a 2D position pattern and a hand-held camera device. The paper describes a user study to examine the psychological changes during experiences of the proposed system.

View full abstract

The purpose of this research is to invent the device that capable of obtaining visual information without direct touching the object. In this paper, we describe the prototype device consisted of a microcontroller, vibrator and distance measurement sensor. Frequency changes appropriately to distances to a target pointing the device at. Using obtainable distance and position of hand by moving the device, target's position is perceptible for user. In addition, user can consider a mental map and recognize target's shape without direct touch.

View full abstract

Haptic feedback system is required to provide reaction force resulting in the robot hand to the operator. In this paper, we develop the force feedback device which can provide operator's multiple finger knuckles with reaction force. The force feedback device presents reaction force in thumb, index finger and middle finger. The device is very compact , lightweight and high haptic resolution by using a Shape Memory Arroy as a clutch brake.

View full abstract

Forward dynamics models are the important basis of motion planning of robots. While recent advancement in Robotics has allowed diverse morphologies of robots to take place, there has been little study on identifying the forward models for such robots. In this paper, we propose to use neuroevolution algorithms, which optimize the parameters of neural networks, to learn such models. We show that a neuroevolution algorithm called CMA-NeuroES can learn models of a simulated double pendulum system with higher accuracy and generalization ability compared to conventional regression algorithms, such as Gaussian Process Regression, Support Vector Regression and Linear Regression.

View full abstract

We clarify a design principle, i.e., emergent mechanism of distance measurement function, by using evolutionary robotic system which designs morphology and intelligence autonomously. In nature, a living creature measures distance to a target by tactile sensors in the case of short range, by a stereo vision in the case of middle range, and by memory of the past state in the case of long range. In this paper, we gave the robot a task keeping distance to a target and investigated functions generated by created body structure and control method. The result showed that the strategy for distance measurement changed from the contact type, to the binocular disparity type, to the binocular disparity with memory type, and to the memory type depending on the target distance as well as nature principle.

View full abstract

In this paper, we propose a GA method to optimize time variant system that has redundancy. GA does not take into account the dynamic environment in general. Accordingly, under such environment, the solution cannot maintain optimality. Therefore, we focus on characteristics of the Distributed Genetic Algorithm in this study. Moreover, in order to make the characteristic more remarkable, we used clustering technique based on genetic information. In a first experiment, to compare the nature of the cluster distribution at the time of convergence. In the second experiments, to examine the validity of proposed method to the time variant problem. We apply our proposed method to the a traveling salesman problem. The results show that the proposed method provides a performance that is superior to that of the conventional Distributed Genetic Algorithm.

View full abstract

Small-world and scale-free properties are important features of networks in the real world, which will make learning in such networks difficult. One of the defining characteristics of small-world networks is a high cluster coefficient. In the present study, processes of reinforcement learning in complex networks with different clustering structures are investigated. From the results of the experiments, clusters in the networks of learning spaces could be learned in more trials with relatively high discount rates and/or rewards. Thus, clusters in the learning spaces may destabilize a process of reinforcement learning within a certain range of parameter settings. It is discussed that solutions affected by the clusters are less optimal, but they may be more robust.

View full abstract

Robots are expected to be able to cooperate with and assist us in our daily life. It is necessary for robots to perform a variety of tasks that were required by ordinary people who is not skilled at the control of a robot. In such a case, a direct teaching seems effective to program robot's motions. Since ordinary people may sometimes teach improper motions, it is necessary for a robot to learn appropriate motions autonomously using instructions including improper motions. To achieve this, we propose a control method inspired by a nonparametric model. In this method, a robot selects an action out of instructions where the fitness of each data is determined by a reinforcement learning algorithm. We also analyzed the properties of the trained controller.

View full abstract

Swarm robotic systems (SRS) are a kind of multi-robot systems which consist of many homogeneous autonomous robots without any global controllers. SRS is expected to have some valuable properties, however it is difficult to generate the desired behavior because of intrinsic redundancy in itself. As a problem dealt in the SRS, generation of flocking behavior has been cited. In these researches, each robot can perceive wide area and communicate between the robots. However, wide perception area and communication may adversely affect the design of SRS. In this paper, flocking behavior is generated in mobile autonomous robots characterized local sensing and non-communication between the robots. Moreover, as a method to generate flocking behavior, we implement the algorithm based on self-propelled particles model to each robot.

View full abstract

Well-trained craft-worker's touch plays an important role in the manufacturing industries. For example, the sheet metal inspection in automobile industries is conducted by human hands. We developed a tactile display which presents designed surface by controlling height, slope and curvature of a sheet. In this display, designed surface is presented on a human palm. And the display can move freely and work on a flat surface. As testing results, it is confirmed that the proposed tactile display has a potential to present designed surface.

View full abstract

We are developing a multi-modal 2.5D display capable of stimulating tactile and force sensations to achieve a new human-friendly interface. In the present study, we extended the versatility of the operation in 2.5D space for object manipulation in 3D space. Although the surface rotation of the virtual tactile pad was limited around the z-axis in the previous display, the rotations around the x-axis and y-axis are managed by the centroid position of the operator's fingers in the new display. We conducted three series of evaluation experiments. In Experiment I, we confirmed that the centroid estimation was stable around the center of the display pad. In Experiment II, we revealed the proper proportional constant for the vertical movement of the virtual tactile pad. In Experiment III, subjects tried to compare the size of two virtual spheres to evaluate the 3D shape presentation capability of the present display. Since the sensation thresholds for each subject were low enough, we proved that operators could recognize a fine distinction of sphere radii generated by the proposed display.

View full abstract

This paper describes a haptic device that represents a viscous force and that consists of nonmagnetic metal and a permanent magnet. Using an aluminum plate and neodymium magnets for the device, a user gets resistance at the time when he moves the aluminum plate over neodymium magnets. Then, the device generates force in proportion to an aluminum plate velocity because the force is based on an eddy current. In order to verify the properties of this force, we measure it by mean of motorized linear stages and a load cell. The paper shows the produced power is proportional to relative velocity between an aluminum plate and a neodymium magnet. The paper also describes the device can give us the force which have a periodic vibration.

View full abstract

For the purpose of designing autonomous robot that can be utilized at a social scene inhered in living environment of humans, developmental model of the robot based on internal behavior representation is presented. Framework of the model is discussed in terms of subjective interpretation of sensory-motor flow that reflects temporal interactions between behaviors of the robot and an external environment. Depending on constraints due to bodily natures and environmental characteristics, internal behavior representation is segmented in proposed framework. Alternatively, target-oriented time series would be induced by internal evaluator corresponding to emotion of living objects. The model is oriented to deal with an unsteady environment, including physical and social dynamics. In this paper, importance of equipping invariant criteria of value in such environment is advocated.

View full abstract

In order to realize a robot which works in cooperation with a human, the dance partner robot which dances ballroom dance with human is researched. The dance partner robot needs to estimate which step human steps on next. In this paper, the authors pay attention to the dance partner robot's step estimation system and propose the method to estimate dance step transition based on Hidden Markov Model and "Mimesis Theory" in order to estimate step transition in higher accuracy. In addition, the authors make "Proto-Symbol Space" from the power that is added to the robot from a human when a human dances with the robot.

View full abstract

It is difficult for a model predictive control to deal with an alternation of the inside and outside of systems. This paper proposes a new predictive control system using a recurrent neural network and a fuzzy control system. The proposed method can predict a ball track by forming the orbital model of a ball on recurrent neural networks.

View full abstract

This paper proposes a new reinforcement learning approach for acquiring conflict avoidance behavior in multi-agent systems. To verify the effectivity of the proposed method, we apply the proposed method to the narrow road problem that many agents go by each other in a narrow road. In the narrow road problem, it is the optimal strategy that agents select the different behavior from other agents. The proposed method can differentiate into agents preferring to move forward and agents preferring to give way, by means of the reinforcement learning using reliability parameters.

View full abstract

This paper proposes a new reinforcement learning algorithm which can learn in the multi-dimensional input spaces containing different-species sensors by means of CMAC and neural networks. We apply the proposed method to a robot arm reaching problem and verify the effectivity of the proposed method.

View full abstract

Recently, the rice cultivation without herbicide is required. Moreover, the autonomous systems for the rice cultivation is needed for manyfanners because of their works are very hard. Therefore, we propose a "Weeding Robot" which can travel all over of the paddy field for the rice cultivation. Furthermore, we design that the operation of the proposed robot is simple without difficult operation and is not complicated set up. In this paper, we describe the development status for each element of the proposed weeding robot. Thereafter, we report about the autonomous experiments.

View full abstract

This paper proposes the new design of thermal cutting system for 5 degree of freedom robotic arm, designed to harvest sweet peppers in horticultural green house. The design consists of two parallel gripper bars mounted on a frame connected by specially designed notch plate and operated by servo motor. The special electric device was developed to obtain the high frequencies with high voltage to perform the cutting operation. The output from the device was connected to the electric rods mounted at the front of the specially designed notch plate. At higher frequencies, the rods generate thermal arc which helps to cut the stem of the sweet pepper. The recognition system assists to locate the fruit.

View full abstract

To accurately estimate the beef marbling standard (BMS) number of live cattles using ultrasound echo imagings, we have developed a image recognition method by use of a neural network. This paper examines the efficiency of applying independent component analysis (ICA) to the compression of multidimensional image features extracted from imagings. ICA can accurately separate a target signal because of its independence assumption, while principal component analysis (PCA), a conventional method, involves decorrelation of the components. We have implemented the estimation tests by use of ultrasound echo imagings of 103 live cattles. Multidimentional texture features extracted from the imagings were compressed by ICA, and then the estimation of BMS number was conducted by using a neural network. The results confirmed that the estimation accuracy of BMS number by ICA was higher than that by PCA.

View full abstract

The automated harvesting robot has been developed ranging from grain and root vegetables. However, the robot that target soft vegetables has not been developed yet. The reason is very difficult to collect without any damages the leaves and stems of vegetables soft. The structure of the collection and transport unit of the spinach harvesting robot is designed in this paper. Mechanism that applies the oscillating crank mechanism is used in the collection unit. The design result and experimental results using the unit are shown in this paper.

View full abstract

This paper proposes height sensor for the ground level position detection and a control system. The system is used in the automatic harvest robot of spinach which is under developing by our group. The proposal sensor can measure the average displacement of the ground surface of the field. By the field experiment using the control system, the effectiveness of the control system could be verified.

View full abstract

This paper is concerned with modeling and control experiment of a casterboard, which is a variant of skateboard having roll-twistable footplates and a pair of inclined casters. In this paper, we develop an autonomous 'robotic casterboard' in which the footplates are actively driven by servo motors, and show that it can be arbitrarily steered by appropriate choices of oscillatory control inputs.

View full abstract

In this paper, we propose the method of route and sensor's priority setting by environment information(e.g. many buildings, many trees, a clear or unobstructed space) from the internet map read by human for outdoor autonomous mobile robot. In many cases, making autonomous movement perform to a robot, in advance the data of the course needed to be collected at fieldworks. Thereby, the range which a robot can move autonomously was limited by collected data at fieldworks. Therefore, in the future, by acquiring environment information from the internet map information, we can set the route for the robot without prior fieldwork.

View full abstract

We have organized "Tsukuba Challenge 2011" between July and November in Tsukuba city. This year was the final year of five years planning of the Tsukuba Challenge organized since 2007. "Tsukuba Challenge" is The Real World Robot Challenge which aims to realize actual intelligent robots which can work in the real daily life environment. As the subject of Tsukuba Challenge 2011, the robots are required to run autonomously and safely for 1.4 km of a promenade street, a park and a public plaza in as well as the inside of the building in Tsukuba city center.

View full abstract

In Tsukuba Challenge 2011, the elevator and slope were adopted as a new challenge environment. This paper describes implemented behavior of a mobile robot getting on and off the elevator by the team OKUGAIGUMI at Intelligent Robot Laboratory, Univ. of Tsukuba. SOKUIKI senser are utilized for the behavior. Judgment of opening and closing door, path planning, the experiment and result in the Challenge are discribes in the paper.

View full abstract

This paper addresses the image-based visual servoing (IBVS) problem for a nonholonomic wheeled mobile robot with an active camera. We propose an IBVS scheme using a virtual barycenter of all feature points. The wheeled mobile robot is considered as a kind of nonholonomic mobile manipulator. De Luca, et al. have proposed an IBVS scheme for mobile manipulators based on the task priority method. However, De Luca's scheme dose not enough the original control objective in some cases. In this paper, we introduces the virtual barycentric point to perform IBVS for it in the primary task and IBVS for each real feature point in the secondary task.

View full abstract

The objective of this study is to develop a robust human tracking method for mobile robot using only one laser range finder (LRF). When a mobile robot runs along a human, it is always necessary to observe positions of the human. Therefore, the LRF must be mounted on a position easy to recognize human. Furthermore, the LRF may lose the position of the target human, and another human might enter between the robot and the target human. Therefore, this paper discusses the mounting position of the LRF, as well as a robust tracking method using the LRF.

View full abstract

Recently, plant factories and agricultural robots have been studied and developed. In these cultivation systems, however, it has been difficult to manage the state of each individual plant. We propose a cultivation system using the swarm of plant pot robots which move to an optimal environment based on sensory information of each plant and the surrounding environment. In this paper, we propose a control method of the swarm of plant pot robots using artificial potential fields for temporally and spatially effective utilization of sunlight, and show its effectiveness by simulation.

View full abstract

This paper proposes a new weeding device which enables to weed around furrows. First, we make a new version of this machine to reduce size and weight of the previous one so that it can be used in actual fields. Next, a noncontact mechanism which enables to weed around furrows is proposed. We design a controller of the angle of the end effector which carries the proposed mechanism. Finally, in order to show the validity of the system, we experiment with it in a pseudo furrow. From the result, the possibility of this system for noncontact weeding to furrows is verified.

View full abstract

This paper proposes a novel vegetation classification method using an airborne imagery obtained by the small unmanned aerial vehicle (UAV). A conventional method used in the vegetation classification uses the value of reflected intensity in airborne images obtained by airplanes and satellites. However, it is difficult to classify the type of vegetations because of the restriction of the image resolution. The small UAV can fly at low attitude, it can obtain a high resolution airborne imagery. The proposed method aims to make a vegetation classification map using the texture analysis and Adaboost machinery learning. We use the texture, color, and NDVI (Normalized Difference Vegetation Index) calculated from airborne imagery as the classification features. From the result of classification test, we concluded that proposed method is effective to classify the type of vegetation using a small UAV.

View full abstract