The paper describes practical design and use of an Outdoor Cleaning Robot. We developed an autonomous Out-door Cleaning Robot, which is equipped with a gyroscope sensor, a triangular surveying sensor, and a communication device. The robot uses the gyroscope sensor and odometery to calculate its current position. The triangular surveying sensor, together with reflection panels strategically placed in the environment, is used for calibrating and improving the accuracy of the calculated position. The communication device is used for multiple cleaning robots to know each other's status. The paper also reports experiments in which the developed robots were put into operation for six months in the outdoor environment, and satisfactory results were obtained.
For humanoid robots which act in daily life environments, integration of vision, force and audio sensors are important to achieve various tasks. In this paper, we focus on trash separation behavior for bottles and cans, which is one of hard problems because sensor patterns would be widely changed. In such situation, not only using passive sensors but also acquisition of advanced sensor information by active behavior such as grasping bottle, watching in close distance and listening to sound with hitting. In this paper, we propose integration method of these sensors for realization of trash separation task.
We believe that humanoids will take an active part in our daily lives in the near future as important media, since a human-size humanoid enables people to more easily accept a sense of humanlike expressions or emotions, wishes, and so forth in addition to capabilities of other media, such as the ability to collect and provide information from the Internet and via ubiquitous sensors connected through a network. Once this time arrives, humanoids will be expected to have the social skills necessary for interacting with people in addition to the ability to carry out their own tasks. We call an interaction that increases familiarity and makes communication smoother a “social interaction.” We have recently developed a human-size humanoid, called “Robovie-IV.” Robovie-IV features interaction abilities such as the ability to chat vocally, which are intended to be “social interactions.” Using Robovie-IV we have conducted an experiment in which it interacts with people in an everyday office environment. This paper discusses the design requirements of Robovie-IV and introduces an overview of its hardware and software architectures. Then, the experimental results, discussions, and conclusions are presented.
This paper presents a humanoid walking control system that generates and updates the motion pattern frequently, with a cycle time on the order of 20 [ms] . Frequently updating the motion pattern enables a robot to quickly respond to changes in the commanded walking direction and upper body posture with minimal delay. Our experiments also show that short cycle times are useful for building a walking system that provides motion status information and sensor feedback to the motion generation layer in order to maintain consistency of the overall dynamics. Using preview control theory, we generate horizontal torso trajectories that can realize desired ZMP trajectories. Trajectories for the free leg are generated using a method that allows for changes to the landing position while taking a step, and adjusting the foot placement to avoid large ZMP errors which may result from sudden changes in the desired motion. We present an implementation of the proposed method on the full-size humanoid HRP-2 and show experimental results designed to validate the overall functionality.
This paper describes a passive dynamic model of a passenger for a biped walking vehicle. The model consists of a lower-limb's model fixed on the robot's waist and a particle model of an upper body connected to the robot's waist through springs and dampers. The stiffness and damping coefficients are verified through waist-shaking experiments by using a force/torque sensor under the seat that is mounted on the robot's waist. A walking pattern generation that enables stable walking even if a passenger sits naturally on the seat is also described. This stable walking pattern is generated by an iteration method based on Fourier Transformation. The effectiveness of the proposed method is confirmed through walking experiments.
Many researchers have been studying on walking control methods for biped robots. However, the effectiveness of these control methods was not verified in outdoor environments such as pedestrian roads and gravel roads. In this paper, a landing pattern modification method adaptable to uneven terrain in a real environment is proposed which is based on a predictive attitude compensation control and a nonlinear compliance control. This method does not require any other sensors except force sensors. Also, a new biped foot system is described, which can form larger support polygons on uneven terrain than conventional biped foot systems. Using the pattern modification method and the foot system, WL-16R11 (Waseda Leg - No. 16 Refined II) achieved a stable walk on bumpy terrain with 20 mm height and 10 degrees inclination. Furthermore, a stable dynamic walk was realized in outdoor environment, when a human rode it. Through various walking experiments, the effectiveness of the method is confirmed.
In the present paper, we introduce a new method by which to generate motions having flight phases such as jumping or running. Such motions are expected to expand the speed and stability of legged robots. Motions with flight phases are examined with respect to jumping, stability and implementation. A novel method of determining the Center of Mass (COM) and foot locations is proposed that is independent of the target robot actuators or assignments of joints. Moreover, the proposed method allows an on-line system to be built easily and is useful for actual robots. A one-legged jumping robot having a toe joint is built in order to demonstrate the feasibility of this method. Using the toe joint of the robot, the proposed method can reduce the angular velocities of the joints remarkably. Experiments are conducted to investigate the control of the direction, velocity and turning of the robot.
MR fluid (Magneto-Rheological fluid) is a kind of functional fluid. In this study, we developed Shear-type Compact MR Brake (SCMR-Brake) applying shear mode of MR brake. To employ multi-layered and minute-gap structure, we developed SCMR-Brakes. By using this brake on the part of ankle torque control unit, we developed Controllable Ankle-Foot nrthosis.
This paper proposes a control method for a human-assisting manipulator using acceleration sensors, which consists of an arm control part and a hand and wrist control part. The arm control part controls manipulator's shoulder and elbow joints using acceleration signals, while the hand and wrist control part controls the corresponding joints using mechanomyogram (MMG) signals measured from a human operator. A distinctive feature of our method is to estimate force and motion information from the measured acceleration signals using MMG signal processing and the probabilistic neural network. It is shown from experiments that the MMG patterns during hand and wrist motions can be classified sufficiently and that the prosthetic manipulator can be controlled using the measured acceleration signals. It may be useful as an assistive device for the physically disabled.
This paper describes an experimental study of power assist systems based on harmonization of human operators, actuators and mechanical elastic elements. This power assist system proposed by the authors simultaneously realizes amplification of operator's torque and optimization of stiffness of a mechanical elastic element. The optimization in the proposed power assist system means that the minimum torque from the actuator can generate periodical motions even if the parameters of the operator's periodic torque are not known. The amplification and the optimization were mathematically proven in our previous papers. In this paper, effectiveness of the proposed power assist system is experimentally investigated by using a mechanical device with a stiffness adjustable mechanism. Next, another type power assist system of a mechanical structure with better backdrivability is proposed and effectiveness of the new structure is also verified through simulation and experimental results.
Providing parents with information related to child's injury has been considered to be a useful method for preventing child's injuries for many years. However, statistics on child's injuries shows that conventional methods failed to reduce injuries. One cause lies in the lack of feedback loop for personalizing information. In this paper, the authors describe“an information circulation system, ”which collects information on parents' cognition on child's injuries through offering personalized information in parallel. This paper describes a new service which offers injury scene videos on the web for injury prevention as a concrete example of the information circulation system. The feasibility of the new service is confirmed by conducting experiments on web for over 13 months in cooperation with a company. This paper also proposes a new method for personalizing information by selecting optimal videos on the web content service based on feature values describing injuries. The effectiveness of the proposed method is proved by questionnaire survy for 20 parents.
When touching an object with a tool held by our hand, we tactually feel the object as if touching with our bare hand. We address what comprises this cognitive ability of “tool-body assimilation.” We compose a computational model of the tool-body assimilation and tool-use backed by physiology and robotics. Two processes of identification of the tool during a swing play a crucial role in the model. Exploiting the model, a robot on simulation can use first-appeared tools of various shapes to retrieve a target object. It synthetically instantiates how computation works in human's brain during tool-body assimilation. It also exemplifies a scheme realizing autonomous tool using robots.
In this paper, we propose a robust online action recognition method based on boosted sequential classification. Our method utilizes boosting algorithm that is one of ensemble learning algorithms. This algorithm is also known as a feautre selector and has been utilized in the fields of image processing and natural language processing in recent years. Based on the boosting scheme, our method can automatically and efficiently select significant features for action recognition. Additionally, the method leverages temporal dependency of actions based on Ising model to improve recognition performance. We evaluated our method to action recognition, such as walking and running, using motion capture data only with posture features. In the result, our method can classify the actions more robustly than the method that does not utilize temporal dependency of actions.
In the recent years, neural networks or other learing networks are frequently used in the field of robotics. However, the needed conditions of the learning system are not fulfilled enough in autonomous robot, because the variety of the needed conditions let it difficult to accomplish. So, integration of the functions is inevitable to create an effective learning system in autonomous robot. In traditional methods, it was difficult to accomplish “autonomous exploration of the effective output”, “simple external parameters”, and “low calculation cost” together in a learning system. Thus, we proposed a new learning method self-organizing network elements (SONE) against this problem. All of these conditions are fulfilled by SONE, however there is a need to enhance the ability against noises. Therefore, we propose a technique to restrain noises in SONE. In our experiments, more resistance against noises was confirmed with this technique. Also in a robot simulation, the performance of the robot was improved by this novel method.
This paper proposes an iterative learning control (ILC) method for robots with redundant joints to acquire the desired control input signals that produce an endpoint trajectory specified in task space. The learning update law of control input signals is constructed only in task space by modifying the previous control input through adding linearly endpoint position and velocity trajectory errors. Although the dimension of the task space is strictly less than the DOF (Degree-of-freedom) of robots, the proposed method need neither consider any inverse kinematics problem nor introduce any cost function to be optimized and determine the inverse kinematics uniquely. Convergence of trajectory trackings to the specified one is shown by numerical simulations in both cases (1) free-endpoint motion and (2) constraint-endpoint motion with specified contact force. A theoretical proof of convergences is also given on the basis of a dynamics linearized around a joint trajectory when the endpoint tracks a desired trajectory.
This paper discusses the classification and enumeration of topological structures of robotic mechanisms, aiming to develop the high performance robotic mechanisms with unconventional topological structures. Fist, features of topological structures of conventional robotic mechanisms are considered. From the consideration, it is pointed out that almost all of the topological structures can be constructed by using only one simple basic unit and two simple connection rules. Then, the efficient way to enumerate the topological structures which cannot be constructed from the basic unit and the connection rules, or unconventional topological structures, is shown. After that, the sequence to construct robotic mechanisms from the enumerated unconventional topological structures is discussed. In the discussion, some effective tools, which support designers to construct variety of mechanisms from the same topological structure are introduced. Finally, two examples of the sequences to generate robotic mechanisms with unconventional topological structures, 3DOF planar mechanism and 6DOF spatial mechanism, are shown.
An electro-conjugate fluid (ECF) is a kind of functional fluid which generates a powerful jet flow when subjected to high DC voltage. Using this fluid, this paper proposes a novel earthworm-type peristaltic micromachine. First we describe a motion pattern of earthworm-type peristaltic micromachine, and propose a micro cell using ECF. The cell, composed of a fiber-reinforced silicone rubber tube, contracts in axial direction and expands in radial direction simultaneously by using the jet pressure of ECF. A large model prototype of earthworm-type peristaltic microma-chine having four cells in series, 10.2 [mm] in diameter and 90 [mm] in length, is developed and its basic driving characteristics are confirmed by experiments.
Lift force and specification requirements for a dragonfly-scale flapping flight robot are quantitatively investigated by using fluid-structure interaction analysis. For the robotics application of flapping flight, the specifications of actuators and energy sources to generate a certain lift force are essential to design a realistic robot. However, these parameters have been poorly studied because of the difficulty of small-scale unsteady aerodynamics interacting with flapping wings. In the present paper, we show the quantitative evaluation of these parameters in a design of realistic 2DOF actuator by using a fluid-structure interaction analysis based on arbitrary Lagrangian-Eulerian finite element method.
This paper describes dynamics of object manipulation performed by a pair of soft fingertips. Soft fingertips en able secure grasping and stable manipulation yet the mechanics of soft-fingered manipulation has not revealed well. Based on the observations of soft-fingered grasping and manipulation, we found that grasping and posture control of a rigid object can be performed by a pair of 1-DOF fingers with soft tips, which contradicts to previous theories on soft-fingered manipulation. In this paper, we propose a parallel distributed model with tangential deformation of hemispherical soft fingertips to formulate the grasping and posture control by a pair of soft fingertips. The proposed model reflects the structure of a human finger consisting of a soft fingertip with a hard fingernail on its reverse side. We then formulate the dynamics of grasping and manipulation performed by a pair of soft fingertips. Experimental results validate the proposed model and formulated dyanamics.
Recent advances of robot technology raise increasing demands for tactile distribution sensor for safe and natural human-machine interaction. For a richer variety of tactile interaction, and for realizing a full surface coverage of various structure robots, not only flexibility, but also stretchability is very important. However, stretchability has been an impossible characteristic due to circuit wiring networks embedded in the sensors. This paper presents a tactile distribution sensor based on inverse problem theory. It solves the wiring problem and realizes a novel stretchable tactile distribution sensor, with potentially a broad range of functionality and applications. EIT, an inverse problem theory, can estimate the resistivity distribution of the inner space of a conductor, by measuring only from the boundary of the conductor. By applying EIT to a force sensitive resistive rubber sheet, a flexible, stretchable and thin tactile distribution sensor is realized with no embedded wiring within the sensing area. To verify the basic characteristics and potential of the EIT-based sensor, experimental results on pressure distribution and stretch dis-tribution sensors are reported. Morever, we propose a porous stretchable sensor with novel tactile sensing capability; pinching, rubbing and pushing. Proposals on possible augmentations of the sensor are also presented, including a construction technique for various sensitive conductors, a multi-layered sensor for multi-stimulus perception.
To achieve a human like grasping by the multi-fingered robot hand, grasping force should be controlled without information of the grasping object such as the weight and the friction coefficient. In this study, we propose a method for detecting the slip of grasping object by force output of the Center of Pressure (CoP) tactile sensor. CoP sensor can measure center position of distributed load and total load which is applied on the surface of the sensor within 1 [ms] . This sensor is arranged on finger of the robot hand, and the effectiveness as slip detecting sensor is confirmed by experiment of slip detection on grasping. Finally, we propose a method for controlling grasping force resist the tangential force added to the grasping object by feedback control system of the CoP sensor force output.
Real-time and robust sound source tracking is an important function for a robot operating in a daily environment, because the robot should recognize where a sound event such as speech, music and other environmental sounds originates from. This paper addresses real-time sound source tracking by spatial integration of an in-room microphone array (IRMA) and a robot-embedded microphone array (REMA) . The IRMA system consists of 64 ch microphones attached to the walls. It localizes multiple sound sources based on weighted delay-and-sum beamforming on a 2D plane. The REMA system localizes multiple sound sources in azimuth using eight microphones attached to a robot's head on a rotational table. A particle filter integrates their localization results to track multiple sound sources. The experimental results show that particle filter based integration improved accuracy and robustness of sound source tracking even when the robot's head was in rotation.
We built an acoustical telepresence robot, called TeleHead, which has a user-like dummy head (i.e., the shape of the dummy head and that of the user is very alike) and whose movement is synchronized with the user's head movement in real time. We are trying to clarify the effects of reproducing the user's head movement. In this article, we evaluated the sense of incongruity caused by the delay time of reproducing head movement by means a psychophysical approach. Discrimination tasks for head-movement delay clarified the users' perceptual thresholds against the dead time. The results indicate that head-movement control should have a dead time shorter than 28 [ms] . In addition, this dead time does not depend on the characteristics of an acoustical telepresence robot.
This paper presents a method of recognizing objects and estimating their 3-D poses from a monocular image. This method integrates image edge points and a 3-D edge model into an object model. The method retrieves candidate objects from an object database using image edge points with SIFT feature vectors. Then, the method estimates the 3-D pose of each candidate object by minimizing the re-projection errors of the 3-D edge model. Experimental results show that the method successfully recognized non-textured objects and complex-shaped objects in real environments.
This paper describes a real-time three-dimensional sensing system for applications based on feedback automation. The proposed system enables to observe a moving/deforming object at high frame rate and can acquire data in realtime. These are provided by three distinctive features, three-dimensional sensing by a single frame, high-frame-rate imaging and high-speed image processing. We also present some results of evaluation experiments. The experimental results show the advantages of our system compared with conventional approaches. Our system is expected to achieve improvements in a wide range of three-dimensional-sensing applications.
In general, the sampling period of vision sensor is longer than that of joint servo in visual servo systems. To compensate the time delay of vision sensor, this paper applies multirate control method to visual servo systems of 6 DOF robot manipulator. The proposed method applies perfect tracking control (PTC) method with signal generator in order to follow the moving object with high-order periodic motion. With this method, computation cost is reduced compared with a conventional observer based method.