Showing papers by "Keigo Watanabe published in 2002"

An ART-based fuzzy controller for the adaptive navigation of a quadruped robot

Abstract: An adaptive-resonance theory (ART)-based fuzzy controller is presented for the adaptive navigation of a quadruped robot in cluttered environments, by incorporating the capability of ART in stable category recognition into fuzzy-logic control for selecting the adequate rule base. The environment category and navigation mechanism are first described for the quadruped robot. The ART-based fuzzy controller, including an ART-based environment recognizer, a comparer, combined rule bases, and a fuzzy inferring mechanism, is then introduced for the purpose of the adaptive navigation of the quadruped robot. Unlike classical/conventional adaptive-fuzzy controllers, the present adaptive-control scheme is implemented by the adaptive selection of fuzzy-rule base in response to changes of the robot environment, which can be categorized and recognized by the proposed environment recognizer. The results of simulation and experiment show that the adaptive-fuzzy controller is effective.

Fuzzy-chaos hybrid controller for controlling of nonlinear systems

Abstract: A novel concept for controlling of nonlinear systems using chaos and fuzzy model-based regulators is presented. In the control of such systems, we employ two phases, the first of which uses open-loop control forming a chaotic attractor or using chaotic inherent features in a system itself. Once the system states reach a predefined convex domain, open-loop control is cut off and a fuzzy model-based controller is employed under state feedback control in the second phase. The relaxed stability conditions and linear matrix inequalities (LMIs)-based design for a fuzzy regulator is introduced to construct a fuzzy attractive domain, in which a global solution is obtained so as to achieve the desired stability condition of the closed-loop system. The proposed controller architecture has been tested using three nonlinear systems: the Henon map, the Lorenz attractor, and a two-link manipulator. The simulation results show the effectiveness of the proposed controller.

Path Tracking Based on Closed-Loop Control for a Quadruped Robot in a Cluttered Environment

Abstract: In this paper, the path tracking control is studied for a quadruped robot, named TITAN-VIII, walking in a cluttered environment. A simple and efficient algorithm of path planning is proposed, which is characterized by finding turning-point in the walking environment for the robot. The generalized gait algorithm based on the static stability is presented for the continuous and omnidirectional crawl of the robot. Especially, the real-time robot localization in the walking environment, which is the key to the settlement of the path tracking control, is realized by dead-reckoning for the quadruped robot. Based on the above work, we design the closed-loop control architecture so that the robot is able to track the desired path in an obstacle-strewn environment. The reliability and effectiveness of the proposed method is demonstrated through the experimental results.

Application of Multiple Fuzzy-Neuro Controllers of an Exoskeletal Robot for Human Elbow Motion Support

Abstract: A decrease in the birthrate and aging are progressing in Japan and several countries. In that society, it is important that physically weak persons such as elderly persons are able to take care of themselves. We have been developing exoskeletal robots for human (especially for physically weak persons) motion support. In this study, the controller controls the angular position and imped- ance of the exoskeltal robot system using multiple fuzzy-neuro controllers based on biological signals that reflect the human subject's intention. Skin surface electromyogram (EMG) signals and the generated wrist force by the human subject during the elbow motion have been used as input information of the controller. Since the activation level of working muscles tends to vary in accordance with the flexion angle of elbow, multiple fuzzy -neuro controllers are applied in the proposed method. The multiple fuzzy-neuro controllers are moderately switched in accordance with the elbow flexion angle. Because of the adaptation ability of the fuzzy-neuro controllers, the exoskeletal robot is flexible enough to d eal with biological signal such as EMG. The experimental results show the effectiveness of the proposed controller.

Obstacle Avoidance of Three-DOF Underactuated Manipulator by Using Switching Computed Torque Method

Abstract: Obstacle avoidance of underactuated robot manipulators using switching computed torque method (SCTM) is presented One fundamental feature of this novel method is to use partly stable controllers (PSCs) in order to fulfill the ultimate control objective Here, we use genetic algorithms (GAs) to acquire the optimum switching sequence of the control actions for a given time frame with the available set of elemental controllers, depending on which links/variables are controlled The effectiveness of the concept is illustrated by taking a three-degrees-of-freedom (DOF) manipulator and showing enhanced performance of the proposed control methodology

Energy-optimal gait analysis of quadruped robots

Abstract: It is important for walking robots such as quadruped robots to have an efficient gait Since animals and insects are the basic models for most walking robots, their walking patterns are good examples In this study, the walking energy consumption of a quadruped robot is analyzed and compared with natural animal gaits Genetic algorithms have been applied to obtain the energy-optimal gait when the quadruped robot is walking with a set velocity In this method, an individual in a population represents the walking pattern of the quadruped robot The gait (individual) which consumes the least energy is considered to be the best gait (individual) in this study The energy-optimal gait is analyzed at several walking velocities, since the amount of walking energy consumption changes if the walking velocity of the robot is changed The results of this study can be used to decide what type of gait should be generated for a quadruped robot as its walking velocity changes

Translational Crawl and Path Tracking of a Quadruped Robot

Abstract: Translational crawl and path tracking are presented for a quadruped robot, named TITAN-VIII, to walk on rough ground. The generalized and explicit formulation is derived to generate the translational crawl gait in an arbitrary direction automatically, to control the joint positions, and to estimate the robot localization in a walking environment. Compared to conventional gaits, the proposed gait is characterized by a natural and continuous transition between any successive gait cycles, by a maximized stride of the robot in each gait cycle, and by different foot trajectories corresponding to the uneven terrain. Especially, the proposed approach enables the quadruped robot to track a reference path in a complex walking environment, based on dead-reckoning localization for the robot. The effectiveness of the proposed method is demonstrated through the experimental results. © 2002 Wiley Periodicals, Inc.

Control of nonholonomic mobile robot by an adaptive actor-critic method with simulated experience based value-functions

Abstract: An adaptive actor-critic algorithm is proposed under the assumption that a predictive model is available and only the measurement at time k is used to update the learning algorithms. Two value-functions are realized as a pure static mapping, according to the fact that they can be reduced to nonlinear current estimators, which can be easily constructed by using any artificial neural networks (NNs) with sigmoidal function or radial basis function (RBF), if all the inputs to the present value-functions are based on simulated experiences generated from the predictive model. In addition, if a predictive model is assumed to be used to construct a model-based actor (MBA) in the framework of adaptive actor-critic approach, then this type of MBA can be viewed as a network whose connection weights are composed of the elements of feedback gain matrix, so that the temporal difference (TD) learning can also be naturally applied to update the weights of the actor. Since the present method can update the learning by using only one measurement at time k, a relatively fast learning is expected, compared with the previous approach that needs two measurements at times k and k + 1 to update the actor-critic networks. The effectiveness of the proposed approach is illustrated by simulating a trajectory-tracking control problem for a nonholonomic mobile robot.

Design of an exoskeletal robot for human shoulder motion support considering a center of rotation of the shoulder joint

Abstract: The exoskeletal robots are expected to support the daily activities of physically weak persons such as elderly persons or handicapped persons in these days. In our previous research, the exoskeletal robots for elbow or shoulder motion support have been developed since the elbow and shoulder motion is especially important for people to perform tasks in everyday life. In this paper, we present a mechanism of the moving center of rotation of the shoulder joint of the exoskeletal robot for human shoulder motion support in order to fit the center of rotation of the robot shoulder joint to that of the physiological human shoulder joint during the shoulder motion. The effectiveness of the proposed mechanism of the robot has been evaluated by experiment.

A simple, low-cost non-dispersive infrared CO/sub 2/ monitor

Abstract: A non-dispersive infrared (NDIR) CO/sub 2/ monitor is developed for air quality assessment, The NDIR CO/sub 2/ sensor consists of one IR source, a pipe in which IR beam interacts with CO/sub 2/ molecules in the air, and a beam splitter terminated by two thermopiles with optical band-pass windows. One thermopile measures the spectrum absorption due to CO/sub 2/ molecules, while the other monitors the light emission level of the IR source. The signal processing electronics take the ratio between the outputs of the two detectors to make the CO/sub 2/ concentration measurement insensitive to the aging of the IR source and temperature. The ratio is a nonlinear function of CO/sub 2/ concentration and the cubic spline function is used for the linearization and the calibration. A prototype monitor demonstrates performances quite satisfactory for practical applications.

Control for a rings gymnastic robot using fuzzy reasoning and genetic algorithms

Abstract: The rings event is part of men's apparatus gymnastics. The ring exercise have free-floating characteristics of the gripping point. Here, we propose a novel “rings gymnastic robot” aimed at an application in gymnastic coaching. The purpose of this article is to develop fuzzy control rules to realize performances on the rings by considering torque minimization by genetic algorithms (GAs), because these rules should be useful in coaching. The effectiveness of the controllers obtained is illustrated by a simulation.

Adaptive output tracking of partly known robotic systems using SoftMax function networks

Abstract: In this paper, a neural-network-based adaptive control scheme is presented to solve the output-tracking problem of a robotic system with unknown nonlinearities. The control scheme ingeniously combines the conventional resolved velocity control technique and a neurally-inspired adaptive compensating paradigm constructed using SoftMax function networks and neural gas algorithm. Results of simulations on our active binocular head are reported. The neural network model constructed to has two neural subnets to separately control the robot head neck and eye movement, simplifying the design and leading to faster weight tuning algorithms.

Performance test of a force controlled robot sander using a surface following controller based on cutter location data

Abstract: This paper describes two indispensable techniques to develop a flexible sanding robot. One is the impedance model following force control for industrial robots which can delicately control the contact force acting between the sanding tool and the workpiece. The other is the surface following controller using the cutter location data to suitably keep the position and orientation of the sanding tool without complicated teaching process. The techniques are applied to an open architectural industrial robot FS-20 with an air-driven double action sander, so that a useful robot sander can be easily realized. Some sanding experiments have demonstrated that the proposed robot sander can successfully accomplish the sanding tasks of wooden workpieces with curved surface.

Acquiring performance skill of backward giant circle by a rings gymnastic robot

Abstract: We (2001) have proposed the "rings gymnastic robot" aiming at an application to the gymnastic coaching by understanding the ring exercises through the robot. The ring exercises have the characteristics that the apparatus can move in all directions freely unlike the other gymnastic events. Fuzzy control is adopted as a control method for the rings gymnastic robot. As the joint torque in performance is useful for the coaching and gymnasts can easily understand the skill to realize performance if the control method is represented as "if-then" rules. In this paper, we derive a three-dimensional model of the robot and acquire the performance skill of a backward giant circle using the genetic algorithm. Three-dimensional graphics sequence based on simulation results shows that the obtained skill is effective.

Enhancing the Autonomy of Teleoperated Redundant Manipulators Through Fusion of Intelligent Control Modules

Abstract: This paper presents a method for redundancy resolution of an industrial manipulator in a teleoperated force control task A seven degree-of-freedom (DOF) industrial manipulator manufactured by the Mit-subishi Heavy Industries Ltd. is used for experiments. The task involves obeying a force command sent from a remote computer while autonomously adapting the posture to avoid unexpected obstacles moving toward the manipulator. Redundancy resolution is employed for autonomous adaptation of the configuration to avoid the obstacle while continuing the force control task This self-adaptive skill on the slave manipulator side is very important because teleoperation is often performed in dangerous or partially unknown environments where unexpected changes such as moving obstacles can well be expected. In such situations, the control ability of the master side is very limited due to the practical limitations of vision sensors to capture a comprehensive view of the environment and the limitations of the degrees of freedom on the master ma-nipulator. The proposed method relies on two modules of an intelligent controller on the slave side. The first is an on-line fuzzy neural network (FNN) for intelligent force control, and the second is a configuration controller that works in harmony with the first to exploit redundancy to react to avoid moving obstacles such that the latter does not inhibit the progress of the former. The second controller generates joint velocity commands in null space of the hand Jacobian, so that its activation does not affect the force controller. Here we show that the proposed method can skillfully avoid a moving obstacle without stopping the force control task This skillful adaptation ability can significantly improve the efficiency and safety of teleoperated force control tasks with less burden on the master side. This paper presents some promising experimental results to demonstrate the effectiveness of the proposed method.

Dynamic Control for a Holonomic and Omnidirectional Mobile Robot with Active Dual-Wheel Caster Assemblies

Abstract: Recently, omnidirectional mobile robots with holonomic property have been studied actively. We have already developed a holonomic and omnidirectional mobile robot which consists of two active dual-wheel caster assemblies and also derived the kinematic models for the assembly and the mobile robot. In this paper, we further derive a dynamic model for the mobile robot by using the forces acting on the steering axis and construct a resolved acceleration control system based on the model to realize a dynamic control for the mobile robot. An experiment using a prototype robot is performed to prove the validity of the model and the effectiveness of the control system.

Fuzzy rules extraction by GA for controlling 2-DOF underactuated planar manipulators: switching computed torque approach

Abstract: We propose a method of fuzzy rule extraction for switching partly stable controllers (PSCs) for online operations of underactuated manipulators The fuzzy rule base of switching the PSCs is optimized using GA and the optimization is performed off-line Design parameters of the fuzzy rules are encoded into chromosomes and shapes of the Gaussian functions are evolved to minimize the angular position errors The angular position errors are used as the inputs to the Gaussian membership functions in the antecedent part and the one index of the PSCs is assigned in the consequent part of the fuzzy reasoning Then, this trained rule base can be brought into the online operation of the underactuated manipulator The effectiveness of the concept is illustrated by taking a 2-DOF RR planar manipulator and results show the effectiveness of the proposed control methodology

A study on rings gymnastic robot: (Fuzzy control rules for realizing exercises)

Abstract: The rings event is one of men's apparatus gymnastics, and has free-floating characteristics at the gripping point that do not exist in the other gymnastic events. In this paper, we propose a rings gymnastic robot intended to apply it to the coaching. Joint torque in exercises is informative for the coaching and gymnasts can easily obtain the knowledge to realize exercises if a control law of the torque is represented as "if-then" rules. Therefore, we adopt fuzzy control as a control strategy for the rings gymnastic robot. The purpose of this paper is to acquire suitable fuzzy rules for realizing a series of exercises, which is a handstand from backward giant circle. Simulation results show the effectiveness of the fuzzy rules and a guideline for the coaching can be read from the fuzzy rules.

Intelligent interface and control of an exoskeletal robot for human shoulder motion support considering subject's arm posture

Abstract: We have been developing exoskeletal robots in order to support the motion of physically weak persons such as elderly persons or handicapped persons. In our previous research, the 2-DOF exoskeletal robots for shoulder motion support and its control method have been developed since the shoulder motion is especially important for people to take care of themselves in everyday life. In this paper, we propose intelligent interface, which realizes the fuzzy-neuro controller adjustment in accordance with the human subject's arm posture, in order to effectively control the 2-DOF exoskeletal robot for shoulder motion support. The intelligent interface is realized by applying a neural network. The effectiveness of the proposed intelligent interface of the exoskeletal robot has been evaluated by experiment.

Intelligent interface of an exoskeletal robot for human elbow motion support considering subject's arm posture

Abstract: We develop exoskeletal robots for the motion support of physically weak persons. In this study, the human elbow motion is assisted by the exoskeletal robot system, since the elbow motion is one of the simplest and the most important motion of human beings in everyday life activities. The angular position and impedance of the exoskeletal robot system are controlled by multiple fuzzy-neuro controllers. The skin surface electromyogram (EMG) signals and the generated wrist force by the human subject during the elbow motion are used as input information of the controller. Since the activation level of working muscles tends to vary with the human subject's arm posture, an intelligent interface that cancels out the effect of posture changes of the human subject's arm is proposed in this paper. The experimental results show the effectiveness of the proposed intelligent interface.

A study of human motion support with an exoskeletal robot (2nd report, control of an exoskeletal robot for human elbow motion support using multiple fuzzy-neuro controllers)

Abstract: We have been developing exoskeletal robots for human motion support. The exoskeletal robots enable physically weak people such as elderly people to take care of themselves. In this paper, we introduce an effective control method of a 1DOF exoskeletal robot for human elbow motion support. Multiple fuzzy-neuro controllers have been applied to control both the angular position and impedance of the exoskeltal robot system in any elbow angle based on force signals and biological signals that reflect the human subject's intention. Skin surface electromyogram (EMG) signals and the generated wrist force by the human subject during the elbow motion have been used as input information of the controller. The proposed system is able to adapt itself to the physiological condition of any human subject because of the adaptation ability of the fuzzy-neuro controllers. The experimental results show the effectiveness of the proposed control method of the exoskeletal robot.

Reinforcement Learning with Expectation and Action Augmented States in Partially Observable Environment

Abstract: The problem of developing gwd or optimal policies for partially observable Markov decision processes (POMDP) remains one of the most alluring areas of research in artificial intelligence. Encourage by the way how we (humans) form expectations from past experiences and how our decisions and behaviour are affected with our expectations, this paper pmposes a method called erpeetation and action augmented stares (EAA.9 in reinforcement learning aimed to discover gwd or near optimal policies in partially observable environment. The method uses the concept of expectation to give distinction between aliased states. It works by augmenting the agent’s observation with its expectation of that observation. Two problems from the litera” were used to test the proposed mahod. The results show promising characteristics of the method as compared to some methods currently being used in this domain.

Coordinated Transportation of a Single Object by a Group of Nonholonomic Mobile Robots

Abstract: In this paper, we propose a decentralized control system for transporting a single object by a group of nonholonomic mobile robots. One of these mobile robots acts as a leader, who is assumed to be able to plan and manipulate the omnidirectional motion of the object. Other robots, referred to as followers, cooperatively transport the object by keeping the initial relative position to the object. Different from the conventional leader-follower type system of transporting a signal object by multiple robots in coordination, the present follower can plan an action based on its local coordinate and does not need any absolute positional information. The simulation result exhibits a good performance of the proposed system.

Global stability condition of fuzzy model-based controllers via evolutionary computation

Abstract: This paper presents a new concept for solving fuzzy controller stability problems on matrix inequalities via evolutionary computation (EC). The gain scheduling problem of a multi-model fuzzy system, which satisfies the Lyapunov stability criteria, is solved. The generalized eigenvalue problem can be directly introduced into EC in searching positive definite or positive semi-definite matrices, by making a penalty for an individual, that violates the inequality condition in order to solve the nonlinear constraints or linear matrix inequalities. Examples are given to illustrate the effectiveness of the proposed methodology.

Reinforcement learning with augmented states in partially expectation and action observable environment

Abstract: The problem of developing good or optimal policies for partially observable Markov decision processes (POMDP) remains one of the most alluring areas of research in artificial intelligence. Encourage by the way how we (humans) form expectations from past experiences and how our decisions and behaviour are affected with our expectations, this paper proposes a method called expectation and action augmented states (EAAS) in reinforcement learning aimed to discover good or near optimal policies in partially observable environment. The method uses the concept of expectation to give distinction between aliased states. It works by augmenting the agent's observation with its expectation of that observation. Two problems from the literature were used to test the proposed method. The results show promising characteristics of the method as compared to some methods currently being used in this domain.

Voice Communication in Performing a Cooperative Task with a Robot

Abstract: This paper investigates the credibility of voice (especially natural language commands) as a communication medium in sharing advanced sensory capacity and knowledge of the human with a robot to perform a cooperative task. Identification of the machine sensitive words in the unconstrained speech signal and interpretation of the imprecise natural language commands for the machine has been considered. The system constituents include a hidden Markov model (HMM) based continuous automatic speech recognizer (ASR) to identify the lexical content of the user’s speech signal, a fuzzy neural network (FNN) to comprehend the natural language (NL) contained in identified lexical content, an artificial neural network (ANN) to activate the desired functional ability, and control modules to generate output signals to the actuators of the machine. The characteristic features have been tested experimentally by utilizing them to navigate a Khepera® in real time using the user’s visual information transferred by speech signals.