Motion planning of a biped robot based on ZMP is quite popular in recent years. In this method, ZMP reference trajectory is planned at first so that ZMP stays inside of the support polygon. Then, walking pattern is generated based on the trajectory. Conventional methods based on ZMP reference trajectory did not take the dynamics of bipedal locomotion into account. Therefore, the basis of the motion planning was ambiguous and the biped robot did not get human-like walking. This paper proposes ZMP reference trajectory generation with inverted pendulum model by using VSP (Virtual Supporting Point). With this, smooth ZMP reference trajectory based on the dynamics of the inverted pendulum is generated uniquely according to the desired stride and walking cycle.

View full abstract

There are two types of force control system, force control with inverse Jacobian matrix and transpose Jacobian matrix. Both of the controllers, however, have a defect. The characteristic of the former controller is that it has singular point due to inverse Jacobian matrix. Motion of a robot becomes unstable at this point. On the other hand, with transpose Jacobian matrix, it is used only in static state.
In this paper, a novel force control method is proposed. This method does not have any singular points and is able to be applied to dynamic state. Furthermore, tracking accuracy of the proposed controller is better than that of the conventional methods.

View full abstract

This paper describes a pushing motion for an unknown object by a humanoid robot. This method deals with an object whose parameters are not known. In order to push the unknown object, the robot motion must be decided according to the object. Therefore, the robot judges whether the unknown object is movable or not. When the robot judges the object is unmovable, the robot stops pushing. On the other hand, when the robot judges the object is movable, the robot continues pushing and walking. When the robot walks during a pushing motion, the influence of the pushing motion may destabilize a walking motion. Therefore, ZMP trajectory is modified by reaction force response before one step. Additionally, cycle time in double support phase is also modified by using sigmoid function. In the result, the heavier the object is, the more slowly the robot moves like a human. With these methods, the robot is able to push the unknown object during walking. Furthermore, even if parameters of the object change, this proposed method is available to continue pushing the object.

View full abstract

This paper discusses robustness on model error of time delayed control systems with communication disturbance observer (CDOB). A feature of the time delay compensation method based on the concept of network disturbance (ND) and CDOB is that the method can be utilized without delay time model (predictive control methods like Smith predictor usually need delay time model). Therefore the method can effectively compensate time delay effect even in the case with time-varying delay. However the method needs a model of controlled system for implementation. Then the robustness on model error of controlled system model is a very important issue for development of the time delay compensation method. In this paper, the robustness on model error is discussed from the perspective of robust stability and steady-state characteristic. At first, robust stability problem is analyzed comparing to that of Smith predictor. Then steady-state characteristic is also studied in the same manner. The validity of analytical results on both robust stability and steady-state characteristics are verified by experimental results.

View full abstract

Variations and/or uncertainties in environments of mechatronic systems, e.g. electrical/mechanical parameter changes and nonlinear components, generally deteriorate the motion control performance. In the research, the fast and precise position settling performance for parameter variations in positioning devices can be improved by techniques of an on-line parameter identification and an initial value compensation. The proposed technique allows the positioning systems to be adaptive and robust for unknown parameter variations. The effectiveness of the approach has been verified by numerical simulations and experiments using a prototype.

View full abstract

This paper proposes novel micro/macro bilateral control based on acceleration control with standardization matrix. In bilateral control, force control and position control should be realized simultaneously. However, they are not able to be realized in one real axis at the same time. Thus, force control and position control are realized in virtual mode space in this paper. Then, the proposed standardization matrix is able to harmonize the standard of macro master system with the standard of micro slave system in the virtual mode space. With the proposed method, the transmission of force sensation from the real micro environment is realized. The experimental results are shown to verify the viability of the proposed method.

View full abstract

This paper proposes an available method of obstacle avoidance control by using a single CCD camera and ultrasonic sensors mounted on the mobile robot. First, depending on the change of the brightness on the image that occurs from the moving of the mobile robot, we calculate the optical flow by the block matching method based on the normalized correlation and detect the obstacle area on the image. Further, in order to reduce the error of the detection area, by combining the distance information obtained by ultrasonic sensors on the image shown the obstacle area we decide the position of obstacle with high accuracy. Dealing with the position information, we make the reference points for generating the trajectory of the mobile robot. This trajectory is smooth and is generated by minimizing a certain cost function. Then, the mobile robot moves according to the trajectory while avoiding around the obstacle. Finally, usefulness of our proposed method is shown through some experiments.

View full abstract

This paper presents a modeling methodology for unknown disturbances in mechatronics systems, based on a disturbance estimation using an iterative learning process. In the disturbance modeling, the nonlinear friction is especially handled as the disturbance in mechanisms, which mainly deteriorates the trajectory control performance. The friction can be mathematically modeled by using the learned estimation data, as a function of displacement, velocity, acceleration, and jerk of the actuator. This model has a distinguished feature that the friction compensation can be achieved with a generalization capability for different conditions. The proposed positioning control approach with the disturbance modeling and compensation has been verified by experiments using a table drive system on machine stand.

View full abstract

This paper proposes an LMI-based design approach for command shaping in the fast and precise positioning control of industrial table drive systems on machine stand. In the command shaping, the position response time and the vibration suppression of machine stand can be directly considered as the constraint conditions in LMI design frame work. The effectiveness of the proposed shaping has been verified by numerical simulations and experiments using a prototype with a 2-degrees-of-freedom robust positioning controller.

View full abstract

This paper deals with a design method of fault-tolerant control system based on Generalized Internal Model Control (GIMC) structure consisting of a standard outer loop feedback controller and an extra inner loop controller. The distinguished feature of GIMC structure is that the controller design for performance and robustness may be done separately. The outer loop controller is designed for nominal performance using some controller synthesis to meet (nominal) control specification, while the inner loop controller is designed to make a trade-off between robustness and performance. This feature is suitable for fault-tolerant control. The outer loop controller is designed for fault-free case, and the inner loop controller for faulty case. In the conventional methods, the inner loop controller is designed to maximize the robust stability margin without information on fault. Therefore, the performance in the faulty case tends to become conservative. In this paper, the inner loop controller is directly designed from experimental data collected from the faulty system. Since the collected data contains information on the fault, conservativeness in the conventional methods is decreased. The inner loop controller is designed by Virtual Reference Feedback Tuning (VRFT). VRFT is a direct design method from input-output data without identifying any models. Since complexity of the controller can be specified by the designer, no complexity reduction has to be required, which becomes advantageous upon implementation. The effectiveness of the proposed design method is confirmed by an experiment.

View full abstract

In this paper, the estimation method of only gravity and the compensation method of only the disturbance caused by gravity are proposed. The proposed method is able to estimate gravity, even when the system has the unknown posture. The proposed method is based on the concept of “environment quarrier”. Then, environment quarry is expanded to multilateral control. In this paper, two dummy robots that are the environment quarrier are implemented besides master-slave system, in order to abstract only gravity disturbance. The proposed method is adaptable, even when mass of master system differs from one of slave system. To verify the viability of the proposed method, experimental results are shown.

View full abstract

This paper proposes a data-driven method named Fictitious Correlation-based Tuning (FCbT) for tuning of Linear Time Invariant (LTI) multivariable controllers. The parameters of the controller are updated directly using the data acquired in closed-loop operation. This approach allows one to tune diagonal elements of the controller transfer function matrix to satisfy the desired closed-loop performance, while the other elements are tuned to mutually decouple the closed-loop outputs. Moreover, FCbT requires only one-shot experimental data for an off-line nonlinear optimization in order to obtain the optimal parameter minimizing a “Fictitious” cross-correlation function. In this paper, we enhanced FCbT to tuning of 2DOF controllers, which are often adopted in various control problems. In model-based control synthesis, a feedforward controller is generally synthesized as a product of a reference model and an inverse model of the plant. It means that closed-loop identification of MIMO system can be performed by FCbT. The effectiveness of FCbT and application to closed-loop identification are shown by simulations and experiments.

View full abstract

This paper proposes a high-performance robust tracking control system for high speed optical disk recording system. The proposed tracking control system consists of robust feedback control system and perfect tracking control system with prediction of tracking error. Improving a free parameter, the proposed robust feedback control system has a good performance for the sudden disturbance suppression. For suppressing the periodic-disturbance in condition of high disk rotation speed, the perfect tracking control system with prediction of tracking error is improved by using zero phase low-pass filter (ZPLPF), in comparison with the ZPET feedforward control system. The experimental results point out that the proposed tracking servo system has a precise tracking response against both the periodic disturbance and the sudden disturbance.

View full abstract

This paper proposes a new generation method of the position reference having both the vibration suppression performance and fast tracking performance for industrial robots. It is important for industrial robots to drive high speed and high accuracy. In such cases, the vibration phenomenon is generated. Conventionally, the notch filter is used in order to suppress the vibration phenomenon. It is able to eliminate the natural frequency component. However, the reference phase error is generated. The reference phase error causes the locus error in the robot. Therefore, the accuracy of the robot is deteriorated by using the notch filter. The proposed method overcomes this problem by using the compensation gain. The proposed compensation gain is used in order to calculate the reference phase error. The compensation of reference phase error is attained by the feedforward input. The numerical and the experimental results confirm that the proposed method is valid to suppress the vibration phenomenon and improves the phase error.

View full abstract

Artificial acquisition and reproduction of human sensations are basic technologies of communication engineering. For example, auditory information is obtained by a microphone, and a speaker reproduces it by artificial means. Furthermore, a video camera and a television make it possible to transmit visual sensation by broadcasting. On the contrary, since tactile or haptic information is subject to the Newton's “law of action and reaction” in the real world, a device which acquires, transmits, and reproduces the information has not been established. From the point of view, real-world haptics is the key technology for future haptic communication engineering.
This paper proposes a novel acquisition method of haptic information named “haptograph”. The haptograph visualizes the haptic information like photograph. The proposed haptograph is applied to haptic recognition of the contact environment. A linear motor contacts to the surface of the environment and its reaction force is used to make a haptograph. A robust contact motion and sensor-less sensing of the reaction force are attained by using a disturbance observer. As a result, an encyclopedia of contact environment is attained. Since temporal and spatial analyses are conducted to represent haptic information as the haptograph, it is possible to be recognized and to be evaluated intuitively.

View full abstract

Recently, the demand for man-machine cooperative system is growing in many industrial fields, particularly, in the cell production and the flexible manufacturing systems. In the design of man-machine cooperative system, the characteristics of the human behavior must be considered. This paper presents a new design strategy of a switched assist controller of a man-machine cooperative positioning task considering the human behavior model based on a continuous/discrete hybrid dynamical system. First, the human behavior model is identified as a Piece-Wise ARX model using the k-means clustering and support vector machine. Secondly, the switched assist system is designed as switched impedance controller based on the identified hybrid human behavior model with optimization. The optimization was solved by executing a forward calculation of the closed loop response of the human behavior model and controller with criterion reflecting the task time and the force effort. As the result, we can find the optimal parameters of switched impedance controller.

View full abstract

This paper presents a modeling and analysis strategy of advanced complex human motion based on a hybrid system framework. The motion modeling problem is addressed as a parameter estimation problem of the piecewise ARX model which is a class of the hybrid dynamical systems. The advantages of motion modeling based on the hybrid dynamics are: (1) The complex dynamical behavior can be represented by underlying some primitive dynamics together with switching mechanism between them. (2) The obtained model could be useful for various application, such as, design of assist controller, recognition of the human behavior, and so on. Finally, the developed strategy is applied to a driver's vehicle following task and a peg-in-hole task which represent two typical classes of the complex human motion. The obtained results show the usefulness of the hybrid dynamics for representing complex human motion.

View full abstract

Recently, skill preservation of an expert has been a serious problem of the medical or production fields. It is difficult to acquire a human motion including touching motion by the conventional visual-based system. A touching motion is inherently bilateral, since an action is always accompanied by a reaction. A bilateral force feedback control is necessary to acquire the human's skill.
This paper proposes a haptic teletraining system by multilateral force feedback control. Since the multilateral control transmits haptic information among three or more remote systems, it is possible to train remote trainee by one skilled trainer simultaneously. In the proposed haptic teletraining system, each local system consists of a master-slave system. Since a touching motion is subject to the “law of action and reaction”, it is possible to decompose the force information into action force and reaction force by using the bilateral control. Thus the skilled motion by a trainer is acquired and saved as a digital database integrated with his haptic information.
The proposed system is applied to one-trainer/two-trainee system. As a result, a motion including contact with the environment is transmitted and trained with vivid force feedback. The experimental results show viability of the proposed method.

View full abstract

In case of decelerating a permanent magnet motor, an over-voltage occurs on the inverter DC voltage due to the regenerative energy of the motor. In order to reduce the over-voltage, a brake circuit is usually applied to the DC circuit.
In this paper a reduction technique of the over-voltage without the brake circuit is described. We proposed the method for controlling the motor d-axis current optimally according to the motor q-axis current to reduce the over-voltage. This method is applied to the salient-pole machine. As a result over-voltage reduction is achieved on a certain condition.

View full abstract

The paper presents analysis and improvement of the control accuracy for a positioning device with a ball screw-driven table system. In the analytical studies, effects of variations of mechanical parameters on the positioning accuracy are examined, where a precise simulator with static and dynamic friction models are provided to simulate actual responses in the positioning. The analyses clarify that the mechanical variations behave as equivalent disturbances with components in low frequency range and deteriorate the positioning accuracy. A robust disturbance observer, therefore, is designed in the frame work of 2-degrees-of-freedom positioning controller, in order to improve the disturbance suppression characteristics. The effectiveness of the proposed design has been verified by experiments with a ball screw-driven table system on machine stand.

View full abstract

Generally, intelligent vehicle robots are strongly needed in many fields. They are expected to assist our daily lives. In order to implement them in practice, we should mount useful functions including path planning, navigation, and obstacle avoidance, etc. on the robots. This paper focuses on obstacle avoidance control technique to multiple obstacles. The used vehicle as a platform is the mecanum wheel omnidirectional vehicle which is driven on floors and mounts a scanner range sensor. The presented algorithm has the feature that multiple obstacle avoidance control is simply come down to the mono-obstacle avoidance control scheme. Finally, the experimental results are shown in order to evaluate the presented algorithms.

View full abstract

In Hard Disk Drive (HDD), there are two control modes. One is a head positioning control mode, the other is a seek control mode. In the head positioning control mode, a feedback controller is optimally designed to suppress a disturbance. In the long span seek mode, a velocity feedback control system is applied in order to move a head fast. Thus, a HDD has plural control systems, and the head is moved to the target position while changing from one control system to the other. However, the changing of the control system causes a discontinuous control signal, which activates the resonance mode of an actuator. The past methods only can decrease the discontinuous control. Therefore, a single control system that can be used for both a seek control mode and a head positioning control mode is necessary for narrow track pitch. In the proposed method, the feedback controller is decomposed to an integrator and a phase compensator. The VCM model is updated by the output of the phase compensator, and the integrator and the output of the velocity feedback controller control the VCM. The validity of the proposed method was confirmed by numerical and experimental results using a miniature 2.5-inch Hard Disk Drive.

View full abstract

A single-phase voltage-tripler rectifier operating in the switch-mode is studied. This is based on a diode pump circuit, in which the voltage across the boost capacitor adds to the supply to pump twice the peak supply voltage onto the output capacitor. An only active power switching device also provides the necessary control over the current. The experimental prototype using an insulated-gate bipolar transistor is implemented to investigate the operation under the current-controlled pulse-width modulation. The experimental results confirm that the input current can almost be waveshaped sinusoidally with a near-unity power factor.

View full abstract