Takahiro Kosaki

Model-based fuzzy control system design for magnetic bearings

This paper describes the application of a model-based fuzzy control technique to active magnetic bearing systems. The stabilization problem of the magnetic bearing systems is reduced to a two degrees-of-freedom control problem about rotational motion including interaction. We derive a multivariable fuzzy controller for this problem from the fuzzy model of rotor dynamics using the concept of parallel distributed compensation. Moreover, the stability of the designed control system is proved by means of the Lyapunov stability condition. A common Lyapunov function is found by solving an optimization problem based on linear matrix inequality conditions. The performance of the developed fuzzy controller is verified through simulation.

Control of pneumatic artificial muscles with the just-in-time method based on a client-server architecture via the internet

This paper proposes the concept of “networked JIT control” based on a client-server architecture via the Internet, which is an expanded version of the Just-In-Time (JIT) control. In the proposed concept, the server has a database for the JIT control and the client has a plant to be controlled. The server performs the computation of an inverse model of the plant by utilizing data stored in the database. The client executes a feedback control algorithm, communicates data for the JIT control with the server through the Internet, and feeds control commands to the plant. In addition to the general advantages of the JIT control, this concept enables to share the database on the server with the clients and to reduce the costs of the addition of more clients. A control system of a robot manipulator driven by pneumatic artificial muscles, which are elastic and safe for humans but nonlinear and difficult to model, is constructed based on the proposed concept. Experimental results show that data obtained by communicating with the server lead to an improvement in the control performance and that the proposed concept is realizable.

A pneumatic arm power-assist system prototype with EMG-based muscle activity detection

Wearable robotic assist devices need to be safe for the user. Pneumatic actuators are suitable for driving those devices because they are highly safe for humans. However, the responsiveness of pneumatic actuators is lower than that of electric actuators. In this paper, we propose an approach to muscle-burst onset detection by monitoring surface electromyogram (EMG) signals, and utilize it to trigger the actuator of a pneumatic power-assist device for the human arm. Experimental verification demonstrates the validity of our approach for detecting muscle-burst onset by EMG signals online, and the applicability of the EMG trigger for the pneumatic power-assist device.

A parallel-link robot prototype driven by pneumatic actuators with variable inclination mechanisms

Parallel-link robots, in general, possess high power and high precision owing to the parallel arrangement of actuators; however, their workspace is smaller than that of serial-link robots. In this paper, we develop a parallel-link robot prototype with pneumatic actuators in which a mechanism for varying the actuator inclination is incorporated for the purpose of enlarging the workspace. Our parallel-link robot realizes the rotational and translational motions of the end effector principally by the linear reciprocating motions of pneumatic linear drives mounted on the base. Auxiliary pneumatic actuators are used to adjust the inclination angles of those main pneumatic linear drives. The use of pneumatic actuators to realize the proposed parallel-link robot results in lightweight, compact, and low-cost construction. The workspace and motion transmissibility of our parallel-link robot are analyzed through simulations based on kinematics; then, experimental investigations are carried out using the prototype.

Drone were me: Flying drone as avatar by aligned stereo view and head pose

In this paper, we present a natural interface between a drone, i.e., a miniature aerial vehicle (MAV), and a human, in which a stereo-vision-MAV is flied via a HMD (Head Mounted Display) with inertial sensors. A stereo camera is mounted on a MAV and the stereo images are transmitted to a remote PC. Then, a user controls the motion of the MAV by his/her head pose, which is measured from the inertial sensor of the HMD, while looking at the stereo images via the HMD as if the MAV is an avatar of the user. Thanks to a stereo-vision camera, it makes it possible that not only a user can sense the distance information between the MAV and obstacles from the HMD, but also a MAV can avoid collisions using the measured 3D information even if human error occurs during the operation.

Performance improvement of networked Just-In-Time control for pneumatic artificial muscles

Here we present techniques for improving the performance of our previously proposed networked Just-In-Time (JIT) control, which is based on a client-server architecture via the Internet. In this control, a database for the JIT control is stored on a server and shared with clients. The clients communicate mutually with the server through the Internet at a constant interval, and they feed control commands to their respective plants. However, since the communication between a client and a server involves a time delay, and since the overgrowth of a database prolongs the time required to process data, in the present study an algorithm for storing data in a database and for handling data of the database in a shorter time was incorporated into the server. In addition, because a longer communication time extends the interval for adding a JIT control command and causes performance deterioration, we vectorized the sending and receiving control commands to reduce such deterioration. The experimental results for a pneumatic artificial muscle test apparatus showed that the control system with the proposed techniques achieved higher performance than that developed in our previous work.

Dwell Time Switching Control for Cooperative Work of Two Mobile Manipulators

Abstract We propose a new switching control methodology for cooperative work of two mobile manipulators. For the purpose of controlling these multiple nonlinear systems with constraints, it is effective to use multiple controllers and switch them optimally by the supervisory system. An introduction of dwell time is useful in designing the controller. We developed a sophisticated software program to design a switching control system and also to operate the real system directly in consideration of constraints. Finally, we will demonstrate the effectiveness of the system by simulations and experiments.

Sliding mode fuzzy control of magnetic bearing systems

Abstract The stabilization of a magnetic bearing system is reduced to the two degrees-of-freedom control problem about rotational motion of a rotor. In this study, a sliding mode fuzzy controller is taken as one of robust approaches for this problem. When the rotor rotates, it is required for the reliable operation to overcome two major obstacles: imbalance force and gyroscopic effects. For imbalance force compensation, a VSS disturbance observer is used. Fuzzy decoupling units are also attached to the control law to reduce coupling effects between the rotations. Experimental results show the stable rotation of the rotor is achieved using the fuzzy controller.