This paper proposes walk pattern modeling for powered exoskeleton based on complex-valued neural network and reports its validity through experiments. We have been developing a powered exoskeleton to support workers at a nuclear power plant in time of hazard. The objective of the powered exoskeleton is to support a worker wearing a heavy radiation protection suit. We believe that conventional reactive power assist control based on EMG sensors is not feasible because they fail to measure the worker muscle activity robustly in the radiation protection suit as the worker has a lot of sweat in the high temperature and the humidity in the suit. Therefore, we have developed feed-forward control to assist the worker's motion based on the recognition of the worker's motion. Our previous studies use a simple k-nearest neighbor algorithm to model the motion of the powered exoskeleton, however, the algorithm is not for online learning and the estimated trajectory is not as smooth as we expected. This paper proposes a new modeling of walk motion of the powered exoskeleton based on a complex-valued neural network. The complex-valued neural network generally has good properties on learning speed and stability. This paper shows its validity for the modeling of powered exoskeleton's walk through experiments.

Recently, power assist suits which support human physical activities have been researched. We are developing a power assist suit for workers in a nuclear power plant. If a disaster happens, the workers have to wear heavy radiation protective equipment. The power assist suit is supposed to support the wearer so that it reduces the load of the radiation protection equipment during an operation. This paper proposes a power assist control method based on motion prediction using 9-axis motion sensors. The power assist suit enables rapid power assist because the motion sensor can detect the start of walking motion in real time. A motion database consists of angles and angular velocities of wearer's chest, lower limbs, and joint angles of the power assist suit. The power assist suit recognizes the motion of the wearer based on the database. Then, it assists the wearer based on the estimation of the motion of the wearer. We conduct experiments to evaluate the proposed method.

In this research, we develop a real-time fluid simulator, which uses smoothed particle hydrodynamics (SPH), for virtual environments including 3-dimensional fluid. SPH is a type of particle method and easy to control computational time by reducing or increasing the number of particles, however, it is difficult to change the number of particles while maintaining the same volume of fluid and the stability during a simulation. In this paper, we propose a new method to automatically adjust the number of particles, mass, radius in kernel functions and gains of density feedback for a real-time process while maintaining the stability. This paper also proposes to apply the method to a system using parallel processing with multi-thread programming. Finally, simulations are performed to estimate the effectiveness of the proposed method for parallel processing.

Stability of haptic interface is an important issue in virtual reality because ail operator touches directly a haptic interface device. The stability is influenced by sampling period and discrete-time property of control system. In this paper, for decreasing the influence of sampling system, we propose a haptic device with an analog circuit, which is placed between the computer and the haptic device and works as springs and dampers. The control system can specify the stiffness and the damping coefficients and their equilibrium. For displaying virtual objects that can move in virtual environment, we propose two methods to utilize the device, continuous-time coupling impedance (CCI) method and continuous-time object impedance (COI) method. We also analyze the passivity of each method for 1-DOF display system. Finally, some experimental results in two-dimensional virtual environment are presented to show the validity of the proposed approach.

In this paper, we discuss displaying a virtual object with coupling impedance by a haptic interface device with long sampling period. For such system to be stable, coupling impedance has to be set low. But it means that the operator feels soft surface of the virtual object and it is not suitable for displaying a virtual object with hard surface. This is because the system is a discrete time system. So we propose a haptic interface device with analog springs and dampers that exert reaction force continuously. The method can decrease the influence of the sampling period and make haptic interface more stable. Finally we perform some experiments with two dimensional virtual environment to confirm the effect of the proposed method.

We discuss a haptic display for 1-DOF grasping of a virtual object by two fingers. Much research has been done on fundamental analysis for stability of haptic displays. But it is difficult to apply the results immediately to grasping situations by two fingers, since the studies deal with a single device and a single object. This is because the fingertip force in grasping situations has two components, internal and external force. The conventional methods, which calculate the coupling impedance at each contact point separately, have no other alternative but to calculate the impedance for the sum of the internal and external force components. So even if only the impedance for external force should be changed, the impedance for internal force is also changed at the same time. We propose a method, in which the coupling impedance is calculated separately for internal and external forces. Second, we discuss the stability of the proposed method and compare it with the conventional method. Finally, we perform some experiments to confirm the effects of the proposed method.