論文

Two-wheeled robots are expected to be used at sites where humans work together. The subject of this paper is a two-wheeled robot with an arm, which can take out loads from shelves in the same manner as a human. When the center of gravity (CoG) of the two-wheeled robot shifts back and forth due to placing a load on the arm, the wheel position control performance deteriorates. A possible method to compensate for the shift in the CoG is to use repulsive compliance control (RCC). However, RCC has the disadvantages that the parameters must be adjusted experimentally and the CoG compensation is slow because the pitch angle is determined dynamically. To address these disadvantages, this paper proposes a method that uses model-based CoG compensation and position control for pitch angle. The model-based controller design simplifies parameter setting and achieves fast CoG compensation by statically determining the desired pitch angle. The validity of the proposed method is confirmed through experiments. The obtained results show that the performance of wheel position control is improved when the load is placed on the arm.

This paper presents a hybrid position/force control with equivalent mass matrices for position-and-force-sensorless manipulators and the evaluation of the control system. Position and force estimation methods in the controller are cascaded to estimate the position and force according to the respective characteristics of the electrical and mechanical systems. The proposed controller is designed to reduce vibration of the force control axes' motion from the cross-coupling factors of the dq-axis inductance, which cause position estimation error, and the cross-coupling effects among control axes from equivalent mass matrices, which expresses the relationship between the workspace force and acceleration. The position estimation based on the voltage equation, including the cross-coupling factors, derives compensation terms of the estimation error. The cross-coupling effects between position and force control axes in the equivalent mass matrices are designed to suppress the undesired force response vibrations. The experimental results show that the proposed system kept contact with an object during the rubbing motion by a 4-degree-of-freedom manipulator without position and force sensors. The utilized equivalent mass matrix reduced the force vibration caused by the vibration of the position estimation.

This paper presents an underactuated control system for a two-wheeled mobile robot (TWMR). TWMR is expected to use as an autonomous mobile robot (AMR) that works in human-collaborative spaces. TWMR has a smaller footprint and higher center of gravity (CoG), than typical AMRs. Furthermore, in this study, TWMR has an arm. Therefore, TWMR is superior in handling loads in human-presence spaces. However, TWMR is the underactuated system in the body's translational position and tilting angle. In addition, TWMR is affected by disturbances caused by loads. As a robust underactuated control system, the cascade PI/PD and synthesized pitch angle disturbance observer (SPADO) are considered. The cascade PI/PD results in significant translational position error and oscillatory responses. To cope with this problem, a proposed method minimizes both errors of the translational position and tilting angle within torque constraints. The proposed method utilizes the predictive functional control (PFC) for the underactuated control considering constraints. SPADO improves the accuracy of an internal model of PFC. The validity of the proposed method was confirmed compared with the cascade PI/PD through the simulation.

Upper limb exoskeletons are a current focus of technology for rehabilitation and assistance. Although the exoskeletons are actively researched and various applications are being explored, it poses many challenges because of their robot-human fusion of dynamics. In particular, control strategy is one of main problems to solve. Important points of the exoskeleton control are safety, usability, and prevention of insufficient/excessive assistance. Control of physical HumanRobot Interaction (pHRI) without electromyographic (EMG) measurements is one of solutions for the first and the second points. However, the exoskeletons assist based on only payload information in conventional researches and they have disadvantages in third point. Since ease of output of muscles is affected by the limb posture, assistance should be determined based on both payload and musculoskeletal information of each user. Hence, this paper aims to realize musculoskeletal model-based assistance considering ease of output of muscles. For this aim, hexagon distribution (HD)-based assistance torque calculator is proposed. HD expresses human output characteristics of a hand tip by modeling muscles of an upper limb including bi-articular muscles as 6 muscle groups. In proposed control strategy, first, the maximum human output is derived by HD. Second, it is judged whether assistance is sufficient or insufficient. Finally, assistance torque is adjusted in the case of insufficient assistance. To verify effectiveness of the proposed method, an experiment is conducted. As a result, it is confirmed that the proposed method can assist users based on their musculoskeletal characteristics.

Service robots working in human environments must have the ability to grasp a wide variety of unseen objects with appropriate grip force without knowing their properties and its response to action in a human environment. This research uses a novel Cyber Physical System framework to estimate and apply minimum grasping force for unknown objects under a task motion using only tactile motion data captured from sensor-less acceleration-based controlled common gripper. The goal is to use this novel framework for developing functions for position-controlled robots to perform real-time cognitive grasping and enable quick implementation across multiple robots. This paper uses soft sensing inputs from the earlier research to characterize unknown object properties like stiffness, mass and surface interaction and show how a data-based analytics algorithm learns a wide range of object properties and slip status along with task action dynamics features, to propose the grasping action with minimum applied force. We validated that the minimum grasping force is able to hold the object firmly during any task motion. In order to improve the learning, the paper proposes a part of the framework to use virtual simulation to gather more learning data and show results that compare well between simulated results with experimental data. We use the simulated data to train Reinforcement Learning approach and show that small variation in object width can be learnt to identify optimum gripping position.

This paper aims to verify the detection accuracy of Zero Moment Point (ZMP) measurements of the human body using the Kinect sensor. In this paper, ZMP is measured based on a multi-link model of the human body. ZMP calculation algorithms using a Low Pass Filter (LPF) and Savitzky-Golay (SG) filter were constructed to investigate filtering methods that can measure ZMP with high accuracy and small lag. The accuracy and responsiveness of these filters were evaluated through experiments measuring ZMP during human walking. Experimental results show that the detection error in ZMP is dominated by the differential noise in the acceleration calculation, and applying the filter only to measured acceleration values increases responsiveness with the same accuracy as when applied to ZMP. We also found that the SG filter is superior to LPF in noise suppression and responsiveness of ZMP measurements, though it occurs overshoot in measurements.

Evaluation of the falling risk is significant for older people not to fall and become in need of care. The static posturography test is adopted to evaluate the lack of balance ability that causes the falling. This test measures the center of pressure (COP) of humans and evaluates it by analyzing the data. There are three main analysis methods: descriptive statistics, stabilogram diffusion analysis (SDA), and mathematical modeling. Descriptive statistics and SDA represent only statistical values without the mechanism of the posture control system. However, mathematical modeling evaluates the system from the model coefficients containing physical meanings, and we focus on it. The stochastic differential equation of stiffness, viscosity, and stochastic terms are introduced to the mathematical model for the static posturography test. The COP trajectory reproducibility of the model is evaluated using SDA. SDA is a time-based analysis suitable for evaluating the dynamic posture control system. In the experiment, we compare the SDA parameters between older and younger subjects and evaluate the reproducibility of the model using the SDA parameters. As a result, we confirm that the SDA parameters can classify based on age to a certain extent, and the model has higher reproducibility than the conventional model against SDA.

This paper presents a force sensorless workspace hybrid position/force control method with equivalent mass matrices switching for rubbing motion. The equivalent mass matrices express an interference between control axes. Since the determinant of the matrix computed from the design values of robots becomes near zero around the singular configuration of robots, the utilized matrices are often arbitrarily designed. However, inappropriate matrices conditions deteriorate control performance and could cause vibration during rubbing motion. This study reveals suitable equivalent mass matrices for rubbing motion depending on the velocity of the force control axes. Moreover, a hybrid controller with equivalent mass matrices switching depending on the expected velocity of the force control axes in tasks is proposed. The validity of the proposed hybrid controller is demonstrated by the experimental response of rubbing motion with a parallel link type direct-drive four-link manipulator.

With the recent decreasing birthrate and aging population, cooperative robots are attracting attention for a new workforce. As cooperative robot, robots with a small footprint and high center of gravity, like humans, can be considered more suited. Robots with these features are generally unstable, and posture stabilization control becomes important when the center of gravity is shifted due to holding objects or other factors. Therefore, in this study, we select a two-wheeled robot as the target robot and carrying luggage as the target motion, and establish the posture stabilization. Many studies about two-wheeled transport robots do not adequately consider the shift in the center of gravity, which may lead to rapid acceleration and negatively affect position tracking performance. The purpose of this paper is to compensate the shift in the center of gravity of the vehicle, and can perform transport operations in a steady state. To achieve this purpose, we propose to apply Repulsive compliance control(RCC) to a two-wheeled transport robot. Several experiments show its effectiveness.

A caregiver has an important role to not only help patients with high levels of disability to move around, but also be aware and manage the patient's state of health. Visual cues on the patient's expression is an especially important source of information on the patient's well-being, especially for those with high levels limitations on cue sending, such as Amyotrophic Lateral Sclerosis (ALS) patients. Thus, this paper proposes a system that enables the caregiver to control the movement of the wheelchair while waking alongside the wheelchair where they can also interact with the patient directly. The proposed approach has been validated through experiments.

Tracked vehicles generally involve slippage owing to the interaction between the road and track surfaces, which renders accurate motion control difficult. This paper proposes a velocity estimation method for a tracked vehicle with slippage, and its application to driving force control. In this method, the disturbance estimated by a disturbance observer was used as information related to slippage, and a neural network was constructed for velocity estimation. In addition, a driving force observer was designed using the estimated velocity. The driving control of the tracked vehicle to suppress slippage was achieved by using the feedback of the estimated driving force. The proposed method was evaluated experimentally through the velocity estimation performance and slip suppression performance tests.

This paper proposes a tracking control method that suppresses slippage by using the driving force of a tracked vehicle. First, the velocity of the tracked vehicle including slippage is estimated using a disturbance observer and machine learning technique. This estimated velocity is utilized to design an observer that can estimate the driving force of the crawler. By distributing and controlling the driving force, tracking control with reduced slippage can be realized. The experimental results demonstrate the tracking performance of the proposed control system.

The operational frequency of a transformer is an important factor in the configuration of an isolated DC-DC converter. However, discussions of this frequency are inadequate. This study focuses on the frequency characteristics of the transformers. The frequency characteristics of the impedance and loss of the transformer were obtained to explain their fundamental attributes. When a transformer is employed in a power electronic transformer (PET), the waveform of the voltage is rectangular. The loss under this waveform was obtained in order to observe its characteristics. Furthermore, the problems with the isolated DC-DC converter are explained. The natural frequency, composed of leakage inductance and equivalent-parallel capacitance, determines the performance of the isolated DC-DC converter. When the operational frequency of the isolated DC-DC converter is below its natural frequency, the loss increases under the no-load or light-load conditions. This increase in the loss is caused by the current, which contains the natural frequency of the transformer owing to the ringing.

The purpose of service robots is to work and handle a wide range of objects in a human environment. Therefore, they must have the ability to grasp unknown objects with the appropriate force, by dynamically identifying their stiffness, mass, and surface-interaction properties. To ensure the successful grasping of unknown objects, it is significant to estimate the object response properties. In this paper, the cognitive functions to efficiently perform such operations with only tactile motion data and supervised learning using limited data, have been demonstrated. The study uses pinch-grasping performed by a 2-fingered robot experimental setup with a pre-designed motion sequence, over 7 different objects having a wide range of properties to capture the motion data. The proposed approach avails pre-processed motion data to train neural networks for accurately predicting object response properties for 3 unseen test objects. We have done a comparative study among three neural network architectures CNN, LSTM, and CNN-LSTM and we found that novel CNN-LSTM outperforms the other two models in terms of accuracy. The CNN-LSTM regression model achieves 0.98 and 0.97 of R2 value for the prediction of stiffness and mass, respectively and the CNN classification model achieves 94% of accuracy for the surface interaction classification.