論文

Exoskeletons are nowadays at the core of research for human enhancement, rehabilitation and prevention of musculoskeletal disorders (MSDs). MSDs are one of the major causes of long-Term disability and illness among workers, and some of them could be prevented by the use of exoskeletons. However, the control of those devices is not always intuitive, and thus, they are poorly accepted. Some control methods using surface electromyography (sEMG) information are currently under research and give promising results. This paper aims to introduce a robust sEMG-based intention detection relying on an LSTM neural network. The method consists of collecting three channels of sEMG together with the position and acceleration angle and rotational speed of the end effector to teach a long short-Term memory neural network to predict the angle of contraction of the elbow. The negative effect of external factors on the prediction, such as the presence of load in the hand or the differences between individuals, are highlighted, and the benefits of calibration methods to solve this problem are presented.

This paper proposes a mutual entrainment method at the hip joint using tactile stimuli and analyzes the effects of the degree of synchronization between the gait cycle and the cycle of stimuli. Previous research insists that Central Pattern Generator (CPG) creates the rhythm of behavioral patterns and the rhythm of CPG can be changed by external stimuli. Some previous research proposed entrainment methods that apply stimuli and bring the CPG cycle to the ideal one. Then, this paper uses pneumatic gel artificial muscles (PGM) to apply tactile stimuli to subjects. To reach the ideal gait cycle, this paper proposed a mutual entrainment method. The proposed method changes both the gait cycle and the cycle of the stimuli applied by PGM. Entrainment is completed when both cycles match, and the gait phase and the phase of the stimuli match. The entrainment uses a phase oscillator model to express the relationship between the gait cycle and the driven cycle of PGM. The method changes the driven cycle of PGM based on the relationship. In the phase oscillator model, coupling strength is a coefficient that expresses how the relationship between the subject and PGM is strong. This paper focuses on the coefficient. By changing the value of the coefficient, the gait cycle and the PGM cycle match more quickly. This paper sets the hip joint as the target of mutual entrainment. The hip joint is the first part to operate when the heel contacts the ground in the gait. Moreover, the hip joint consumes 40 - 50% of the total energy while walking. Thus, The hip joint plays a role in the movement of the legs. The experimental results show that the convergence speed and error to the target phase change depending on the coupling strength. The coefficient also results in changes in the magnitude of Ground Reaction Force (GRF) during entrainment.

The current labor shortage is a global problem, and human-collaborative robots are important to solving this problem. In particular, there are high expectations for Two-wheeled robots (TWR) with an arm because of their human-like characteristics. To use a TWR more flexibly, it is necessary to devise a path-finding algorithm especially for the TWR. In this paper, path planning algorithm using spline interpolation with a velocity profile is proposed considering not only the time to achieve the task but also the stability of the robot. Simulations and experiments demonstrate the effectiveness of the proposed method.

This paper proposes a platform-independent overload detection system for electric forklifts based on the analysis of contact point jerk. Unlike conventional methods that require hydraulic system modification, the proposed system employs dual Time-of-Flight (ToF) sensors. The core of this method lies in analyzing the jerk (rate of change of acceleration) at the moment when the fork contacts the pallet, enabling overload detection within approximately 50 mm of lift initiation—before the load is fully raised. We developed a power-law extrapolation method based on hydraulic system dynamics to estimate feature distributions in the overload region from limited experimental data in a physically plausible manner. Experimental validation using a TOYOTA 8FBE10 forklift confirmed that contact point jerk features exhibit statistically significant differences between normal and overload conditions. Simulation results considering variable load centers (400 mm to 1000 mm) achieved detection accuracy exceeding 96% in the 400–800 mm range. At LC = 1000 mm, limitations emerged due to class imbalance arising from the relationship between experimental data and threshold positions, resulting in reduced specificity. This system enables early warning before high-altitude operations without any modification to the existing forklift systems, contributing to the prevention of tip-over accidents.

This paper presents a force sensorless knee flexion detection and gait assistance system for active prosthetic legs, evaluating its impact on gait characteristics. The proposed method can estimate the ground reaction force without relying on ground strain or friction coefficients by using Reaction Torque Observer (RTOB). In the proposed approach, an exoskeleton devise is mounted to the prosthetic leg and the users utilize it as an active prosthetic leg. Waist trajectory measurement was used to evaluate whole-body movement, comparing gait characteristics in able-bodied individuals, passive prosthetic leg users, and active prosthetic leg users. These analyses contribute to advancing gait assistance evaluation and developing control methods with a small number of sensors for active prosthetic legs.

Owing to their high center of gravity (CoG) and small footprint, two-wheeled robots (TWRs) are expected to be used in collaborative tasks with humans. When a TWR uses its mounted arm to transport a load, its CoG shifts, deteriorating translational control performance. Conventionally, CoG shift is compensated using the torque applied to the arm by the load. However, because the load-weight information is not considered, gravity compensation is insufficient. This study aims to improve translational control performance through precise gravity compensation. The proposed method is a gravity compensator comprising a model-based compensator, which clarifies the physical characteristics, and a model-error compensator, which addresses errors caused by the load weight. A virtual point is also introduced to improve body stabilization and translational control performance. The proposed method enables precise gravity compensation for a TWR and improves its translational control performance. The experimental results demonstrate the robustness of the proposed method in maintaining translational control performance regardless of the load conditions. This compensation method significantly affects the practical deployment of TWRs in human collaborative environments, such as factories and warehouses.

This research aims to design a bi-copter AUV and evaluate its feasibility. Autonomous Underwater Vehicles (AUVs) that sail independently without cables and human operators are in the spotlight as an alternative to human divers. To miniaturize AUVs to operate in narrow spaces, we propose an AUV with two thrusters and two pitch angle servo motors: a 'bi-copter' AUV. The bi-copter is expected to be power efficient, and have high turning performance. However, precise and robust control performance is required to overcome its underactuated characteristics and disturbances. To enhance pitch motion stability, a Pitch Motion Compensator (PMC) with a Pitch Angle Disturbance Observer (PADO) is implemented and controls the position of a moving mass inside the AUV body. We propose a PID-Backstepping Hybrid (PBH) control composed of the PID position controller and the backstepping attitude (roll and yaw motion) controller for the position and attitude control. In addition, We propose the yaw angle reference value calculation methodology to control the y position indirectly. Our proposed method is evaluated by the simulations using MATLAB/Simulink and Simscape including wave disturbance and its feasibility is confirmed.

This paper investigates the deployment of transport Two-Wheeled Mobile Robots (TWMRs) within warehouse environments. TWMRs have superior maneuverability and a smaller footprint compared to conventional steering-type Autonomous Mobile Robots (AMRs). This feature makes TWMRs well-suited for narrow spaces and high-reach tasks. Despite these advantages, TWMRs are inherently unstable, necessitating sophisticated control systems to maintain balance. In this study, we propose a control system that enables remote operation of the TWMR, allowing the operator to adjust the pitch angle command. The TWMR is equipped with various control mechanisms, including Disturbance Observer (DOB), PD controller and Lyapunov controller. These components work in conjunction to stabilize the TWMR and enhance the translational position tracking performance. Simulation results demonstrated the effectiveness of our approach, showing that the proposed system successfully stabilized the TWMR and reduced pitch angle variations, especially under load conditions. It also suggested the potential of integrating human control ability with advanced control systems for improved TWMR performance.

The research field of muscle rehabilitation and strength recovery is constantly evolving with new techniques and assistive devices being proposed every year. In particular, the need of specialised structures and clinicians, the cost of the rehabilitation and the rising numbers of patients among injured athletes and elderly people, are making rehabilitation robotics a cost-effective alternative. In the last years, researchers have mainly focused on proposing assistive devices to be used during the rehabilitation or new techniques, in order to make the whole process more effective. Nonetheless, the accessibility and the cost of most of the treatments, still leaves a lot of room to improve for a more inclusive access. Besides giving the freedom to the patient to carry out the rehabilitation at home, it is crucial to properly choose the exercise according to the muscles which need to be trained and the condition of the subject. Especially for injured athletes and for the elderly, isometric exercises and static strength training are a good solution to prevent stressing the joints and moving the damaged muscles. Hence, this paper aims to introduce a new musculoskeletal-specific design framework for upper-limb isometric exercises, after having evaluated with a Hexagon Distribution (HD) the muscle condition of the subject. In addition, a 2-DOF wearable and lightweight exoskeleton is introduced to maintain the correct posture of the arm while performing the task. Finally a randomized controlled trial (RCT) is conducted with a cohort of different groups to show the effectiveness of the proposed method.

Contact safety and high control performance are important for human assistive collaborative robots and wearable technology. A series elastic actuator (SEA) has a spring inserted between the motor and the load, which has high compliance. A motor/load-side encoder that acquires motor and load angles enables precise control. SEA with a motor/load-side encoder is appropriate for a human assistive system. However, resonant poles due to the low stiffness of SEA cause resonant vibration and low robustness. Human assistive systems need to achieve control that is precise and robust to disturbances from humans and the environment. This paper proposes a novel control system design that achieves vibration suppression and high robustness. The proposed method consists of torsion torque control and an inverse system filter. By dividing the suppression of resonant pole effects and improving robustness into different controllers, the proposed method makes it easier to design appropriate gains for human assistive systems than conventional methods. The validity of the proposed method and the effect of stiffness error were verified by simulations of position and force control.

This paper proposes a disturbance rejection method for a multi-degrees of freedom (MDOF) manipulator with flexible joints. Mechanical transmitters in joints have low stiffness and the arms of the manipulator vibrate during a fast motion. A motor/arm-side encoder that detects motor and arm angles directly is useful for precise position control. In this paper, the MDOF manipulator has motor/arm-side encoders. The system with flexible joints has motor dynamics between the input and torsional reaction torque. It deteriorates disturbance rejection by acceleration control. The proposed method comprises a torque controller compensating for motor dynamics and a filter improving disturbance rejection. The proposed method enables position control that maintains vibration suppression and improves robustness by utilizing both the motor and arm angle. The validity of the proposed method was confirmed by simulations.

This paper proposes vibration suppression and disturbance rejection methods for multi-degrees-of-freedom (MDOF) manipulators with flexible joints in the workspace. The proposed methods utilize an equivalent mass matrix (EMM) and equivalent stiffness matrix (ESM) that represent the mass and stiffness in the workspace, respectively. In conventional methods, variations of EMM and ESM can degrade vibration suppression and disturbance rejection performances. Therefore, the proposed methods employ a workspace observer (WOB), resonance ratio control, and load-side workspace observer (LWOB). The WOB and resonance ratio control fix the EMM to constant values and realize stable vibration suppression. The LWOB improves ro-bustness by compensating for disturbances. The stability analyses reveal the parameter conditions. The simulation results show that the proposed methods suppress vibrations and disturbances in the workspace.

This study examines the training policies and environmental robustness of a neural network used in velocity estimation for a tracked vehicle with slippage. In the proposed method, the velocity is estimated by a neural network whose input is an estimated disturbance to the driving axle that includes slippage information. First, we experimentally clarify the proposed method’s scope of applicability and effectiveness under different environmental conditions in training and estimation. Subsequently, we experimentally confirm that the estimated disturbance is robust to environmental changes and complementary to environmental information. Finally, the neural network trained on a flat surface is validated in combination with gravity compensation for acceleration to apply it to driving on a slope.

Lumbar spinal disc herniation is a disease in which the protruding nucleus pulposus presses on the nerve due to actions that place loads on the disc, causing pain in the lower back and lower limbs. About 80% of treatments of disc herniation are conservative treatments, and although it is necessary to live with pain for a long time, there have been no studies that clearly define the relationship between pain and biomechanical parameters. In this study, we proposed a method of identifying biomechanical parameters that predict posture-related pain in patients with lumbar spinal disc herniation. The pain values were quantitatively evaluated by the Numerical Rating Scale (NRS) and the biomechanical parameters were analyzed by OpenSim. Lasso regression was performed to narrow down the biomechanical parameters that were related to pain and derive the mathematical model of the relationship. Therefore, many of the parameters of the obtained mathematical model were related to the lumbar spine and were consistent with areas that be related to lumbar spinal disc herniation.

This paper presents a position estimation compensation method considering dq-axis cross-coupling factors of Interior Permanent Magnet Synchronous Motors (IPMSMs) inductance for position-and-torque-sensorless admittance control. The cross-coupling factors are from dq-axis mutual inductance and deteriorate control and position estimation performance in control methods without the model. Since dq-axis cross-coupling factors of inductance vary with the rotor position and the current, measurement and estimation of the values are difficult. The proposed method estimates the cross-coupling factors and position estimation errors using dq-axis current variation from the injected voltage. In addition, the proposed method compensates the estimated position based on a voltage equation, including the dq-axis cross-coupling factors, by adding the estimated error. As a result, the vibration of current, velocity, and estimated reaction torque were reduced, and the performance of a position-and-torque-sensorless admittance control system was improved. The experimental results show the proposed method's validity and response by the sensorless admittance control.

Proximity and tactile sensors have been widely used to control robots. However, each of these sensors has its own shortcomings. Visual sensors have blind spots, whereas tactile sensors cannot acquire information when non-contact. To overcome these problems, this study develops an optical sensor that can measure the distance and two-axis forces using a beam structure and photo reflectors. The sensor is inexpensive owing to its simple structure. Moreover, the sensor is demonstrated to have sufficient accuracy for both the distance and force measurements at 0.2, 0.5, and 1.0 Hz. Additionally, the usefulness of the sensor is confirmed based on reduced impact control at the time of contact and force control.

Proximity sensors measure the distance to an object and detect contact. Hence, proximity sensors are typically mounted and used in end-effectors. There might be situations where it is necessary to acquire information on not only distance and contact; but also the normal force and shear force. The conventional proximity/biaxial force sensor utilizes beam deflection but brittle material and interference limit its performance. Here we present a novel proximity/decoupled biaxial force sensor made of sturdy materials, springs, and photo reflectors. We demonstrate this sensor's performance and advantage of the proposed sensor in teleoperation and during a task requiring rubbing an object.

Wind energy conversion systems (WECS) have become increasingly common in recent years as a result of global warming. WECS needs to release the wind through pitch control during high winds. In this study, a novel method for individual pitch control (IPC) via pitching moment estimation is designed. The conventional IPC method considers the blade load, without keeping the rotor speed at the rated value. This study proposes the application of a reaction force observer (RFOB) to the pitch actuators to estimate the pitching moment of each blade. Using the pitching moment information, we estimate the wind speed for each blade, enabling IPC to simultaneously achieve load reduction and maintain the rotor speed at the rated value. The effectiveness of the proposed method is verified via simulations and experiments.