In this study, in order to improve the accuracy of eccentric magnetic absolute encoders (EMAE), an angle calculation method that combines radial basis function neural networks (RBFNN) and third-order phase-locked loops (TOPLL) is proposed. EMAE consists of a simple multipole magnet, but the eccentricity causes periodic distortion in the signals, which degrades accuracy. In addition, the accuracy of magnetic encoders is degraded by nonideal components, such as amplitude mismatches, phase shifts, low- and high-order harmonics, random noise, and DC offsets. The proposed method uses amplitude normalization and RBFNN to reduce these disturbances. Moreover, the steady-state error that occurs when the ramp frequency is input is eliminated by using TO-PLL. The effectiveness of the proposed method was demonstrated through a simulation.
This paper proposes a high-response torque control method for a surface permanent magnet synchronous motor (SPMSM) with reduction gearing for interactive operation. In robots performing interactive tasks, backforward-drivability of the geared SPMSM is important. The backforward-drivability is defined as the ease of manipulating the load-side of the geared motor by external forces. Torsion torque control has been proposed to enhance this drivability. However, conventional torsion torque control is limited in response speed due to the high system order and reliance on a minor loop current controller. In this study, we propose a high-response sliding mode direct torsion torque control for the geared SPMSM modeled as a two-inertia resonant system. The proposed method achieves rapid response by eliminating the current controller and controlling inverter switches directly. Numerical simulations and experiments confirm that the proposed method provides high-response torsion torque control.
Two-inertia resonant systems are used to represent several motion control systems with resonance modes such as drivetrains and robot joints, offering a more flexible approach for their control compared to single-inertia representations. Controlling the transmitted torsion torque (also known as joint torque) to the load allows to effectively control the driven load of the system regardless of the resonant mode, which can be done by employing torque sensors. However, many real systems contain some degree of joint backlash, which causes a large, impulse-like shock on torque reversals due to the mechanical play of the joint elements, which can lead to undesired behavior and even destruction of the mechanical elements of the system being controlled. While torque sensors are able to sense whether the drive and load-sides are in contact or not, they don't give information on the relative angle between both sides, which is necessary for achieving shockless torque control. This paper introduces a novel nonlinear observer structure based on torque sensing that is able to estimate all the load-side states even when backlash occurs. This new observer is then implemented in a torsion torque control scheme that employs model predictive control during backlash gap traversings to achieve shockless control. Simulation results confirm the effectiveness of this method.
The variation in road conditions and vehicle speeds poses significant challenges in maintaining the desired stability and tracking accuracy of Autonomous Ground Vehicles (AGVs). Uneven, wet, or slippery road surfaces reduce frictional properties, while larger speed variations generate multidirectional motions within the vehicle, ultimately degrading its stability and tracking performance. Additionally, parametric uncertainties, such as cornering stiffness, affect the lateral dynamics and further compromise the tracking performance and stability of the vehicle. To address these challenges, this paper presents a polytopic Linear Parameter Varying (LPV)-based Linear Quadratic Regulator (LQR) controller. The polytopic LPV framework is introduced to handle variations and uncertainties in AGV parameters and assisting the LQR in minimising errors and ensuring yaw stability effectively. The performance of the polytopic LPV-LQR controller is analysed during single-lane and double-lane change manoeuvres considering varying AGV speeds and road conditions with uncertainty. Additionally, the proposed controller's performance is compared with a Robust Controller to provide further insights based on different metrics such as lateral position error, heading angle error, slip angle, and yaw rate. The results show that the proposed LPV-LQR controller is more capable of managing varying and uncertain parameters than the robust controller, facilitating satisfactory AGV tracking performance and stability at higher AGVs speed.
In the literature, fast or real-time magnetic tracking is conventionally achieved by using optimization or machine learning, which requires intensive computation, large computer memory usage and/or tedious data preparation. This could result in high cost and low adaptability of the tracking systems. To address these issues, a combined approach of using weighted coefficients and Monte Carlo algorithm is put forward and analyzed in this paper. The analysis is based on simulating the scenario of tracking a spherical magnetic marker with a diameter of 15 mm within a 200x200x200 mm(3) cuboid region, which is suitable for capsule endoscopy. The statistics of 10,000 case studies show that the approach achieves, averagely, 91.84% position tracking accuracy and 93.97% orientation tracking accuracy with a tracking frequency of 909.1 Hertz.
In orthopedic surgery, robotic technologies to assist surgeons operating surgical drills to cut bones are being de-veloped, since there is a risk to injure delicate tissues around bones. Penetration detection is useful for surgical drills to enhance safety. However, there is still a risk to damage delicate tissues, even if penetration is detected to stop drills. In this study, a teleoperated haptic drill system implementing a safety enhancement system which extends haptic sensation according to the end effector position is presented. The presented method is composed of the object detection phase and the operator assist phase. In the object detection phase, the position and posture of an object are acquired by the drill system which has the shape model of the object. In the operator assist phase, haptic sensation is extended based on the acquired object information to prevent penetration of the object, while the state of the controller of the drill system is visualized by CG images created by a cutting simulator. The feasibility of the method is confirmed by cutting a pseudo bone.
Multi-input multi-output (MIMO) control systems are common in many industrial applications, such as reel-to-reel (R2R) systems, process control in distillation columns, robotic manipulators, and flight control. The coupling between plant inputs and outputs makes controller design challenging, especially when selecting controller gains in transfer function-based approach. This paper proposes a simple yet practical method for designing the controller using a decentralized approach based on transfer functions, along with an intuitive gain-tuning procedure. Unlike state-of-the-art methods, this approach does not require solving complex optimization problems. It is also generalizable to any MIMO control system. The controller design approach simplifies the MIMO system by shaping the open-loop transfer function into a triangular form, allowing it to be treated as a set of individual SISO systems. Each SISO system is influenced by a specific level of internal disturbance. These disturbances are associated with the dynamics of the actuators controlling the control variables, and therefore can be modeled based on the knowledge of the actuator dynamics. This makes controller design and gain tuning simpler compared to the complex algorithms used in the literature. The effectiveness of the proposed approach is demonstrated by its implementation on a 2x2 R2R system hardware. Finally, the stability and robustness of the designed control system are experimentally assessed and various stability margins are obtained.
This paper proposes an advanced disturbance-suppression strategy for a dual-stage actuated 3-degree-of-freedom (3-DoF) positioning system for high-precise optical 3D inline metrology applications. The decentralized control loop based on single input single output (SISO) PID controllers is complemented by a disturbance observer (DOB) in each DoF to suppress disturbance forces, originating from the coarse positioning unit at fractions of the pitch of the mechanical spindle drive. By reformulating the DOB design problem in the H-infinity-synthesis framework, the disturbance sensitivity of the closed-loop system is shaped to suppress disturbances at these specific frequencies. Experiments on an experimental prototype demonstrate the effectiveness of the approach with up to 28% reduction of measurement errors related to the positioning system on a large inspection area (0.7x0.6m).
Upper limb assistive robots have been widely developed due to their ability to aid user movement and reduce caregiver workload. In our previous work, we introduced a wearable end-effector-type upper limb assistive robot, WELiBot. To enhance the robot's portability and reduce the required actuation torque, this work investigates gravity compensation for WELiBot's mechanism using springs. Following the 1-DOF gravity compensation principle, springs were designed for lifting and reaching movement respectively. The effectiveness of this compensation effect was then evaluated by comparing the necessary torque at the actuating joints with and without springs. This work offers insight into gravity compensation for parallel mechanisms and contributes to the future development of WELiBot.
In various industrial mechatronic equipment for processing and assembling electronic components, high-acceleration drives in the internal positioning mechanisms cause machine stand vibrations, degrading their production accuracy and throughput. Generally, a trade-off exists between the fast and precise control of positioning mechanisms and suppressing machine stand vibrations, which complicates control design. This study presents an extended data-driven vibration suppression feedforward (FF) control method that addresses both positioning performance and machine stand vibrations. The proposed method designs FF controllers to pursue the trade-off through a multi-objective optimization problem using predicted responses of both the positioning device and the machine stand. Experiments with a laboratory table positioning system demonstrate the effectiveness of the proposed method, compared to an existing data-driven vibration suppression FF control method.
Membrane deployment structures are used in spacecrafts due to their lightweight nature and high stowage efficiency. Currently, numerous analytical methods have been proposed to model the behavior of such flexible structures. However, the complex dynamics of membrane deployment in space lead to high computational loads in conventional analysis methods. Thus, there is a need to develop a low-cost, computationally efficient method that can accurately reproduce membrane deployment dynamics. This study proposes an analytical approach applying the Nonsmooth DEM to analyze membrane motion modeled by particle systems, aiming to reduce computational cost. To evaluate this method, both simulation and ground-based experiment were conducted using a mesh structure that simulates flexible membrane. Comparison of experimental and simulation results indicated that the proposed method was validated as an approach for analyzing membrane deployment structures.
Magnetic levitation planar motor can move at high speed and with high accuracy, and prevent vibration and noise because of no friction. This technology leads to the application of a transport system that requires high-speed and high-precision movement. However, the design of appropriate control algorithms is essential to achieve stability and high-speed, high-precision control of the maglev planar motor. In this research, we stabilized a maglev planar motor on the basis of system identification. Initially, we designed a whole control system for the movement of the maglev planar motor. Furthermore, we conducted system identification and verified that the maglev planar motor works with the designed control system and evaluated its motion performance. Through experiments, it was confirmed that accurate motion control was possible. This research is expected to be a basis for improving control performance soon.
High-performance flexible sensors characterized by high sensitivity, low detection limits, a wide working pressure range, and rapid response times have gained significant attention recently due to their essential role in advancing wearable smart devices, human-machine interaction, and healthcare systems. In this work, we present the fabrication of highly flexible and sensitive capacitive pressure sensors using Multi-Walled Carbon Nanotube (MWCNT)/Ecoflex composite electrodes. The sensors were fabricated using a straightforward dipping-drying method, facilitating the uniform adhesion of CNTs to the elastomer matrix. The resulting pressure sensors demonstrate exceptional performance, exhibiting high sensitivity (2.2% kPa(-1)), fast response time (< 40 ms), and an extensive working range (0.03-30 kPa), with strong linearity (r(2) = 0.983). Several applications of human machine interaction are successfully demonstrated by using the developed sensors. This work presents a cost-effective and simple way to fabricate flexible pressure sensors for use as wearable devices across a range of applications, including motion monitoring, hand gesture detection, and human-machine interfaces (HMIs).
Myoelectric hand gesture recognition is an effective and promising strategy for controlling prosthetic hands. Currently, most commercial systems rely on manual feature extraction, which is based on prior understanding of surface electromyography (sEMG) signals. While effective, this traditional method is limited by its dependence on handcrafted features. With recent advancements in artificial intelligence, deep learning techniques have been explored for myoelectric gesture recognition, offering the potential for more abstract and automated feature extraction. In this paper, we propose a new model called the Hybrid Multi-stream Convolutional Neural Network-Bidirectional Long Short-Term Memory (Multi-stream CNN-Bidir.LSTM) for myoelectric hand gesture recognition. Our model achieved state-of-the-art performance, with 89.03% accuracy in within-session testing and 40.66% accuracy in cross-day uncalibrated testing. Additionally, we demonstrate the importance of incorporating linear decision boundaries, proper feature selection, and deep model architecture design to optimize recognition performance.
This article investigates the feasibility of using a dual-cantilever atomic force microscope for imaging the electrical properties of a sample below the surface. Principles from macro-scale electrical resistive tomography are adapted to utilize measurements from a dual-probe atomic force microscope. A deep-learning method is employed to perform the inversion process and construct the tomography. Simulation results demonstrate that electrical resistive tomography is possible at the nanometre scale but improvements to the inversion algorithms are needed before moving to experimental applications.
In recent decades, the agricultural labor force in developed countries has steadily declined, leading to severe shortages of workers. To address this problem, automating various agricultural tasks has become increasingly urgent. This paper focuses specifically on the robotic automation of fruit harvesting. Existing automated harvesting robot systems have several limitations: They lack flexibility as they target only specific fruits, require large batteries because of high power consumption, which reduces operating times, and incur high costs. To reduce power consumption and overall costs, the proposed system uses a single RGB-D sensor and a Raspberry Pi as its sole computational resource. To ensure flexibility, the system utilizes shared peduncle characteristics among different fruits. After identifying the fruit to be harvested, the system locates the connecting peduncle and determines the target cutting pose. The experimental results indicated an average success rate of 91% and an average time of 12.7 seconds to harvest each fruit.
In motion control, velocity is required to be estimated with less delay and less error, but the two are trade-offs because of an encoder's quantization noise, and they are difficult to make both small simultaneously. This paper aims to overcome the trade-off when conventional methods are used. This paper proposes an FPGA (Field Programmable Gate Array) implementation method for a model-free velocity estimation using Chebyshev expansion. The use of Chebyshev expansion enables better rejection of quantization noise from the measured data than using a frequency-domain filter. The estimation method is realized by a single inner product calculation, which takes 3 mu s of computation time on FPGA. The performance of the proposed method is validated by a velocity estimation from actual encoder data.
The rapid expansion of the semiconductor industry and the increasing demand for high speed, high precision positioning stages necessitate exceptionally precise active control of pneumatic vibration isolation tables. This study addresses these stringent requirements by employing a novel pneumatic valve to regulate airflow to the air springs of the vibration isolator. The novel valve incorporates an internal sensor for accurate poppet position monitoring which enables inner loop feedback (FB) control of the poppet position. This effectively compensates for airflow induced disturbances acting on the poppet. Consequently, rapid and precise air spring pressure regulation is achieved which results in accurate displacement control of the vibration isolator. The effectiveness of the proposed system is validated through experiments evaluating its ability to isolate vibration caused from a movable stage mounted on the isolator.
This review introduces the trends of the global drone industry and the manufactures, the forecast of drone industry, the use case, the ranking of global drone performances including AAM (Advanced Air Mobility).
This paper presents the practical implementation of an adaptive control strategy for a motion system subject to disturbances primarily caused by the fundamental frequency of the periodic reference signal. To this end, we devise a model for the motion platform which adequately describes the rigid-body dynamics. This physical model has been validated using measured data in the frequency domain to illustrate that it adequately captures the real dynamics of the motion platform. Using this model, we derive a standard motion control architecture with an adaptive augmentation. The control strategy consists of a feedforward controller based on model inversion and a baseline feedback controller for regulation plus an adaptive component resorting to L-1 control theory. The adaptive term is responsible for the rapid rejection of harmonic disturbances while recovering the nominal performance even in the presence of parametric perturbations. The effectiveness of the proposed control strategy is verified through real-time experiments featuring outstanding fast adaptation and rapid disturbance rejection.
