Sensory-Motor Learning for Simultaneous Control of Motion and Force: Generating Rubbing Motion against Uneven Object

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.

In this paper substantiates the need to develop a new design of the receiving device to increase the speed of information transfer between motile objects. This is especially true for cases of large distances in poor climatic conditions. In such a situation, the power of the transmitted signal decreases with increasing distance to the receiver. In this case, due to the small signal-to-noise ratio, information may be lost, in addition, it takes longer to process. This reduces the speed of information transfer. In the work, a new method is developed for constructing a receiver for m-ary radio communication system, which allows for the exchange of information between motile objects. Various methods for processing phase-shifted complex signals in various reception options (with a high level of interference, medium and low) are considered. It has been established that the most appropriate is the use of non-optimal incoherent processing of m-ary phase-shifted complex signals in matched filters and correlators. In addition, the work presents the results of studies of various operating modes of the developed receiving device. The directions of further research and development to improve the development characteristics are identified.

Robotic motion generation methods using machine learning have been studied. Bilateral control-based imitation learning can imitate human motions using force information. Using this method, variable speed motion generation that con-siders physical phenomena such as the inertia and friction can be achieved. However, the previous study focused on a simple reciprocating motion. To learn the complex relationship between the force and speed more accurately, it is necessary to learn multiple actions using many joints. In this paper, we propose a variable speed motion generation method for multiple motions. We considered four types of neural network models for the motion generation and determined the best model for multiple motions at variable speeds. Subsequently, we used the best model to evaluate the reproducibility of the task completion time for the input completion time command. The results revealed that the proposed method could change the task completion time according to the specified completion time command in multiple motions.

<strong class="journal-contentHeaderColor">Abstract.</strong> We have investigated the time delay between substorm onset and related reactions in the dawn and dusk ionospheric electrojets, clearly separated from the nightside located substorm current wedge by several hours in MLT. We looked for substorm onsets occurring over Greenland, where the onset was identified by a LANL satellite and DMI magnetometers located on Greenland. With this setup the MARIA magnetometer network was located at dusk, monitoring the eastward electrojet, and the IMAGE chain at dawn, for the westward jet. In the first few minutes following substorm onset, sudden enhancements of the electrojets were identified by looking for rapid changes in magnetograms. These results show that the speed of information transfer between the region of onset and the dawn and dusk ionosphere is very high. A number of events where the reaction seemed to preceed the onset were explained by either unfavorable instrument locations, preventing proper onset timing, or by the inner magnetosphere's reaction to the Earthward fast flows from the near-Earth neutral line model. Case studies with ionospheric coherent (SuperDARN) and incoherent (EISCAT) radars have been performed to see whether a convection-induced electric field or enhanced conductivity is the main agent for the reactions in the electrojets. The results indicate an imposed electric field enhancement.<br><br><b>Key words:</b> Ionosphere (auroral ionosphere; electric fields and currents) - Magnetospheric physics (storms and substorms)

The paper presents the development of a microprocessor-based relay protection device on open architecture. Currently, there is a problem with modern microprocessor relay protection: the impossibility to replace the damaged element with alternatives from other manufacturers. The solution to this problem is the use of devices with open architecture. The study developed a structural model of a microprocessor-based relay protection device based on an open architecture with the industrial Internet of things application. Open architecture is achieved through open protocols and the principle of modularity. The industrial Internet of things technology transfers the control action of triggering blocking. A microprocessor-based relay protection device prototype based on an open architecture was developed. The simulation of the developed device was conducted. The appearance of higher harmonics and aperiodic components in the short-circuit current was not considered during modeling. Due to the study's limitations in the form of lack of load, current and voltage sensors, such as Hall sensors, and inductance coils, the subject of this study is only the speed of operation. A high multiplicity current generation setup was assembled for experimental testing. The developed relay protection device on an open architecture trips faster than the traditional solution. The application of the Internet of Things allowed it to ensure the blocking of non-selective tripping. The obtained results are provided by structural simplification compared to traditional solutions and speed of information transfer with the application of the Internet of things. The developed open architecture device with the industrial Internet of things technology application gives new possibilities for relay protection systems, including flexibility to meet the requirements in connection with the introduction of distributed power

Similar rate of information transfer on stimulus intensity in accessory and main olfactory bulb output neurons

This paper describes quick and stable standing and sitting motion of an inverted pendulum type robot by whole-body motion including force control. The whole-body motion is achieved by controlling composite center ofgravity of the robot, and damping force control of a wrist-roller is employed to assure the stability of invertedpendulum control on a contact situation with the ground. Motion planning of standing and sitting based on theproposed control method is also proposed. Standing and sitting motion using the proposed motion control andmotion planning is simulated with a dynamic simulator, ODE(Open Dynamics Engine). From the simulationresults, it was con“rmed that the robot successively realized the standing and sitting motion quickly and stably.Keywords: Inverted Pendulum, Assist Robot, Standing and Sitting, Force Control 1. INTRODUCTION A robot with an inverted pendulum mobile platform possesses suitable features as a human supporting robot;it has high work capability which can use its own weight and high mobility coming from a mechanically smallfootprint and a casters needlessness, and a weight saveing is also easy for it because of a simple structure.However, many studies on a wheeled inverted pendulum type robot have focused on improving its mobility, suchas running, steering, and trajectory following

We proposed a fault-tolerant control method based on machine vision to enhance the reliability ofthe robotic arm in the radiation environment. Firstly, we introduced the joint angle visual detection method as an angle detection backup solution, which was used to timely detect the joint angle of the robotic arm in case of encoder fault. Next, we introduced an encoder fault diagnosis mechanism, and the difference between the machine visual feedback value and corresponding encoder feedback values was used as the basis of encoder fault. When the encoder failed, it could be switched to the joint angle visual detection in real time. With the Canny edge detection and the Probabilistic Hough transform, the method could extract the identification tag of robotic arm joint image from the front-end camera, and hence the joint angle value was acquired. Finally, in the fault-tolerance experiment, the proposed approach found the encoder failure and switched detection method, and the scheduled task continued to the end. The error about the angle detection based on the machine vision would not exceed ± 0.008.

Generation of natural human motion is one of key techniques for multimodal dialogue systems with a human-like avatar. In particular, natural and expressive lip motion synthesis is necessary to make conversation between a user and an avatar richer. However, such expressive lip motion is often difficult to be generated automatically because it can be changed depending on phonemic context and prosody. To address this difficulty, we introduce a novel motion generation method on the basis of the modulation of a set of dynamic models learned from neutral motion data. As a suitable model for lip motion generation, we adopt a hybrid dynamical system, which consists of linear dynamical systems for each motion unit and a symbolic automaton for switching between these units. We show that, from the viewpoint of control theory, it is possible to modulate linear dynamical systems for various types of motion. Early results demonstrate the applicability of the proposed method using lip motion synthesis for simple phoneme sequences.

The gleno-humeral (GH) rotation centre is typically estimated using predictive or functional methods, however these methods may lead to location errors. This study aimed at determining a location error threshold above which statistically significant changes in the values of kinematic and kinetic GH parameters occur. The secondary aims were to quantify the effects of the direction of mislocation (X, Y or Z axis) of the GH rotation centre on GH kinematic and kinetic parameters.Shoulder flexion and abduction movements of 11 healthy volunteers were recorded using a standard motion capture system (Vicon, Oxford Metrics Ltd, Oxford, UK), then GH kinematic and kinetic parameters were computed. The true position of the GH rotation centre was determined using a low dose x-ray scanner (EOS™ imaging, France) and this position was transferred to the motion data. GH angles and moments were re-computed for each position of the GH rotation centre after errors of up to ± 20 mm were added in increments of ± 5 mm to each axis. The three-dimensional error range was 5 mm to 34.65 mm.GH joint angle and moment values were significantly altered from 10 mm of three-dimensional error, and from 5 mm of error on individual axes. However, errors on the longitudinal and antero-posterior axes only caused very small alterations of GH joint angle and moment values respectively. Future research should develop methods of GH rotation centre estimation that produce three-dimensional location errors of less than 10 mm to reduce error propagation on GH kinematics and kinetics.

Temperature sensors are widely used in aero gas turbine engines for measurement of air and gas temperature at various locations starting from inlet of fan to exhaust gas from the nozzle. Thermocouples are one of the most widely used sensors for the above purpose. Time constant of thermocouple is a key parameter in the selection of sensor for a particular application. Even though time constant is available for standard stand alone thermocouple configurations, it varies widely based on many physical parameters and inlet flow conditions inside a gas turbine engine. Experimental determination of time constant of stand alone thermocouples with step input is carried out for various thermocouple configurations and a Fibre Bragg Grating sensor using a dry block calibrator with hot air as medium. Two different methods are used to determine the time constant of thermocouple sensors. Further determination of time constant of thermocouple assembled in a rake and tested in real time situation in aero gas turbine engine is carried out. A comparative study reveals that many parameters are involved which influence the value of time constant under different input conditions. Finally the selection of sensors is carried out based on the measurement requirements.

This study focuses on the three-dimensional dynamic analysis of joint angles and the prediction of sports injuries in long-distance runners, addressing the significant limitations of traditional approaches in this field. The research is grounded in the fact that millions of athletes participate in long-distance running events worldwide each year, with nearly 30% experiencing joint-related injuries. Traditional joint angle analysis methods exhibit error rates as high as 15%, and injury prediction accuracy remains limited to around 30%. To overcome these challenges, a novel deep learning-based model was developed, utilizing the public Human3.6M dataset (comprising 500 samples of long-distance running motion) and 300 additional samples collected from 30 professional club athletes. The model integrates a customized feature extraction module, a joint angle encoding component based on spherical harmonics, a temporal dynamics module, and a probabilistic injury prediction mechanism. Experimental results demonstrate that the average joint angle analysis error was reduced to 5.2%, while injury prediction accuracy improved to 75%. Our model adopts a modular deep learning architecture consisting of a feature extraction module with custom kernel functions, a joint angle encoding component based on spherical harmonics, a temporal dynamics modeling component leveraging non-stationary temporal kernels, and a final injury prediction component using a Gaussian Mixture Model integrated with Bayesian inference. Evaluation metrics include joint angle analysis error rate, injury prediction accuracy, precision, recall, and F1 score. On joint angle analysis, the model achieved an average error rate of 5.2%, significantly outperforming the 14.8% of the 3D-Traditional baseline and the 12.3% of the CNN-2D baseline. For injury prediction, the model reached an accuracy of 75%, compared to 35% for the ML-Injury model and 50% for the Simple-DL model. Precision and recall reached 78% and 72% respectively, indicating the model’s superior predictive performance across multiple evaluation dimensions.

Work-related Musculoskeletal Disorders (WMSDs) account for about 34% of non-fatal injuries resulting in days away from work in the construction industry. Particularly in modular construction, due to the repetitive nature of the manual tasks, workers are highly exposed to ergonomic risks. To identify workers’ awkward postures that potentially lead to WMSDs, ergonomic assessment tools have been developed and widely used by ergonomists and Occupational Health and Safety (OHS) practitioners. However, the accuracy of these methods is highly affected by the subjectiveness towards the user’s inputs (e.g., body joint angles), which are difficult to accurately determine in given observation time. Consequently, the imprecise estimates of worker postures may result in inaccurate final results and risk intervention plans. In an effort to address this issue, this study applies fuzzy logic techniques to ergonomic evaluation tools—e.g., Rapid Upper Limb Assessment (RULA). By modelling the range of input values using fuzzy sets rather than discrete boundaries, the imprecision inherit in the inputs has less impact on the final RULA score. As a result, an automated fuzzy expert system has been developed by using membership functions and rules created based on the existing RULA method. An experiment is carried out in order to study the amount of imprecision in joint angle values from human estimations while observing a posture, and also to compare the sensitivity of RULA and the developed fuzzy RULA system to input imprecision. The results reveal that although the fuzzy RULA model has high correlation with RULA, it is more accurate and less sensitive to the variance in input values. The developed model presents a methodology to improve the accuracy of ergonomic assessment methods and handle the uncertainty inherent in ergonomic evaluation, providing the construction industry practitioners with an automated technique to evaluate the ergonomic safety of workers.

Deep learning is a sub-field of machine learning with their models being vaguely inspired by the communication patterns and information processing of a nervous system. In a deep learning model each layer of the network, which is build up of multiple neurons, transforms its input data into slightly more abstract and composite abstractions. Within a convolutional neural network (CNN), the first layer of neurons may be transform its raw input pixels to encoded edges, with the next layer encoding arrangements of edges, with the next layer more abstract features such as eyes and noses. However, locating which exact neurons in each layer are looking for each abstract feature is still a hot area for research. Here we show a novel way to pinpoint each neuron in the hidden layers that is searching for a specific feature that is believed to play a pivotal role in the classification task. We found that by removing the feature of interest from the original image and running this through the CNN, each neurons activation map can be compared to the activation generated by the original image using t-SNE. This allowed us to successfully located the individual neurons in the network that are significantly effected by the removal of this feature and hence changing the classification. This method allows us to get a better understanding of what the network has learnt and how important this learnt information is when coming to the final classification.

In this paper, we apply Reinforcement Learning (RL) to control a manipulator using camera images. Basically, RL algorithm helps the agent to choose actions under some specific policies, which maximize rewards accumulated from a sequential result of the actions. In order to control a manipulator, the agent generally considers joint angles or torques of the manipulator as the states. Deep learning has enabled to deal with a very high dimensional state space so that raw images can be considered as states. However, using raw images as state spaces can require extensive computational resources because of their high dimension, so there has been much research using additory algorithms such as Convolutional Neural Network (CNN) or autoencoder to extract features from images. While those algorithms reduce the computational load, it is still a quite complicated process. In this paper we consider a task for the end-effector to reach a random target and propose a new approach using a vision-based direction vector, which has low dimension and can be simply implemented. We calculate the direction vectors via the camera and exploit them as the states of the policy. Then, based on the direction vectors, the manipulator can learn how to control each joint. For the simulation, we demonstrate a multiple degree of freedom(dof) manipulator of a commercial robot, and the simulation results show that the manipulator successfully tracks the target.