Adaptive effective drive for robotics should have small sizes and weight and self-regulation possibility at origination of non-staff situations. Automatic gear-boxes existing now (CVT) are the most complicated mechanical systems and have the most complicated hydro mechanical control system. Such transmissions are absolutely unsuitable for robotics. Recently adaptive gear (toothed) variators are developed. The adaptive variator represents the self-controlled gear planetary train with constant engagement of the toothed wheels, created on the basis of an author's discovery «Effect of force adaptation in mechanics». The adaptive gear box of the car represents brand new planetary gear variator in the form of two mobile kinematic chain with the elementary additional original constraints. The adaptive variator has ability to drive the executive tool with a speed which is back - proportional external loading at constant power of an engine. The basic advantages of an adaptive gear variator: simplicity of a design, absence of the control system, full adequacy to working conditions. It is expedient to apply an adaptive gear variator in transmission of robot. Theoretical bases of creation of adaptive transmission on basis of using of toothed variator are presented.
Konstantin S. Ivanov has completed his Candidate of Technics from Moskow Technical University, Russia, and D.E. from Kazakh National University, Kazakhstan. He is the Professor of Almaty University of Power Engineerung and Telecommunication, Kazakhstan. He has published more than 125 papers in reputed journals and has been serving as an editorial board member of repute.
Leading experts around the world analyzed geophysical images daily by and with the development of computer vision technologies, attempts should be made to automate this process. Image data can be acquired quickly using consumer digital cameras, or potentially using more advanced systems, such as satellite imagery, sonar systems, and crewless aerial vehicles. The authors of this article have developed several approaches to the automatic creation of seismic images. The amount of obtained images became enough to use algorithms of machine learning for their processing. In the last five years, computer vision techniques have evolved at a high rate and have advanced far from the use of Deep Neural Networks (DNN). It would be reckless to use in work only the latest developments without understanding how they appeared. Therefore, the authors reviewed the approaches of computer vision to determine the most appropriate techniques for processing high spatial images that differ from the most popular tasks of computer vision (face recognition, detection of pedestrians on the street, etc.). Geophysical high-resolution images serve as a source for obtaining new information using computer vision. The authors identified two areas for further research that deserve attention: • Detection of geological objects. Determination of their shape, relative position, and volumetric characteristics, • Determination of the content of geological objects. Energy characteristics of geological objects. These areas of research are not fundamentally new. Nevertheless, with the advent of modern tools and techniques for working with high-resolution images, it is advisable to re-evaluate their capabilities. The initial research hypotheses that the authors have accepted for themselves in subsequent works are as follows: Do the methods of computer vision allow creating a fully automated process for the identification of geological objects by seismic volume? The main result of the paper is to develop research hypothesis for computer vision in geoscience.
Networks informational technologies caused mammoth growth of knowledge and serious problems of scientific big data (SBD) processing attendant innovations. Their solution lies in Internet records semantic filing by artificial intelligence (AI) co-processing. Deep-learned AI could assist unprepared natural intelligence (NI) to investigate meanings combining filtration, integrity study, and knowledge compression. Cognogenesis difficulties are seriously aggravated by necessity to understand philogenetic achievements hidden in constantly growing SBD flow. World sophistication accompanies up-to-day educational crisis. The part and parcel of system-informational culture is inter-disciplinary activity taking place in computer systems needing man’s universal tutoring supported by AI and impossible without its assistance. Rational consciousness auto-building results from it. Sine qua non of successful work with complex meanings becomes lifelong mutual NI and AI tutoring reinforcement learning being unable to cope with arising intellectual intricacies. Partnership of NI and AI becomes basis of up-to-day cognitive revolution. Neuroscience shows that man is able to grow, use, and identify scientific meanings. This investigation objective is to contribute AI implementation technology supporting man’s intellectual breaks and semantic consciousness auto-molding on the ground of language of categories (LC) mathesis universalis. System axiomatic method allows NI and AI interface managed in LC. Knowledge comprehension is proved to be attained on the way. Its auto-obviousness issues from real-ideal presentations discoordination mastering. Knowledge universal core is to be presented in the form of the utmost mathematical abstractions. This is cogno-ontological (Co-Ont) knowledge base. It organizes environment for person’s trained semantic existence. Trance-disciplinary meanings identification and extraction occur by Co-Ont application.
Nicolay Vasilyev has completed his PhD from Lomonosov Moscow State University, Russia, and postdoctoral studies from USSR Academy of Sciences (AS). He worked in Russian Academy of Sciences. Since 1999 he has been working in different universities of Moscow, Russia. He has published more than 20 papers in reputed journals. His fields of interests are optimization, control and games theories, networks routing modeling, universal algebra, and artificial intelligence.
Leaves account for the largest proportion of all organ areas for most kinds of plants, and they are also the main part of the photosynthetically active material in a plant. The observation on individual leaves can help recognizing the growth status and measuring complex phenotypic traits. Current image-based leaf segmentation methods have problems such as highly restricted species and the vulnerability toward canopy occlusion. In this work, we propose an individual leaf segmentation approach for dense plant point clouds using facet over-segmentation and facet region growing. The approach can be divided into three steps: (1) point cloud preprocessing; (2) facet over-segmentation, and (3) facet region growing for individual leaf segmentation. The experimental results show that the proposed method is effective and efficient in segmenting individual leaves from 3D point clouds of greenhouse ornamentals such as Epipremnum aureum, Monstera deliciosa, and Calathea makoyana, and the average precision and recall are both above 90%. The results also reveal the wide applicability of the proposed methodology for point clouds scanned from different kinds of 3D imaging systems such as stereo vision and Kinect v2. Moreover, our method is potentially applicable in a broad range of applications that aim at segmenting regular surfaces and objects from a point cloud.
Dawei Li received the Bac¬¬¬¬¬¬¬¬helor of Engineering degree in Automation in 2006 from Tongji University, Shanghai, China. He received his PhD degree from Tongji University in January, 2013. During 2013-2015, he worked in the Department of Computer Science and Technology Department as a postdoc. In 2015, he became a lecturer with Donghua University, Shanghai, China. He became an associate professor with Donghua University in 2018. His current research interests include image processing, computer vision, and plant phenotyping.
Many reference tracking systems involve the interaction of controlled devices with human operators, where a command center indicates a reference path to a human agent. The performance of such systems depends not only on the controlled devices, but also on the ability of the human to operate such devices. An important challenge in this scenario is to take corrective actions when errors occur during the operation process in an online and non-invasive fashion. There is a need to develop strategies that learn from unmodeled errors during the operation of the system to take non-invasive corrective actions that minimize the tracking errors. In this work, we propose a strategy in which the command center learns from the behavior of the human agent and the performance of the tracking system, and is able to instruct him/her to follow a modified reference path such that the resultant path is as close as possible to the ideal one. In this way, the modified reference path is designed to account for uncertainties and imperfections due to the human agent and the devices during the operation process. We propose a fully online and autonomous command center based on deep reinforcement learning to find the optimal policy to be shown to the human agent such that the error between the system and the ideal paths is minimized. We show through simulations the applicability of reinforcement learning techniques in human-closed loop systems involving scenarios with a high degree of uncertainty and imperfect human agents.
Maria Arroyo is an electrical and electronics engineer with a minor in Technology Innovation from the University of los Andes, Colombia. Nowadays, she is completing her MSE at the same university. She was teaching Assistant of Stochastic Processes, and also a Research Assistant, who organized ActInSpace Colombia in collaboration with CNES and the European Space Agency. She had worked for UTC Aerospace Systems, now Collings Aerospace in AZ, USA and is currently working at Teknia Group, in Nagoya, Japan. She received an award for academic excellence at University of los Andes. Email: email@example.com
The electrocardiogram (ECG) is the oldest and most enduring tool for the clinicians to diagnose diseases related to the heart. The standard ECG uses 10 cables to obtain 12 leads reflecting the angles at which electrodes ``look’’ at the heart and the direction of the heart's electrical depolarization. The ECG trace in a lead comprises three waves, P, QRS complex, and T. Possible diseases can be effectively revealed by investigating the locations, shapes, or sizes of these waves on the 12 leads. A normal ECG has only very small Q waves. However, a downward deflection immediately following a P wave that is wider than two small squares or greater in height than a third of the subsequent R wave is significant: such Q waves can represent previous infarction. Bundle branch block can be diagnosed by looking into the QRS complexes in the V1 and V6 leads. Right ventricular hypertrophy is indicated by a dominant R wave in the V1 lead and right axis deviation. However, manual inspection on the 12 leads by clinicians is inefficient and could lead to misjudgments. In this paper, we propose a pattern-recognition based system that can detect and locate automatically these different waves in the ECGs. The system starts with digitizing the ECGs into individual time series of data. Then by detecting the peaks and troughs in a given time series, the P, Q, R, S and T waves can be identified and located in this series. Based on the results, an expert system or a neural network model will be constructed as a medical assistant to be used in hospitals or clinics. We believe our research can help doctors provide more accurate and efficient diagnosis for heart-related diseases.The electrocardiogram (ECG) is the oldest and most enduring tool for the clinicians to diagnose diseases related to the heart. The standard ECG uses 10 cables to obtain 12 leads reflecting the angles at which electrodes ``look’’ at the heart and the direction of the heart's electrical depolarization. The ECG trace in a lead comprises three waves, P, QRS complex, and T. Possible diseases can be effectively revealed by investigating the locations, shapes, or sizes of these waves on the 12 leads. A normal ECG has only very small Q waves. However, a downward deflection immediately following a P wave that is wider than two small squares or greater in height than a third of the subsequent R wave is significant: such Q waves can represent previous infarction. Bundle branch block can be diagnosed by looking into the QRS complexes in the V1 and V6 leads. Right ventricular hypertrophy is indicated by a dominant R wave in the V1 lead and right axis deviation. However, manual inspection on the 12 leads by clinicians is inefficient and could lead to misjudgments. In this paper, we propose a pattern-recognition based system that can detect and locate automatically these different waves in the ECGs. The system starts with digitizing the ECGs into individual time series of data. Then by detecting the peaks and troughs in a given time series, the P, Q, R, S and T waves can be identified and located in this series. Based on the results, an expert system or a neural network model will be constructed as a medical assistant to be used in hospitals or clinics. We believe our research can help doctors provide more accurate and efficient diagnosis for heart-related diseases.