26 Ноября 2021

International AI and robotics workshop took place at the Financial University: Robot controlling, computer vision, and mathematical problems in cryptography

On November 24, the Faculty of Information Technology and Big Data Analysis held an international workshop on artificial intelligence and robotics. More than ten researchers and scientists from China, Taiwan, Italy, Portugal, Poland, Russia, Kazakhstan, Tajikistan, and Uzbekistan discussed a wide range of research topics, such as computer vision technologies, algorithms for controlling groups of robots, algebraic equations in cryptographic analysis, and decision-making methods in alternative energy. The event came as a result of close cooperation between scientific staff of the Department of Data Analysis and Machine Learning of the Financial University under the Government of the Russian Federation and foreign researchers participating in the Visiting Professor Program. The Visiting Professor Program is aimed at employing foreign experts to share his/her research experience, deliver lectures and conduct seminars to the Financial University students, as well as implement joint research projects and prepare scientific publications in cooperation with the hosting department/faculty. Some of the joint research have already found ground in scientific articles published in highly rated international journals: 1. Krakhmalev, O.; Korchagin, S.; Pleshakova, E.; Nikitin, P.; Tsibizova, O.; Sycheva, I.; Liang, K.; Serdechnyy, D.; Gataullin, S.; Krakhmalev, N. Parallel Computational Algorithm for Object-Oriented Modeling of Manipulation Robots. Mathematics 2021, 9, 2886. https://doi.org/10.3390/math9222886 2. Krakhmalev, O.; Krakhmalev, N.; Gataullin, S.; Makarenko, I.; Nikitin, P.; Serdechnyy, D.; Liang, K.; Korchagin, S. Mathematics Model for 6-DOF Joints Manipulation Robots. Mathematics 2021, 9, 2828. https://doi.org/10.3390/math9212828 3. Korchagin, S.A.; Gataullin, S.T.; Osipov, A.V.; Smirnov, M.V.; Suvorov, S.V.; Serdechnyi, D.V.; Bublikov, K.V. Development of an Optimal Algorithm for Detecting Damaged and Diseased Potato Tubers Moving along a Conveyor Belt Using Computer Vision Systems. Agronomy 2021, 11, 1980. https://doi.org/10.3390/agronomy11101980 On the part of hosting Department of Data Analysis and Machine Learning, the following scientists took part in joint researches: Associate Professors Oleg N. Krakhmalev, Georgy V. Moiseev, Sergey A. Korchagin, Viktoria V. Shamraeva, Vera A. Ivanyuk, Nikita A. Andriyanov, Eldar F. Boltachev, Alexey V. Osipov, Sergey T. Gataullin, Galina N. Kamyshova, Dmitry O. Hort and department assistant Dostonzhon N. Barotov. Huo Jianwen, Associate Professor at Southwestern University of Science and Technology (China) opened the workshop presenting the results of a research work on a motion control system for a group of mobile robots based on visual information from an accompanying drone. He noted the shortcomings of already existing algorithms, such as the inability to take into account dynamic obstacles that are not present on the map, high computational complexity and low stability of group configuration management. Based on the results of field experiments, his modified algorithm showed an acceleration of the path search process by one and a half times compared to previous algorithms. The dynamic window method and computer simulation of constructing a programmed trajectory based on map matching in the ROS environment facilitated the process of planning the trajectory of a mobile robot. The new control system for rebuilding a group of robots demonstrated to be of easy implementation, high flexibility and low computational complexity. The modeling confirmed that the algorithm is more effective for searching the best trajectory when encountering obstacles that are absent on the global map. The developed algorithm is designed to simplify the control of a group of mobile robots, for example, to quickly deliver necessary cargo to areas affected by natural disasters. The theme of controlling a group of robots was continued by Kang Liang, Associate Professor at Shanghai Polytechnic University (China). In his report entitled 'Algorithms for control and simulation of robots', he demonstrated a strategy for controlling a swarm of robots using the so-called bee algorithm, which is suitable for both heterogeneous and homogeneous groups of robots, where each agent controls its trajectory and speed of movement, dynamically bringing them into conformity with the movement experience of other members of the group. Based on the results of experiments, this new PSO+ algorithm demonstrated higher efficiency and positioning accuracy than the canonical PSO algorithms. Zhang Xinbin, Associate Professor at Harbin Polytechnic University (China), in his report entitled 'Soft and Bionic Robots', offered methods of solving one of the main challenges faced by the creators of soft robots, namely a low level of rigidity, which negatively affects the stability of work, productivity, and accuracy of movement of such machines. The researcher concludes that the only way to achieve a high level of both stiffness and softness is to use a mechanism called variable stiffness. Xinbin noted that, as a rule, the stiffness of a mechanical part mainly depends on its material properties and structural features. Thus, variable stiffness can be achieved by structure-based methods (jamming and antagonistic stiffening) or material-based methods (shame memory materials and low-melting-point alloys). The researcher presented a gas-ribbon hybrid driven soft finger made from soft materials containing no rigid parts, with the variable stiffness realized by only adding a soft ribbon and a finger working under a gas pressure of below 35kPa. In an experiment with a gas-ribbon hybrid driven soft finger gripper, the variable stiffness mechanism demonstrated an increased quality of grasping, especially grasping of easy deformation or vulnerable objects. Next year the Financial University hopes to strengthen cooperation with Prof. Zhang Xinbin and cover new areas of application of soft robotic systems, in particular in agricultural robotics. Trung-Hieu Le, Associate Professor of the Department of Electronic Engineering at the National Taipei University of Technology (Taiwan), came up with the results of a study on optimizing deep learning techniques for object detection in inclement weather conditions. His proposed method rethinks the approach to the sequence of deep learning process in image recognition. The developed mechanism does not deal with a cascade of restoration and object detection processes, but carries them out in parallel employing RetinaNet for object classification and localization in the detection subnet, while using feature recovery and visibility enhancement modules in the restoration subnet. Trung-Hieu Le, together with associate professor of the hosting Financial University Nikita A. Andriyanov, reflected the results of their joint research in an article for the scientific journal Remote Sensing (Q1). The research of Adam Ekielski, Professor of the Warsaw University of Life Sciences (Poland), was also devoted to computer vision technologies. In his report entitled 'Methods of image analysis for assessing the quality of crops', he noted the prospects for using computer vision to increase yields while reducing production costs and herbicide use. He included data from video cameras and sensors installed on monitoring drones and agricultural equipment, as well as from sensors that record spectrum range and spectrum wave number in the list of auxiliary techniques that improve the quality of image analysis. Dilshod Z. Muzafarov, Dean of the Faculty of Mathematics, Associate Professor of the Department of Programming, Khujand State University (Tajikistan), devoted his report to solving systems of Boolean algebraic equations. The proposed new solution method supposes that, firstly, systems of Boolean algebraic equations written with logical operations are transformed or approximated in a system of continuous-polynomial equations in a unit n-dimensional cube, where ordinary arithmetic operations are used, such as addition and multiplication. Unlike systems of Boolean algebraic equations, the system obtained in this way can be solved by optimization techniques. The results of this research work can find its application in cryptographic analysis. Maria Antonietta Pascali, Researcher at the Institute of Information Science and Technologies - ISTI CNR Pisa (Italy), delivered a report on topological data analysis, its practical application and new trends. She pointed out that understanding how deep learning works can help find better ways to apply these technologies to image preprocessing for better feature extraction and learning. This approach can also be used to combat visual attacks on image recognition systems. Erbol Erbayev, Senior Lecturer at the Graduate School of Mechanical Engineering of the Zhangir khan West Kazakhstan Agrarian Technical University, demonstrated the results of his research on mathematical and instrumental methods of decision support for infrastructure management in the field of alternative energy. He revealed discovered optimization strategies for more effective usage of power electronics drawing on the example of wind turbines. Furthermore, he discussed the possibility of joint participation with the Financial University in the program for the development of carbon polygons. Bolatzhan A. Kumalakov, Dean of Astana IT University (Kazakhstan), devoted his report to the practical application of GPU and CUDA technology in parallel programming. He disclosed developed machine learning methods to determine body's positions (laying, standing, sitting, running, etc.) using obtained data on Wi-Fi network interferences. The problem of creating a strong AI was raised by Mokhiniso B. Khidirova, Senior Researcher at the Tashkent University of Information Technologies (Uzbekistan). She noted a high dependence of existing AI technologies on data, while the AI itself has low capabilities for cognition. Thus, small changes in the input data, which do not lead to a distortion of human understanding, force machine learning systems to make gross mistakes, since any object of cognition in the surrounding reality is complex and contradictory, and this incomplete coincidence of the phenomenon and essence confuses the machine. Manuel Leote Esquivel, Professor at the New Lisbon University, an expert in the field of probability theory and stochastic financial mathematics, was devoted to the possibilities of joint research activities. As a participant of the Visiting Professor Program, Prof. Esquivel was also expected to deliver a lecture on models of mutual interaction between price and liquidity of assets with switching different modes. At the end of the workshop, Sergey T. Gataullin, Deputy Dean of the Faculty of Information Technology and Big Data Analysis, expressed his hope that such scientific events would become a tradition, demonstrating constantly developing joint scientific activities of Russian and foreign researchers. The Faculty of Information Technology and Big Data Analysis of the Financial University plans to strengthen international cooperation and seek funding for further researches in the field of new technologies.