����APP

Scientific field of Prof. I.G. Tsmotsya – "Development of methods and real- time intellectual tools for the synthesis of mobile smart systems".

Within the scientific field, a theory has been developed based on neural networks for improving the accuracy of navigation data measurement, restoring lost data, predicting movement and spatial coordinates, cryptographic encryption and decryption of data, and controlling the movement of ground-based mobile robotic platforms (MRP). Within the framework of this scientific direction, the department has carried out fundamental and applied research on the development of:

• theoretical foundations for the development of intelligent, high- performance, specialised real-time tools that ensure the coordination of data flow intensity with the intensity of calculations in specialised tools;
• methods for the spatial-temporal mapping of neural algorithms into coordinated parallel structures, which, by taking into account the characteristics of the implementation tools, the requirements of specific applications, and the intensity of data flow, ensure the synthesis of intelligent tools with high equipment utilisation efficiency;
• methods, parallel algorithms, and VLSI structures for accelerated computation of basic neural network operations;
• structures and methods for synthesising exchange devices based on multi- port memory.

During the implementation of research projects "Neural network technology for real-time data protection and transmission using noise-like codes" (DB / "Neuroprotection", state registration number 0119U002256) and "Experimental system of neural network cryptographic protection and real-time data transmission using barcode-like codes" (DB / "Neurocode", state registration number 0121U109503), research was conducted on the development of information technology for neural network cryptographic protection of data in real time. This information technology was developed based on an integrated approach that includes: research and development of the theoretical foundations of neural-like data encryption/decryption; development of new algorithms and structures for neural-like data encryption/decryption; a modern element base and means of automated design of software and hardware components. Neural network cryptographic data protection is implemented using auto-associative type direct propagation neural networks, which are trained using a non-iterative method of sequential geometric transformations. A key feature of such networks is the ability to non-iteratively calculate the weight coefficients of synaptic connections between neural elements. The use of such a neural network for cryptographic data protection ensured the repeatability of results and the hardware and software implementation of data encryption and decryption blocks with high technical and economic performance. The information technology of neural network cryptographic data protection is focused on encryption with symmetric keys, in which the encryption key and the decryption key are the same or the decryption key can be easily��calculated based on the encryption key. Encryption is performed on plain text using a key that encompasses the architecture of a neural network, a matrix of weight coefficients in floating point format, and masking codes.

As a result of this research, the following is proposed:

• a method of neural cryptographic encryption and decryption of data, which is distinguished by the use of a tabular-algorithmic approach to the implementation of neural elements;
• simulation models of encryption and decryption, whose advantages are a user-friendly interface and the ability to work with files of any format;
• a method of tabular-algorithmic coordinated calculation of scalar products with analysis of k-bit slices, which, unlike existing methods, allows the selection of the value of k, the number of bit slices, and the duration of the pipeline cycle;
• an improved method of tabular-algorithmic calculation of scalar product for floating-point numbers, which differs from existing methods by means of reduction to the greatest common order of input data and weight coefficients;
• an improved method for selecting an element base for the synthesis of a data protection and transmission system, which differs in the calculation of an integrated assessment of the effectiveness of the element base and the consideration of the requirements of a specific application;
• a simulation model for automated selection of the element base for the synthesis of a data protection and transmission system, which differs from others in its database filled with modern components.

As part of the research project "Methods and means of neuro-fuzzy control of a group of mobile robotic platforms" (DB / "Neurogroup", state registration number 0123U101688), the development and research of system structures, methods, algorithms and means of neuro-fuzzy control of a group of MRPs, assessment of the navigational state of the surrounding environment, exchange of data and control commands between MRPs was carried out. During the work, requirements for the neuro-fuzzy control system for a group of MRPs were identified, the main ones being: ensuring effective control of a group of MRPs; minimising the time required to perform tasks; flexibility and adaptability to changing operating conditions; reliable and stable operation during the implementation of various scenarios; expansion of functions and scaling in terms of the number of MRPs; accuracy and reliability of motion control for each MRP; response to changes in operating conditions; uninterrupted operation of the MRP group; efficient use of MRP resources; reduction in size, weight and energy consumption; real-time control; collection of data on the environment and the status of MRPs; wireless communication between MRPs; development of software tools taking into account distributed architecture; implementation of a programming interface with the possibility of developing additional software and integration with other systems; storage of data on the status of all MRPs for further analysis and improvement of MRP group management. A structure for an MRP motion control system using fuzzy logic has been developed, the main components of which are sensors, a phasing unit, a decision-making unit, a rule base, and a defasings unit, which ensures the adaptation of MRP and group functioning in conditions of incomplete information.

As a result of this research work, the following were developed:

• a controller for neuro-fuzzy MRP motion control based on phasing blocks, a fuzzy inference and neuro-defasification block, which, thanks to its ability to adapt to the requirements of specific applications and the use of tabular-algorithmic calculation methods, ensures real-time operation and high technical and operational characteristics;
• a method of training a neural network with a teacher, which, by forming an input matrix, normalising training vectors, finding the mean value, and centring the input data, ensures the calculation of coefficients for tuning the neural network to perform defasings;
• a method for operating a neural network during defasings, which, with the help of preliminary calculations and lateral connections, ensures increased accuracy in the formation of MRP control signals at the output of the neural network;
• the method of constructing a neural defasicator, which, through the use of a tabular-algorithmic implementation of the basic operation of neural defasicisation, ensures increased performance and the implementation of a neural defasicator with high technical and operational characteristics;
• a fuzzy logic controller with neural defasification, which, thanks to adaptation to the amount of input data and the use of developed rule bases, ensures increased performance and accuracy of MRP motion control signal formation;
• an improved method for controlling the movement of a group of mobile robotic platforms, which ensures effective control of a group of MRPs in real time, as it takes into account the variable parameters of the platforms and the changing state of the environment;
• the method of autonomous control of MRP movement, which, by taking into account the state of the group, the external environment and the values of MRP characteristics, ensures effective autonomous control of MRP movement in real time;
• an improved method for the time distribution of resources in the storage environment of exchange means, in particular, ensuring the coordination of the intensity of access to it with the intensity of data flow, which allows selecting the required speed of the storage environment and increases the efficiency of equipment use;
• the method of increasing the intensity of access to the storage environment in exchange means using a controller parallel memory (KPP), which�� by introducing pipeline registers ensures a reduction in the time of data retrieval from the storage environment to the time of register operation.

Based on the results of their research, the department's scientists defended the following dissertations: in 2013, O. V. Skorokhoda defended his candidate dissertation "Synthesis of real-time parallel-vertical type neuroelements and neural networks" in the specialty 05.13.23 "Artificial Intelligence Systems and Tools"; in 2015, I. Ye. Vavruk defended his thesis "Control of the movement of a wheeled mobile robotic system using fuzzy logic" in the specialty 05.13.23 "Artificial Intelligence Systems and Tools"; in 2019, T. V. Teslyuk defended his thesis "Methods and Means of Data Collection and Processing in Enterprise Energy

Efficiency Management Systems" in the specialty 05.13.06 "Information Technologies"; in 2021, V. Ya. Antoniv – candidate dissertation “Information technologies for parallel sorting and data search” in the specialty 05.13.06 “Information technologies”; in 2023, Yu. A. Lukashchuk defended his doctoral dissertation "Real-time data protection information technology for mobile smart systems using neural networks" in the specialty 122 – "Computer Science".

The developed means of improving the accuracy of navigation data measurement, restoring lost data, predicting movement and spatial coordinates, cryptographic encryption and decryption of data, and controlling the movement of ground-based mobile robotic platforms are protected by 21 patents for inventions.