The number of used vehicles is growing very fast. This causes the bigger number of accidents. A lot of accidents and crashes are caused by driver factor. The solution might be an unmanned transport control. But use of such transport leads to another problem – how to provide safety drive and to prevent collisions of unmanned transport. The scietific novelty of the doctoral research is the technology for the embedded system, which is able to learn in real time to avoid crashes without any datasets, preliminary training and a teacher. This technology is universal for different transport types. Author named it as Immune Neural Network (INN). It is a symbiosis that uses best features of the artificial neural networks (ANN) and artificial immune system (AIS). Separately each of them has a bunch of disadvantages. ANN requires preliminary training, may be overtrained etc. AIS does not require training, but the optimization requires time and may not be used for real time control. Novel INN system combines the possibilities of ANN and AIS is adapted for real time safety control. The main goal of the doctoral work is to develop immune neural network based technology of machine learning for unsupervised safe vehicle control. The main hypothesis is that an immune neural network can make control decisions to prevent vehicle collisions with better performance than a traditional neural network in this task. Following tasks were defined: - To study the objects of electric transport traffic movement control and their interaction. - To study the existing solutions for the recognition and prevention of dangerous situations in electric transport, which are based on the algorithms of ANN. - To compare centralized, decentralized and distributed system structures, to choose the most suitable one for the proposed task and to develop the novel system structure, which could help to make the proposed system cheaper, faster and easier to implement. - To develop mathematical models and algorithms, that could help to solve different types of transport safety and collision prevention tasks, such as object recognition task, traffic light signal recognition task, possible crossing point detection task, collision probability evaluation task, collision prevention task. - To develop novel INN based algorithm for UEV dangerous situation recognition and prevention task. - To develop electrical circuit diagram with an immune memory (IM) for UEV based on a single board computer. - To make computer simulations and to prove the efficiency of the proposed algorithms. The introduction of the dissertation contains the task statement, the analysis of the existing types of solutions of the problem, grounding and topicality of the chosen topic, practical application of the proposed algorithms and methods.4 First chapter of the doctoral thesis is devoted to the comparing centralized, decentralized and distributed system models and developing the novel system structure, which could help to make the proposed system cheaper, faster and easier to implement. Mathematical models and target function are described in the second chapter of the doctoral thesis. Mathematical models are based on the algorithms, such as convolutional neural network (CNN) and artificial neural network (ANN), and on the novel INN based technology of machine learning for unsupervised safe vehicle control. The third chapter is devoted to the methods and algorithms used in the dissertation. The choice of electric transport communication and control methods are studied and described. Several algorithms are developed: - CNN based algorithm for object recognition task; - algorithm for traffic light red signal recognition task; - ANN based algorithm for collision probability evaluation and minimization task; - novel INN based algorithm for collision probability evaluation and minimization task. The fourth chapter presents the developed prototypes and computer models for testing the proposed algorithms. The electrotechnical description of the developed system with the explanation of the electrical scheme are proposed in this part of doctoral thesis also. The last chapter of the dissertation is devoted to the results of testing the efficiency of the developed algorithms. Developed algorithms were tested with the help of computer modeling.