More and more tasks are aided by technological solutions. Limitations of technical systems, in most cases, result with a person supporting the system during its operation. A person is dislocated from the work area and uses robots as tools. Development of interactive models could change this paradigm by placing a person at the centre with support and assistance provided by the robot as a partner. So far, for security reasons, robots and people worked strictly separate, but development can enable joint working areas, which can be beneficial. In such environments a contact between robots and humans can occur, and the question of the physical interaction becomes an important issue. The issue of contact with humans so far was seen as a binary problem where if a person came to contact the robot stopped, otherwise his action wasn’t interrupted. The thesis proposes the use of new technologies and interaction models based on physical interaction, enabling humans and robots to work and perform interaction in shared working areas. To make this acceptable for the user the interaction must be intuitive. The robot must be able to perform movements and tasks at the request of users, know the workspace and take operator safety into account. The proposed model determines robot’s activities through physical interaction with humans. Classification of tactile stimuli, registered using capacitive sensors, force and spatial position, distinguishes the elements and the meaning of interaction. In order to provide a certain autonomy and the possibility of moving through space the process of motion planning was integrated in the model. As part of the research a multiple criteria interpretation of the working area, in which there is a distinction between objects in the environment, human, goals, the robot and the robot path, is defined. The model of interaction is formed by a sequence of actions that the robot executes which ultimately results in robotic action. Defining the probability of the model variables results from the interaction with humans. Learned patterns are stored as long-term knowledge based on which the robotic action, in accordance with the current state of the environment, is formed. Temporal difference assigns a significantly higher probability factor to more recent events, and to those more distant in time much smaller. This thesis is organized in six chapters, as follows: Chapter 1: Introduction. This chapter presents the advantages and limitations of robotic technology. Senses and perception, along with sign language and tactile interaction are described. With the development of technology tactile language and tactile interaction are applied in interfaces with technology. Since the work is based on physical interaction the importance and benefits of touch are emphasized. Chapter 2: Human robot interaction. After introductory remarks there is a summed overview of human robot interaction. The basic user interfaces are described and the relevance of the research is presented. Also the chapter reviews the literature related to the topic of the thesis and explains the research scope. The interaction model, which will recognize the intent and message of a person and in accordance with them to achieve desired behaviour, is explained. Chapter 3: Tactile stimuli classification. The third chapter deals with the interpretation of the physical interaction. A description of the system components, hardware and software, which can react to tactile stimuli is given. The classification model, which is divided into several categories, is described. Multiple criteria interpretation of space, with different layers and time components, based on which the running planning movement in the configuration space is explained in detail. Experiments that are used to define the parameters of classification, such as speed, displacement, and force, are shown at the end of the chapter. Chapter 4: Probabilistic model. The chapter presents a probabilistic model. Concepts necessary for describing the model, theoretical basis of probability and Markov processes that form the basis of the model are introduced and defined. The model is based on the multiple criteria workspace definition, the stochastic chain status and experiential knowledge with time differentiation of events. The experiment deals with the problem complexity and size of the probability transition matrix. Chapter 5: Experimental results. This chapter deals with the validation and verification of the proposed methodology. Simulations and experiments are conducted to test the safety and efficacy of the developed algorithms and classification. Experiments were carried out in laboratory conditions using lightweight robot arm and capacitive sensors ("artificial skin"). Chapter 6: Conclusion. This chapter summarises the main contributions of the dissertation and presents several recommendations for future research.