Problem statement: Most of modern robotics systems are still lowautonomous, i.e. they can not perform in extreme environments having preloaded scenarios. The reason of this is a complicity to predict possible situations and how those can further evolve. Because of this human-operator keeps being an essential part controlling robot in emerging situations. However this approach has a weakness that in sophisticated situations human-operator may not act correctly. Purpose: to study models supporting decision making by human-operator. These models should simulate intellectual activities of high-skilled operators in various emerging situations. Methods: within these studies we plan to investigate 2 main steps in creating models for supporting decision making. These are the following: 1. Integrated usage of logical–linguistic (based on natural language) and mathematical (calculation) formal means for qualitative and quantitative information processing. This would allow both to formalize logics of high-skilled operators and to consider dynamics of control processes. 2. Creation of procedures which would implement sefl-learning (adaptation in wide sense) in case of a priori uncertainty. Those should be based on logical structure of likely-based (fuzzy) conclusions, using inductive inference and analogy inference, which are inherent to mind of humans. Results: It is expected that this approach to build model for decision support in robotics systems will increase quality of decision making by human-operator. This statement is supported by a fact of intensive usage of human-operators’ experience especially in emerging situations.