Methods of Information Presentation and Pre-Processing in Multimodal Interfaces Based on Self-Organizing Kohonen Maps

Development and implementation of intelligent planning and control systems using multimodal interfaces as elements of decision support systems capable of performing complex operations in all links of the feedback chain is a complex task, but in modern conditions it is increasingly in demand. The paper proposes the use of the Kohonen self-organizing map, SOM (Self-Organizing Maps ), as a link in the feedback chain of the human-machine interface, allowing not only to reduce the space of parameters of the state of the control objects in which the search for solutions is carried out, but also to carry out clustering, on the basis of which in the future it becomes possible to more clearly make informed decisions by a human operator or to implement transparent deterministic or statistical algorithms by a robot. A corresponding algorithm has been developed, which showed a gain of 10–12% on experimental data on a sample of 100 comparative experiments on a dataset dedicated to autonomous driving, with a compression ratio of about 5, …, 10.