Robust algorithms of data classification obtained using set of weights by group of robots

Three adaptive robust learning algorithms for a group of robots are proposed in the paper, provided that each observation obtained by the robots consists of several elements, i.e, is multi-valued. The reason for the multi-valued data is that the robots in the system provide different measurements for a single external parameter observation at a time. The algorithms are based on sets of weights or interval weights of a certain type for all elements of the training set. In addition, to formalize multivalued data and modify weights in the process of obtaining new data, it is proposed to use the Dirichlet interval model. The first algorithm is a modification of the support vector machine that takes into account multivalued data. The second algorithm is a modification of the AdaBoost algorithm for multi-valued data. The third algorithm is a combination of AdaBoost and the Dirichlet interval model. All algorithms are robust and use a minimax decision-making strategy.