We formulate the problems of optimal composition of a heterogeneous group of robots designed to search for and transport loads. The goal is to maximize the expected number of loads found and taken, subject to the resource constraint in the form of the total carrying capacity of robots or the number of modules. The number of found loads and the weight of each load are considered as uncertainty factors. The number of found loads depends on the number of search robots linearly or according to the binomial distribution. To solve the problems, the stochastic programming method is used, maximizing the mathematical expectation of the number of loads transported. The numerical solution results show the effectiveness of the method in comparison with heuristic approaches. Analysis of the solutions allows us to conclude that it is advisable to diversify the composition of a group of robots under uncertainty. In this regard, an analogy can be drawn with biological diversity in populations of living organisms.