In this article, we propose a decentralized approach to formation control in groups of stochastic mobile robots. Different from existing work, we explicitly model the presence of stochastic perturbations affecting the dynamics and model the multivehicle system through stochastic differential equations. We design a decentralized formation control strategy where only a subset of informed agents is aware of the desired target locations and derive sufficient conditions for almost sure convergence to the desired formation pattern. Specifically, we illustrate how the achievement of the control goal is related to the intensity of the noise and to the topology of the communication graph among the robots. The proposed formation control strategy is tested through extensive numerical analyses and validated experimentally on ground robots.