In this article, we address the modeling of uncertainties and errors in the simulation of situated agents. We propose a model of interaction that integrates uncertainties and errors due to effectors and sensors. This model is built upon a stochastic modeling of the interaction between an agent and its environment. We illustrate an application of such a model in case of simulating a set of mobile autonomous robots evolving in a structured environment (the inside of a building). Moreover, the built simulator has been designed to be executed on parallel computers and we describe its parallelization together with major results in terms of speed-up and size-up.