In this paper we propose a model that captures the influence of noise
and speed on the correct behavior of a computing device situated in a
dynamic environment. Within this model we analyze the relation between
structural properties of automata and their immunity to noise. We pro
ve upper and lower bounds on the effect of noise for various classes o
f finite automata. In addition, we show similar relationships between
relative speeds of the automaton and the environment and the accuracy
of computation. Our model, combining basic notions from algebraic auto
mata theory and the theory of stochastic processes, can serve as a sta
rting point for a rigorous theory of computational systems embedded in
the real world.