Component-by-component state smoothing is discussed for multi-dimensional dynamic systems with non-linear random interference such as jamming. Each component of the observation model is a non-linear function of only one state component, arbitrary random interference, and observation noise. Each state component is first approximated by a finite state machine and then, using the Viterbi decoding algorithm of information theory, the state components are sequentially smoothed in parallel. This results in a memory reduction for the implementation of the state smoothing. Simulation results have shown that the proposed scheme performs well, whereas the classical estimation schemes cannot be used, in general, to estimate the states of dynamic models with arbitrary random interference.