© 1973 by British Computer Society
| ||||||||||||||||||||||||||||||||||||||||||||||||||
State estimation algorithms for non-linear stochastic sequential machines
Computer Control Section, Computer Centre, Nuclear Research Centre, Democritus, Athens, Greece
The problem of estimating the state sequence for a number of stochastic sequential machine models is considered. Basically, all models are assumed to have the structure commonly used in the stochastic control field, i.e. that of a deterministic system corrupted by random disturbances. A Bayesian sequential information processing approach is followed which is most convenient for dealing with finite-state systems. The resulting estimators are given in the form of sequential algorithms which are suitable for a digital computer. However, these estimators can be reduced in the form of finite machines with the aid of well-known state equivalence/reduction techniques. An example is presented which illustrates the effectiveness and usefulness of the theory. The results of the paper are applicable to a large variety of digital communication systems involving noisy channels and also to quantised stochastic control processes.
Received July 1971.
* Computer Control Section, Computer Centre, Nuclear Research Centre, Democritus, Athens, Greece