Circular automata

We define a finite-state machine called a circular automata (CA) which processes information in a queue; we show that any function computed (or any language recognized) by such a machine is computable (recognizable) by a Turing machine and vice versa. Space and time bounds are given for the needed simulations. Furthermore, the class of languages recognized by (non-) deterministic linear bounded automata is equal to the class of languages recognized by (non-) deterministic CA which don't expand the length of the contents of the queue. Whether every language recognized by such a non-expanding CA is recognized by a deterministic one is equivalent to the famous LBA problem.CA can be viewed as generalizations of ordinary finite automata and as a Shepherdson-Sturgis single register machine programming language. An interesting model of a non-expanding CA is that of a finite-state machine which process tapes in the form of a loop. This appears to be a very natural way to process magnetic tape which circles back on itself.