An efficient application of dynamic programming usually requires that the task decompose into several instances of the same task for a shorter input. In case of the Levenstein distance, this boils down to prefixes of the two strings and the number of edits required to get from one to the other. I don't see how such a decomposition can be achieved in your case. At least I don't see one that would result in a polynomial time algorithm.
Also, it is not quite clear what operations you are talking about. Depending on the context, a swap or exchange can mean either the same thing as transposition or a replacement of a letter with an arbitrary other letter, e.g. test->text
. If by "transpose/swap/twiddle/exchange" you try to say just "transpose", than you should have a look at Counting the adjacent swaps required to convert one permutation into another. If not, please clarify the question.