Given an example of second order Markov process how can it be transformed to a first order by using augmented set of state variables? Is it possible to do that without increasing the number of parameters needed to show the transition model?

I was thinking that I can just ignore the extra variables in the second order model but the question asks for extra set of state variables?