The nonstationary one-dimensional autoregressive model is investigated, where and are given random variables. We construct new estimates for the parameter that have more simple limit distributions in the unstable case ( = ) in comparison of least square estimates limit distributions of which have complicated form expressing by means of functional from Wiener process (see, for example in the case =1).