In this paper the problem of recursive approximation of the feasible parameter set through parallelotopic sets is studied. A recursive algorithm is proposed providing at each step the minimal volume approximation of the parameter set given by the intersection of the set estimate at time t and the feasible set determined by new measurements at successive times t+1, t+2, .... This algorithm may be useful in two respects. First, it is expected to improve on average the convergence speed of the known one step ahead algorithm. Second, it can be exploited to find a solution to the recursive feasible parameter set estimation problem for error-in-variables identification problems in the set membership context.