Input don´t care sequences in FSM networks

We present an approach to compute all input don´t care sequences for a component in an FSM network with an arbitrary topology. In a cascade FSM network, Kim and Newborn´s (K-N) procedure exactly computes all input don´t care sequences for the driven machine. However, for a component in a general FSM network the problem of computing and exploiting input don´t care sequences is unsolved. We demonstrate that this problem can be reduced to one for a cascade circuit. This reduction uses the notion of an abstract driving machine. In some cases, the exact computation and exploitation of these sequences may be too expensive. We propose methods to compute subsets of input don´t care sequences. We discuss the implementation of these algorithms using implicit methods. We also present approximate methods for managing the complexity of large FSM networks. Finally, we give some preliminary results on small networks.