Local Encoding Transformations for Optimizing OBDD-Representations of Finite State Machines

Ordered binary decision diagrams are the state-of-the-art representation of switching functions. In order to keep the sizes of OBDDs tractable, heuristics and dynamic reordering algorithms are applied to optimize the underlying variable order. When finite state machines are represented by OBDDs the state encoding can be used as an additional optimization parameter. In this paper, we analyze local encoding transformations which can be applied dynamically. First, we investigate the potential of re-encoding techniques. We then propose the use of an XOR-transformation and show why this transformation is most suitable among the set of all encoding transformations. The presented theoretical framework establishes a new optimization technique for OBDDs.