Identifying patterns in sequences of variables

Abstract

Complex rostering problems often require to recognize and count some patterns in the employees' schedules. The number of occurrences of such patterns is then constrained to comply union rules, business requirements, and other workflow constraints. A common approach to deal with these constraints is to model them as cost-regular constraints but the resulting automata are not trivial to encode manually. This paper proposes a new constraint, the pattern constraint, whose goal is to recognize sets of patterns and constrains their occurrences. The pattern constraint is implemented in two different ways, relying respectively on a modified version of the regular constraint and on the cost-regular constraint. Both approaches employ an algorithm that automatically encodes the underlying automaton. As a result, the pattern constraint provides a high-level modeling abstraction, removing the burden of encoding automata for pattern recognition and allowing to automate the creation of complex models for rostering problems. © 2011 Springer-Verlag Berlin Heidelberg.