Title: An Algorithm for Solving Workforce Allocation and Utilization Problems
Speaker:
Dr. Ada Barlatt, University of Michigan, Employee Candidate, 1412
Date/Time:
Tuesday, December 9, 2008, 10:30 – 11:30 am
Location: CSRI Building/Room 90 (Sandia-NM)
Brief Abstract: Many workforce planning problems include two levels of decision making: how many workers to hire and how to assign these workers to tasks. To ensure that the solution can be implemented, it is critical that real-world operational details be taken into account.
Unfortunately, including real-world detail can lead to tremendous challenges. Often these workforce allocation and scheduling problems include many complex and often non-linear relationships that make it difficult to model the problem using traditional mathematical modeling techniques. Even once the model has been created it may not be tractable - many mathematical models of these problems contain a very large number of constraints and have weak linear programming relaxations that make the model difficult to solve.
To overcome these difficulties, we present a model based on composite variables (variables that capture multiple decisions simultaneously) as a way to capture the real-world complexities and ensure an implementable solution. In addition, we present a novel solution approach, based on solving a set of carefully selected feasibility problems, to find high-quality solutions in acceptable run times.
We discuss the class of problems for which our approach is appropriate and will use an automotive stamping facility with a heterogeneous workforce as a demonstrative example. In this example, we do not restrict batch sizes, or sequencing of part types; nor do we fix the number of set ups a priori. In addition, we allow sequence- dependent changeover times and varying due dates. Computational results are presented using data from a major automotive manufacturer.
CSRI POC:
Mark D. Rintoul, (505) 844-9592 |