EXPRESS: Tactical Workforce Planning and Deployment with Timing Coordination Requirements

We address a tactical planning problem facing organizations that must perform geographically dispersed jobs with timing dependencies using appropriate teams of workers and equipment. This Tactical Workforce Planning and Deployment (TWPD) problem, which is motivated by the annual planning and budgeting activity for rail infrastructure maintenance and upgrades, requires making simultaneous decisions on resource sizing and selection, resource-to-job assignment, and resource repositioning to complete all jobs while minimizing resource and transportation costs. Key challenges include timing coordination among jobs, varying resource capabilities, and sequence-dependent changeover costs. Prior scheduling and routing models often focus on short-term decisions, assuming a given set of resources, and do not incorporate complex timing coordination requirements. We propose a large-scale optimization model, with strong timing coordination constraints, that integrates all the features of the TWPD problem. To solve this difficult problem, we propose three new families of valid inequalities to further strengthen the model and develop problem reduction procedures as well as an optimization-based multi-stage algorithm to identify good initial feasible solutions. Our Composite method applies these acceleration strategies before initiating an exact branch-and-bound solution procedure. To test the effectiveness of this method, we first apply it to real data sets for track maintenance from a leading railroad company and compare its performance with a method that uses a standard solver without our enhancements. We find that our method dramatically outperforms the standard method. We also test our method on many randomly generated problems to validate its superior performance and identify problem features that make this problem difficult to solve. When compared to a manual tactical plan, our method quickly generates a near-optimal solution that saves over 15% in annual total cost, demonstrating the practical value and viability of our model and method. We conclude by discussing some managerial implications and insights from our results.