Date of Graduation


Document Type


Degree Name

Master of Science in Industrial Engineering (MSIE)

Degree Level



Industrial Engineering


Kelly Sullivan

Committee Member

Sarah Nurre

Second Committee Member

Ashlea Milburn


Applied sciences, Facility location, Logistics, Network optimization, Operations research, Vehicle routing


The location arc routing problem (LARP) is a network optimization problem combining strategic facility location decisions and tactical or operational vehicle routing decisions for customer demand located on arcs of a network. The LARP seeks to locate facilities, or depots, and create vehicle delivery routes to minimize costs. The total cost is comprised of three components: fixed facility locations costs, fixed route creation (or vehicle acquisition) costs, and variable arc traversal costs. The applications of the LARP are varied and often include public services such as mail delivery, garbage collection, and street sweeping. In all of these applications, the magnitude of customer demand may be unknown at the outset of the problem and realized uncertainty can greatly affect the final solution. To the author’s knowledge, there is currently no discussion of formulating or solving a LARP with uncertainty.

This paper presents an iterative tabu search, augment-merge heuristic to solve the LARP with stochastic customer demand. Each realization of customer demand for a particular network, represented by an individual scenario, was generated using the deterministic mval instances (with 24-50 nodes and 44-138 arcs) created by Hashemi Doulabi and Seifi (2013) and a truncated normal probability distribution. The tabu search phase handles the depot location decisions and chooses a set of depots to be used across all scenarios. The augment-merge phase creates a set of vehicle routes for each scenario. One-third of the initial experiments resulted in stochastic solution costs less than their deterministic counterparts indicating the promising value of considering customer demand uncertainty using the proposed stochastic LARP algorithm.