Comparing Modeling Approaches for Distributed Contested Logistics

In any military operation, reliable logistics is essential to maintaining a combat-effective force. Without the continual resupply of ammunition, food, and other materiel, forces cannot sustain their operations. Currently, logistics routes are made manually based on the judgment of individual logisticians. To automate this process, we develop two models that incorporate Geographic Information System (GIS) data to generate distributed logistics networks between two locations. These models build upon existing approaches to non-military transportation problems. The first model, the Route Selection Algorithm (RSA), modifies the $K$ Shortest Paths algorithm and employs a Mixed Integer Linear Program (MILP) to generate multiple routes that minimize tactical risk while maximizing route dissimilarity. The second model, the Route Generation Algorithm (RGA), iteratively determines the optimal path and then applies a penalty factor to previously used arcs, discouraging their selection in future routes to enable route dissimilarity. Both models return multiple dissimilar routes with minimized risk that provide commanders with several viable resupply options. After a small-scale model comparison of 100 simulated scenarios, our results indicate that the RSA produces routes with lower risk, while the RGA generates routes with higher dissimilarity and has a lower runtime. These models serve as initial formulations that can be further refined into a robust, comprehensive risk-avoidance model to be used by military logisticians. This paper introduces a novel comparison between two routing algorithms and presents an innovative method for quantifying the risks posed by enemy forces and weapons systems in military transportation problems.