Resilient Logistics in Contested Networks
Abstract
This proposal tackles the complex decision-making challenges in ensuring the resilience of logistics network operations within contested environments. Resilient logistics calls for routing personnel, equipment, and essential supplies from their points of origin to their final destinations while considering the risks associated with uncertain capacities and interdiction efforts by an intelligent adversary. Ensuring the resilience of a network necessitates assessing the impact of threats to the network. Therefore, PI will also investigate the closely related dual counterpart problem of persistent network interdiction, in addition to resilient network design and routing.The PI will develop new models and effective computational methods for large-scale solution of network design, routing, and interdiction problems with probabilistic and adversarial constraints. In particular, PI will investigate (i) probabilistic cut constraints, (ii) probabilistic path constraints, (iii) adversarial cut constraints, (iv) adversarial path constraints. These are fundamental substructures of resilient network design, routing, and persistent interdiction models.The developed methods will require the iterative solution of optimization problems with incrementally introduced constraints that will build resilience (excess capacity) into the network, considering the risk of reduced capacity or interdiction by an adversary. Identifying such constraints willrequire the frequent solution of constraint generation (separation) problems. In order to solve these subproblems fast, the PI willdevelop novel deep reinforcement learning methods based on optimization proxies for the corresponding separation problems. The fusion of optimization and deep learning is expected to bring substantial leaps in decision-making for resilient logistics in contested environments.If successful, the models and computational methods developed will be used not only for identifying critical vulnerabilities in logistics networks, but also for the efficient allocation of scarce resources to mitigate those vulnerabilities. Furthermore, they may be used to quantify the trade-off between the resilience and the cost of resilience for logistics networks for informed decisionmaking.The employment of new computational methods for persistent network interdiction is also expected to enhance the effectiveness of ISR and interdiction operations by ensuring the efficient allocation of appropriate assets and resources, and be valuable in simulating attacks on logistics networks. The research project will help to quantify the trade-off between the cost and effectiveness of interdiction operations.APPROVED FOR PUBLIC RELEASE
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Mar 15, 2024
- Source ID
- N000142412149
Entities
People
- Alper Atamtürk
Organizations
- Office of Naval Research
- United States Navy
- University of California Regents