A trilevel defender-attacker problem with partial interdiction

Abstract

Terrorism is one of the most serious problems of today's world. Thus, the reliability and robustness of network systems with critical infrastructure are crucial. In this thesis, the best location-allocation strategy is sought to serve the customers in a network system in case of intentional disruptions caused by terrorist attacks. For this purpose, a trilevel mixed integer programming model is introduced. In the upper level, the system planner (leader) determines the facility locations, the capacities of the facilities, and the assignment of customers to the facilities. The attacker (follower) makes the choice of the interdiction fractions on the opened facilities in the middle level. In the lower level, the system planner sets the reassignment of the customers considering the capacity reduction of the facilities and decides which customers are served by outsourcing. In the upper level, a tabu search heuristic is used to locate facilities and commercial solver Cplex is employed to determine the capacities of the opened facilities and pre-attack allocations. The best interdiction strategy is found by the Electromagnetism Like Algorithm (EMLA) in the middle level. Finally, the postattack allocations are found with Cplex. This solution method gives quite satisfactory results.