Seminars

The threat posed by hostile unmanned aerial vehicles to both civilian and military
targets is real and escalating. Kinetic interceptors have emerged as a cost-effective,
low-collateral alternative to traditional air defense measures; however, a key chal-
lenge lies in enabling the strategic and coordinated deployment of multi-interceptor
networks to counter large-scale, synchronized threats. The adversarial interaction be-
tween interceptors (defenders) and malicious drones (attackers) naturally lends itself
to the framework of noncooperative dynamic games. This perspective has recently
reignited the study of pursuit-evasion games in the context of multi-agent target de-
fense. However, much of the existing work relies heavily on geometric intuition and
reasoning, often lacking provability and generalizability.
This work fills these gaps and advances the field with three main contributions.First, we justify and generalize the widely used geometric method for single-attacker target defense games through the lens of parametric convex optimization, extending its applicability to more advanced scenarios involving multiple defenders and moving targets. Second, we introduce multi-phase game decomposition methods to address multi-attacker scenarios, deriving optimal defender paths against both cooperative and noncooperative attackers and establishing conditions for the absence of dilemmas. Third, for multi-defender multi-attacker scenarios, we propose combinatorial local game decomposition methods and present an approach to integrating discrete task-level decisions with continuous control strategies on assignment singular surfaces. The effectiveness of the proposed strategies for all scenarios is validated through simulated experiments.