Model predictive control based cooperative pursuit evasion for uav

Abstract

This thesis proposes game theoretically model predictive control based guidance approach for pursuit-evasion problem of uav's. The main idea is that guided swarm uavs pursue towards to adversary uav which evade to survive as long as possible. Game theoretical approach of pursuit-evasion is based on designing the cost functions for each pursuer to converge adversary evader. Proposed approach is examined as decentralized. Therefore, each pursuer can be able to handle its mission independently without being affected by the other pursuer. The main contribution is the formulation of swarm pursuit-evasion problem as the game theoretical which can enable to develop optimization-based algorithms that bring superior strategies to pursuers for one-to-one, two-to-one scenarios during the air combat. This work proposes an algorithm to enhance applicability of the game theoretic non-convex model predictive control problems on real-systems that have nonlinear controland state constraints. Proposed algorithm provide a model predictive control-based guidance system which orientates the pursuers according to the evaders dynamics and positions. Nonlinear constraints are convexified along the finite-horizon time without loss of generality in successive linearizations. After discretization of dynamics, the sub-optimal convex problem can be applied in model predictive concept for time-critical scenarios such as collaborative pursuit-evasion of aerial vehicles.