Pekiştirmeli öǧrenme ile silah tareti platformunun yönelim kontrolü

In this study, deep reinforcement learning methods are applied for the attitude control problem of nonlinear gun turret platforms. In order to create the problem scenario, mathematically modeled gun turret dynamics are applied in a game engine based simulation. Deep Q Learning (DQN) and Deep Deterministic Policy Gradient (DDPG) algorithms were applied in the solution of the problem and the reward function was designed iteratively. In order to compare the results, a classical control theory algorithm was developed and the controller responses obtained by providing the same reference signals to these algorithms were compared. Without providing system dynamics information to the relevant learning algorithm, approaching as model-free, the problem of reference attitude tracking of the gun turret was solved under certain assumptions and the results obtained were shown and compared in detail.