智慧型雷達預估器於追蹤高速運動目標之研究

Abstract

隨著科技的進步，飛機的飛航速度持續加快，而彈道飛彈也高度發展，使得雷達的功能也必須隨之提昇，以飛彈的防禦系統而言，其構想是，當彈道飛彈來襲時，地面發射攔截飛彈在空中予以摧毀，實現技術上難度極高，必須仰賴電腦整合衛星雷達、地面雷達、和攔截飛彈，三位一體的鎖定來犯目標，並在空中予以摧毀。如果要成功地攔截飛行極快的彈道飛彈，則反飛彈必須具有相當快且精確的操控能力外，利用雷達來估測目標的運動狀態更是重要，為了提高雷達的功能，未來的發展趨勢是研發智慧型雷達預估器，以達到精確地追蹤高速運動目標，並且可以做長時距的預測。傳統的Kalman filter雖然可以估測目標的運動狀態，但必須事先知道環境中雜訊的統計量及良好的初始狀態設定，才能使Kalman filter得以最佳化。基於Kalman filter的缺點，我們在這篇論文中，提出一個方法，不用事先知道環境中雜訊的統計量及良好的初始狀態設定，利用類神經網路來估測目標的運動狀態，我們以模擬的方式來驗證此方法的可行性，並與傳統的Kalman filter進行比較。

Due to the development of new technologies, the plane and missile fly much faster. Thus, the function of the radar must be enhanced. For an air-defense system, when the missile invades, the system should launch intercepting missiles to destroy the invading missile in midair. It is by no means an easy task, and demands the integration of satellite radar, ground radar, and intercepting missiles together with the computer. For a successful interception of the missile flying with such a high speed, the intercepting missiles must be equipped better Guidance Laws, and the radar should be able to predict target dynamics. We aim to develop an intelligent radar predictor to track the high-speed moving target precisely. Although a conventional Kalman filter can also predict target dynamics, it needs to know the statistics of the noise in the environment and the initial state in advance. To avoid the limitation of the Kalman filter, in this thesis, we propose using a neural-network approach for prediction. The feasibility of the proposed approach is demonstrated through simulations and its performance is compared with that of the conventional Kalman filter.