基於多輸入多輸出雷達在多路徑環境中的二維到達角估計及真實目標與虛假目標識別

Abstract

時分複用多輸入多輸出雷達系統在現代雷達應用中廣泛使用，具備高角度解析度的優勢。然而，在目標具有速度的情況下，速度會引起相位失真，因此需事先進行相位補償，以確保二維到達角度的估測不受影響。在實際操作中，雷達系統經常面臨多路徑效應的挑戰，這種效應會導致接收信號中出現複雜的直射和反射現象，從而降低信號性能與定位準確度，進一步影響雷達對目標的精確定位與追蹤能力。 本論文首先建立一套多路徑信號模型，以模擬多路徑效應對複雜車用環境中雷達性能的影響。該模型考慮了經由不同路徑返回雷達的多重信號、因傳播距離變化所造成的路徑損耗，以及目標反射對接收訊號的影響。此外，亦納入造成多路徑效應之反射物體的特性。為進一步提升角度解析度，本研究結合時分複用多輸入多輸出雷達與稀疏天線陣列。在此基礎上，我們提出一套新的目標分類與識別方法框架，包括：用於區分直視與非直視目標的分類方法，以及分別針對多角度反射與平面反射情境所設計的真實與虛假目標識別方法。相較於現有技術，本方法具備更高的可靠性，無需進行複雜運算，亦能準確分類與識別真實目標與虛假目標，並有效利用所提取之資訊，提升雷達後續處理的準確性與可靠性，增進對周遭環境的整體理解能力。

Time-Division Multiplexing Multiple-Input Multiple-Output (TDM MIMO) radar systems are widely used in modern radar applications due to their advantage of high angular resolution. However, when the target has velocity, the motion induces phase distortion, necessitating phase compensation in advance to ensure the accuracy of two-dimensional direction-of-arrival estimation. In practical operations, radar systems often face the challenge of multipath effects, which result in complex direct and reflected signals in the received data. These effects degrade signal performance and positioning accuracy, further impacting the radar's ability to precisely locate and track targets. This thesis first establishes a multipath signal model to simulate the impact of multipath effects on radar performance in complex automotive environments. The model considers multiple signals returning to the radar via different paths, path loss caused by variations in propagation distance, and the influence of the target's reflection on the received signals. In addition, the characteristics of reflecting objects that cause multipath effects are also taken into account. To further improve angular resolution, this study combines TDM MIMO radar with sparse antenna arrays. Based on this, a new target classification and identification framework is proposed, including a classification method to distinguish between line-of-sight (LoS) and non-line-of-sight (NLoS) targets, as well as true and ghost target identification methods designed for both multi-angle and planar reflection scenarios. Compared to existing techniques, the proposed method offers higher reliability, requires no complex computation, and is capable of accurately classifying and identifying true and ghost targets. It also effectively utilizes the extracted information to improve the accuracy and reliability of subsequent radar processing and enhance the overall understanding of the surrounding environment.