應用於毫米波通訊系統中的波束成型訓練之機率性樹狀搜尋法

Abstract

在本篇論文中，我們提出兩個應用於毫米波通訊系統中尋找波束方向的搜尋法，並且將他們命名為機率性單界線搜尋法和機率性雙界線搜尋法。傳統上基於編碼簿的波束方向搜尋法會使用二元搜尋法，亦即量測訊號兩次，每次量測使用不同的碼字。在我們提出的第一種搜尋法(單界線搜尋法)，我們只量測訊號一次，也就是我們節省了一半的搜尋時間。儘管我們減少了一半的搜尋時間，但是估計波束方向的成功機率並沒有因此衰減到剩一半。在我們提出的第二個搜尋法(雙界線搜尋法)，我們可以控制搜尋時間，讓他介於原本二元搜尋法時間的一半到一倍。搜尋的時間其實會直接影響成功機率，搜尋時間越長，成功機率越高。所以根據我們提出的第二個搜尋法，我們必須在搜尋時間和成功機率間做取捨。

In this thesis, we propose two search methods for finding beam directions in mmWave communication system, namely opportunistic one-threshold algorithm and opportunistic twothreshold algorithm. In conventional codebook based beamforming training algorithm, we utilize binary search which means we need to measure signals twice with different codewords. In the first proposed search method, opportunistic one-threshold method, we reduce the search time into half of the original time which binary search requires. More specifically, we only measure the signal one time. Although we reduce the search time, the success rate dose not decay dramatically. In the second proposed search method, opportunistic two-threshold method, we could control the search time, setting it 50% to 100% of the binary search time. Due to the controllable search time, there is a trade-off between search time and success rate. In other words, the longer search time results in the higher success rate.