Neural/Fuzzy Access Attempt Control in cdma2000

Abstract

於第三代無線行動通訊規範IMT-2000中，共通過五種規格，其中屬於美規的cdma2000系統為第二代IS-95窄頻CDMA系統的寬頻延伸版本；本篇論文旨在對cdma2000規格中的接取意圖（access attempt）程序做控制研究，此程序主要應用於手機與基地臺間的建立連線、手機註冊及發送簡訊。 首先，我們運用乏晰邏輯（fuzzy logic）設計接取意圖控制系統，稱之乏晰接取意圖控制器（FAAC），此控制器可依照不同的系統表現狀況而動態給定接取參數（access parameter），這些參數包括：接取機率（access probability）、初始功率（initial power）及功率增值（power increment）。 接著，我們提出以類神經網路（neural network）所建構的類神經╱乏晰接取意圖控制器（NFAAC），因其加強性學習演算法（reinforcement learning algorithm）能更進一步地根據系統迴饋的加強性訊號（reinforcement signal）以給定接取機率。最後，我們得到乏晰接取意圖控制器模擬的結果相對於傳統的方式在流通量（throughput）、接取延遲（access delay）、阻隔率（blocking rate）及功率消耗（power consumption）的幾個方面都有較佳的表現；再者，類神經╱乏晰接取意圖控制器則在各方面上優於乏晰接取意圖控制器與傳統的方式。

In this thesis, we study access attempt control for cdma2000 cellular system. The access attempt mechanism is applied when the cellular system performs call setup, registration, and burst data transmission. We propose an access attempt controller with fuzzy logic system, named fuzzy access attempt controller (FAAC). By taking system performance parameters into consideration, FAAC can assign adequate access parameters including access probability, initial power, and power increment for mobile stations to achieve higher performance in terms of throughput, delay, blocking rate, and power consumption. Next we propose the neural/fuzzy access attempt controller (NFAAC) with the reinforcement learning algorithm in order to avoid fluctuation of access probability in FAAC. The NFAAC adopts a critiron to be the learning signal then outputs proper access parameters for mobile stations. Simulation shows that FAAC performs better than the conventional one in overall performance, and NFAAC overrides FAAC more.