DS/CDMA/FRMA第三代無線通訊系統之乏晰/類神經壅塞控制

Abstract

在第三代無線通訊系統中，能否提供較大的頻寬、較高的傳輸速率及傳送多媒體資料已成為主要的設計考量。其中分碼多工擷取（CDMA）的系統由於具有高度頻寬使用效率、抵抗多路線干擾的能力等，已成為未來無線以通訊系統的一個選擇。而對於直接序列-分碼多工擷取（DS - CDMA）來說，系統的容量是受到干擾限制的，如何去控制干擾量就是一個重要的課題。 在本論文中，我們首先提出一個在直接序列-分碼多工擷取╱時框保留多工擷取（DS - CKMA / FRMA）環境下的平行迴路類神經綱路（pipeline recurrent neural network）干擾預測器。有了這個干擾預測器的幫助，對於系統干擾量的控制將會更有效。再來我們提出一個乏晰╱類神經壅塞控制器，其內包含了平行迴路類神經綱路干擾預測器、乏晰（fuzzy）效能指標器及乏晰允諾機率控制器。藉由預測的干擾值及系統效能指標，乏晰允諾機率控制器可以調整允諾機率來控制競爭使用者的人數以使干擾量低於某個標準。由模擬的結果中，我們可以看到平行迴路類神經綱路干擾預測器在預測誤差及收斂速度上有較好的表現。而我們的壅塞控制器在傳送失敗率（corruption ratio）、語音封包漏失率及適應傳送資料特性改變的能力上，都有較好的表現。

In the third generation wireless communication, a system which has wideband, high speed, and ability to transmit multimedia traffic is primarily concerned. The CDMA (code division multiple access) systemis a good choice due to its high spectrum efficiency, soft capacity (or graceful degradation), multipath resistence, anti - jamming ability, inherent frequency diversity, and no cell planning requirement. Because for direct - sequence CDMA (DS - CDMA), the capacity is limited by interference level, the control of interference level is an important design issue. In this paper, we first propose a pipeline recurrent neural network (PRNN) interference predictor for DS - CDMA / FRMA wireless systems. With the aid of predicted interference, the control of interference level can be more effectivly. Second, we propose a fuzzy / neural congestion controller which includes the PRNN interference predictor, fuzzy performance indicator, and fuzzy permission probability controller. According to the predicted interference and performance indicator, the congestion controller can adjust the permission probability to control the number of contention users and let the interference level be under a level. From simulation result, we can seethat the PRNN interference predictor has better performance than RLS interference predictor in prediction error and the improvement of prediction is about 7%. We can see also the PRNN interference predictor takes less time to convergence by comparing with RLS interference predictor, and the improvement is about 25 sec. And our congestion controller has better performance than DS - CDMA / PRMA with channel access function in corruption ratio, voice packet dropping ratio, and adaptive property to change of traffic source characteristic. To set the corruption ratio and voice packet, dropping ratio to the lvel of 0.01, our algorithm has the improvement about 36.67% and 15.15%, repectively.