| 摘要: | 在無線通訊頻道(Channel)配置的運用上，可區分為固定式(Fixed)與動態(Dynamic)，因為通訊系統在尖峰時段(Hot-sports)會出現話務量大增導致服務系統負載不平衡(Load Unbalance)，即系統內部份的基地台工作量過大而所配置的頻道匱乏而另一部份的基地台卻只執行少量的工作且配置之頻道多閒置未用。當發生負載不平衡的情況時，將造成整個系統的話務阻塞率(Blocking rate)及話務中斷率(Dropping rate)增加，而降低通訊系統的服務品質，因此如何動態地控制系統內部各基地台達到負載平衡，以求系統最佳的整體服務工作效率，是一個相當重要的問題。針對此問題已有相當多的學者進行研究，發表了許多相關的論文。傳統控制的設計方式必須以數學模型來描述受控系統，但是系統中所有變數一定要徹底了解才能完成、對於龐大、非線性、或難以量測等情況的受控系統，考慮的變數越多，所要處理的問題越複雜，因此該系統數學模式的建構就越不易，以致於取得的數學模式與真實系統相差甚遠，所設計出來的控制器不易滿足預期的要求，這種傳統的方法會造成系統的震盪使系統整體效率變得更糟。 為了彌補傳統設計的缺點，我們提出利用Fuzzy Logic Control(FLC)，簡化系統設計的複雜性，特別適用於非線性、時變、模型不完整的系統上，模糊控制是一種容易控制、掌握的非線性控制器，具有適應性、強健性(Robustness)、容錯性(Fault Tolerance)，有鑑於此我們將此問題歸納整理之後, 提出針對動態負載平衡及頻道配置的設計及策提出新的解決的方法及策略，亦即是Cell load decision making、Cell Involved Negotiation以及Multi-channels Migration。此新研究及策略可使通訊系統具有快速選擇且準確的方式來判斷系統之負載狀態，且根據不同的輸入變數對系統產生的影響及關係，利用各基地台的之負載狀態指標來決定在各基地台間之頻道借用(Channel Borrowing)方式與數目，在此研究過程中，我們針對數種不同方式的通訊服務需求及不同系統情境的頻道借用演算法(Simple borrowing、Direct retry、CBWL、LBSB )，提出表現更佳之新演算法Fuzzy based Dynamic channel borrowing Schemes (FDCBS)。我們所提之系統設計在效能評估上有最少的話務阻塞率(Blocking rate)及系統話務中斷率(Dropping rate)、除此之外還可改進通訊系統間取得頻道之訊息交換的複雜度暨數量，及取得可用頻道延遲時間(Channel acquisition delays)。 利用模糊邏輯控制定義出每個負載指標的歸屬函數，可增加系統負載評估整體之客觀性，其次當負載過重時利用糢糊邏輯控制決定出頻道配置的數目，以降低當兩基地台間負載差距過大時所不斷發出頻道配置要求的次數。當頻道配置動作完成後，如果系統負載並沒有明顯改善，我們提出了結合類神經網路學習(Neural Network Learning)的特性，根據基地台的狀態動態的增加或減少基地台之負載指標及修正糢糊歸屬函數輸出及輸入的中心值，如此系統便能夠更精準且隨著系統在不同的話務量及時間作不同的負載屬性及決定頻道配置的數目。為了驗證我們所提出方法之執行效率，我們除了比較前所敘述的四種傳統頻道配置方法及FDCBS外，研究設計了一個新的具有適性化頻道配置且具動態負載平衡機制的方法Neural-Fuzzy Controller for Dynamic Channel Borrowing Scheme (NFDCBS)。我們所設計的動態頻道配置及負載平衡系統是一個極有效率的策略及方法，經各種實驗結果顯示我們所提出的策略及方法不但準確，而且可以積極提升通訊系統的效能，更可適用於下一代通訊系統之頻道配置及負載平衡問題。 In the third generation of wireless cellular system, adaptive dynamic channel borrowing is presented to maximize the number of served calls in wireless cellular networks. This has led to an intensive research in mobile computing to provide mobile users accesses to an internet. Efficient use of limited radio channels with a simultaneous increase of traffic capacity requires proper channel assignment. The uneven traffic load may create hot-spot cells and possibly causes a high blocking rate in hot-spot cells. Most conventional methods use load indices with a threshold value to determine the load status of a cell. However, those exists a ping-pong effect, as loads are around the threshold value. This result causes an unstable system and unnecessary message passing overhead. In addition, the estimation of traffic load is difficult and time-consuming. Thus, an intelligent prediction mechanism is needed. In this dissertation, a fuzzy-based dynamic channel-borrowing scheme (FDCBS) is presented to maximize the number of served calls in a distributed wireless cellular network. We develop a method to predict the cell load and solve the channel-borrowing problem based on the fuzzy logic control. A new channel borrowing algorithm with multi-channels borrowing is also presented in this dissertation. A borrowing mechanism supporting the present facility has been built on the application-level of wireless cellular networks. Performance results show that the FDCBS has better adaptability, robustness, and fault-tolerant capability thus yielding better performance compared with other algorithms. Fuzzy logic control can deal with perceptual uncertainties in classification problems. In order to design an adaptive channel borrowing in wireless cellular networks, it is an important task to construct the fuzzy membership function for each attribute and generate fuzzy rules from training instances for handling a specific classification problem in wireless cellular networks. However, there are some drawbacks in the proposed DFCBS: (1) Need human experts to predefine initial membership functions, i.e., these methods can not construct membership functions from the training data set fully automatically. (2) Too complicated and need a lot of computation time, and generate too many fuzzy rules. Neural-fuzzy controllers, based on a fusion of ideas from fuzzy control and neural networks, possess the advantages of both neural networks (e.g., learning abilities, optimization abilities, and connectionist structures) and fuzzy control systems (e.g., human- like IF-THEN rule thinking and ease of incorporating expert knowledge). In this way, we can bring the low-level learning and computational power of neural networks to fuzzy control systems and also provide the high-level, human-like IF-THEN rule thinking and reasoning of fuzzy control systems to neural networks. In brief, neural networks can improve their transparency, making them closer to fuzzy control systems, while fuzzy control systems can self-adapt, making them closer to neural networks. We shall focus on structure learning and parameter learning in neural-fuzzy controllers, and examine cases when they can take place sequentially and separately in two phases, when only parameter learning, or only structure learning is necessary, when they can occur simultaneously in a single phase, and when they can be performed with reinforcement signals. All these variations of structure and parameter learning of neural-fuzzy controllers are discussed with simulation examples. In order to adaptive for channel-borrowing in wireless cellular networks, a neural-fuzzy controller for the dynamic channel-borrowing scheme (NFDCBS) is presented to provide multimedia services and to support increasing number of users. In a cellular network, the call arrival rate, the call duration and the communication overhead between the base stations and the control center are vague and uncertain. Therefore, we propose a new efficient dynamic-channel borrowing for load balancing in distributed cellular networks based on NFDCBS. The proposed scheme exhibits better learning abilities, optimization abilities, robustness, and fault-tolerant capability thus yielding a better performance than other algorithms. It aims to efficiently satisfy their diverse quality-of-service (QoS) requirements of multimedia traffic. The results show that our algorithm has lower blocking rate, lower dropping rate, less update overhead, and shorter channel-acquisition delays than previous methods. Through simulations, we evaluate the blocking rate, update overhead, and channel acquisition delay time of the proposed method. The simulation results demonstrate that our algorithm has lower blocking rate, less updated overhead, and shorter channel acquisition delays. |
