在這篇論文中，首先我們提出兩種智慧型(包括乏晰與類神經乏晰)訊務塑型器(TS)-使用參數控制器(UPC) ，來監控一個連線的平均封包速率。它是由兩個智慧型漏水桶法所構成的TS和UPC連結而成。此智慧型漏水桶法含有一個智慧型水增額(increment)控制器，它利用此連線的長期平均封包速率和短期平均封包速率，以及一組語言性法則，來調整水增額的量。水增額的調整可以在選擇性、反應性、和平均佇列延遲方面有較佳的表現。模擬結果顯示智慧型TS-UPC在選擇性方面略優於傳統(非智慧型)TS-UPC，但智慧型TS-UPC在反應性與平均佇列延遲方面的效能明顯優於傳統TS-UPC。其中，類神經乏晰TS-UPC的效能又比乏晰TS-UPC來的好。
接下來我們提出兩種類神經乏晰處罰方案來處罰違反平均封包速率的連線--類神經乏晰線性處罰方案和類神經乏晰非線性處罰方案。在這兩種類神經乏晰處罰方案中，除了採用效能最好的類神經乏晰TS-UPC外，並強加一個額外的封包丟掉機率，於一個類神經乏晰UPC所認定的違法連線之封包丟掉機率上。類神經乏晰線性處罰方案將一個處罰控制法引入UPC所使用的類神經乏晰漏水桶法中，使得此類神經乏晰漏水桶法有能力丟掉一個違法連線的一些合法封包，以作為處罰。而類神經乏晰非線性處罰方案則是將一個水增額函數控制器(IFC)引入類神經乏晰TS和類神經乏晰UPC兩者所使用的類神經乏晰漏水桶法中。水增額函數控制器利用函數來調整水增額，使得一個違法連線的違法程度被類神經乏晰UPC的類神經乏晰漏水桶看起來比實際更為嚴重，結果就會有比原先更多的封包被丟掉。模擬結果顯示兩種類神經乏晰處罰方案除了在選擇性方面仍然非常接近離理想值外，在反應性與平均佇列延遲方面的也有極佳的效能。

In this thesis, we first propose two intelligent TS-UPCs, which are conjunctions of the TS (traffic shaper) and the UPC (usage parameter controller) for sustainable-cell-rate usage parameter control in QoS-Controlled High-Speed Networks. One is the fuzzy TS-UPC, which is composed of the fuzzy leaky bucket; the other is neural fuzzy TS-UPC,
which is composed of the neural fuzzy leaky bucket. The fuzzy/neural fuzzy increment controllers (FIC/NFIC), which are incorporated with the conventional leaky bucket algorithm in fuzzy/neural fuzzy leaky bucket
algorithm, properly choose the long-term mean cell rate and the short-term mean cell rate as input variables and intelligently compute to determine the increment value. Three performance measures such as
selectivity, responsiveness, and queueing delay are considered. The intelligent fuzzy leaky bucket algorithms can have performance better than the conventional leaky bucket algorithm. And the neural fuzzy leaky
bucket algorithm outperforms the fuzzy leaky bucket algorithm.
Next we propose two types of neural fuzzy penalty schemes for punishing SCR-violating connections -- the neural fuzzy linear penalty scheme and the it neural fuzzy nonlinear penalty scheme, which impose an additional cell
dropping probability onto the cell dropping probability of a violating connection experienced at UPC. The neural fuzzy linear penalty scheme introduces a penalty control algorithm within the neural fuzzy leaky bucket algorithm employed by UPC, enabling the neural fuzzy leaky bucket algorithm to drop the conforming cells of a violating connection as a
punishment, whereas the neural fuzzy nonlinear penalty scheme introduces an increment function controller (IFC) within the neural fuzzy leaky bucket algorithms employed by both neural fuzzy TS and neural fuzzy UPC.
IFC adjusts the increment to make the connection's violation degree worse than the neural fuzzy leaky bucket in neural fuzzy UPC, and as a consequence, cells will be dropped more than should have originally. Simulation results show that the neural fuzzy penalty schemes are almost identical to the ideal cell loss ratio in selectivity. And theb neural
fuzzy penalty schemes have great performance in responsiveness and in mean queueing delay.

Chinese Abstract.....i
English Abstract.....ii
Acknowledgement.....iv
5Contents.....v
List of Figures.....vii
List of Tables.....ix
1 Introduction.....1
2 Intelligent Sustainable-Cell-Rate Usage Parameter Contro for QoS-Controlled High-Speed Networks.....4
2.1 Introduction.....4
2.2 Leaky Bucket Algorithm.....9
2.3 Fuzzy Leaky Bucket Algorithm.....11
2.4 Neural Fuzzy Leaky Bucket Algorithm.....15
2.4.1 Structure of NFIC.....16
2.4.2 Reinforcement Learning.....19
2.5 Simulation Results.....22
2.6 Concluding Remarks.....29
3 Two Neural Fuzzy Penalty Schemes for Punishing SCR-Violating Connections in QoS-Controlled High-Speed Networks.....31
3.1 Introduction.....35
3.2 Two Penalty Schemes.....35
3.2.1 Two design Objectives.....35
3.2.2 Neural Fuzzy Leaky Bucket Algorithm.....37
3.2.3 The Neural Fuzzy Linear Penalty Scheme.....38
3.2.4 The Neural Fuzzy Nonlinear Penalty Scheme.....39
3.3 Simulation Results and Discussions.....41
3.3.1 Simulation Environment.....41
3.3.2 Selectivity.....43
3.3.3 Responsiveness.....48
3.3.4 Queueing Delay.....49
3.4 Conclusion.....52
4 Conclustion.....53

" Trffic control and congestion control in B-ISDM," ITU-T, Recommendation I.371, May 1996.
E. P. Rathgeb, "Modeling and performance comparison of policing mechanism for ATM networks," IEEE J. on Selected Areas in Commun., vol. 9, no. 3, pp. 325-334, Apr. 1991.
L. Dittmann, S. B. Jacobsen, and K. Moth, "Flow enforcement algorithms for ATM networks," IEEE J. on
Selected Areas in Commun., vol. 9, no. 3, pp. 343-350, Apr. 1991.
S. Shioda and H. Saito, "Satisfying QoS standard with combined strategy for CAC and UPC,"IEEE ICC'95},
pp. 965-969.
M. Butto, E. Cavallero, and A. Tonietti, "Effectiveness of the `leaky Bucket' policing mechanism in ATM networks," IEEE J. on Selected Areas in Commun.}, vol.9, no.3, pp. 335-342, Apr. 1991.
Zohnshiun Eul, "Fuzzy leaky bucket algorithm for usage parameter
control in ATM networks", M.S. thesis, Department of Communication
Engineering, National Chiao Tung University, Taiwan, 1998
C. T. Lin, and C. S. G. Lee, Neural Fuzzy Systems New York, Prentice-Hall, 1996
H. J. Ziemmermann, Fuzzy Set Theory and its Applications, 2nd ed. New York: Kluwer,1991, pp. 11-17.
T. D. Ndousse, "Fuzzy neural control of voice cells in ATM networks,"IEEE J. on Selected Areas in Commun., vol. 12, no. 3, pp. 1488-1494, Dec. 1994.
V. Catania, G. Ficili, S. Palazzo, and D. Panno, "A comparative analysis of fuzzy versus conventional policing mechanisms for ATM networks," IEEE/ACM Trans. on Networking., vol. 4, no. 3, pp. 449-459, June 1996.
A. A. Tarraf, I. W. Habib, and T. N. Saadwi, " A neural fuzzy controller using reinforcement learning method for ATM traffic policing,"IEEE MILCOM'94, pp. 718-722, 1994.
R. G. Garroppo, S. Giordano, S. Miduri, and F. Russo, "A prediction based UPC mechanism for VBR video traffic,"Proc. IEEE ICC'98}, pp. 1119-1123, 1998.
F. Guillemin, C. Rosenberg, and J. Mignault, "On characterizing an ATM source via the sustainable cell rate traffic descriptor,''IEEE INFOCOM '95, pp.1129-1136.