Asynchronous Transfer Mode (ATM) is a important transfer technique in the Broadband Integrated Services Digital Networks (ISDN). In the ATM networks, it is necessary to provide a fast and efficient transfer technique because of using wavefiber and transfering more information. As ATM needs to support many services, and the networks is large, policing and congestion must be adopted to monitor the traffic. Then the network resources can be used efficiently. It is difficult to use conventional schemes to get satisfying results for imprecise quantities and varities in the ATM networks. This paper applies "soft computing" theory to the ATM networks. It mainly researches ATM switching, ATM policing and congestion control. This paper proposes some new schemes to improve the networks and to use network resources efficiently. The main works and contribution of this paper include: 1. Firstly, it introduces the ATM networks and thoery of "soft computing" in detail. Especially it mainly describes the parts with relations to our researchs. It analysises the features of the ATM networks. It points that it is feasible and advantageous to apply the thoery of "soft computing" to the ATM networks, because of fast switching, much information to be tranfered, its imprecise quantities and varities in the ATM networks. We outline the objects and methods of our research. We also review the development history of the research field. So people can see that our research is vary important and advanced. 2. Secondly, We propose an ATM switching fabric with input buffer. It uses a cellular neural network to control cell scheduling. It adopts the longest- queue-first-allocation (LQFA) scheme to do optimal decision, so more cells can be switched during a time slot. It can improve the QOS of the networks. The cellural neural networks can do parallel computing. And its VLSI implementation is easily done. The algorithm cannot converge to a local optimal solution. At last, we prove the cellural neural network is convergent. 3. Thirdly, we propose an ATM switch Fabric with Fuzzy back-pressure function. It controls back-pressure function with the methods of intelligent control. It has a low cell loss rate (CLR), and has a little cell delay. With simulation results, the switching fabric proposed in this paper has a good QOS. It is a good application of"soft computing" to ATM networks. 4. Fourthly, a usage parameter control (UPC) scheme with fuzzy simulated annealing algorithm is proposed. It improves the conventional Leaky Bucket (LB). It learns the relations between the states of traffic and the parameters of controller by itself. The controller has a good selectivity and transparency. From simulations, we can see that the fuzzy simulated annealing controller has a better performance than the conventional LBs. It al
修改评论