Research Of Intelligent PID Control For Congestion In ATM Network

The demand for higher transmission capacity and higher bandwidth to the user by multiservice and multimedia application, has put Asynchronous Transfer Mode (ATM) based on architectures at the forefront of several competing switch technologies. ATM is expected to support various types of multimedia traffic (such as data, voice or video applications) with different bit rates and quality of service (QoS) requirements. Due to the unpredictable fluctuations and bushiness of traffic flows within networks, congestion may occur frequently. Therefore, it is necessary to design the proper congestion control mechanisms to ensure the promised QoS.One of the important characteristics of ATM networks is that the ATM networks support several service types to meet applications with different QoS. In UNI 4.1, the service types are Constant Bit Rate (CBR), Variable Bit Rate (VBR), Available Bit Rate (ABR), and unspecified Bit Rate (UBR). Among these services, CBR and VBR services guarantee both the bandwidth that a connection will receive and the maximum delay that will be incurred by its cell. Different from them, ABR connections can only utilize the residual bandwidth of CBR and VBR connections. Hence, congestion control of ABR service is especially important.There are two different schemes to implement the ABR service congestion control, binary rate feedback (BRF) scheme and explicit rate feedback (ERF) scheme. Because the switch performance lies on the simplicity of the algorithm in high-speed ATM networks, many switch manufacturers choose the BRF scheme that is simpler.The classical congestion control algorithms for binary ABR service mainly based on heuristics and do not have a formal theory to support them, thus they have two drawbacks. One is that the queue length and allowed cell rate (ACR) oscillate with great amplitude, the other is that the control performance is not as good as we had expected in networks with a long bandwidth delay. Considering these, the paperstudies the congestion control of binary ABR from the view of the modern control theory. The main contributions of this paper are as follows:Firstly, the binary ABR congestion control model is deduced based on the flow theory. Subsequently, the nonlinear model is locally linearized at equilibrium point. With this linear model, the performance of standard EFCI algorithm that is used in BRF scheme is evaluated. The conclusion is that oscillations are caused by the nonlinear component existed in EFCI algorithm, but not intrinsic attribute of the binary congestion control mechanism. Therefore, the BRF scheme can be expected a bright future.Secondly, based on the linear FID controller, the paper does research on congestion control of binary ABR service. The simulation results show that this controller constrains the oscillation of ACR and the queue length of switch. In some degree, the performances of steady and instantaneous state have been improved. After this, three reformative FID controllers are put forward in this paper. They are Integral-Separated FID Controller Shift Integral FDD Controller and Differential Ahead FID Controller. They further improve network performances, but the enhancement is not good enough.Thirdly, through the combination of fuzzy logic and FID, the paper designs three fuzzy FID controllers and applies them in congestion control of binary ABR service successfully. They are Direct-action-type Fuzzy FDD Controller, Fuzzy Self-tuning FED Controller and Fuzzy Immune FID Controller. The simulations show that these kind controllers not only include the fine performances of the traditional FID controller but also do well when great disturbances are met with or the parameters of the model are changed.Fourthly, through the combination of neural network and FED, the paper designs three neural network FID controllers and applies them in congestion control of binary ABR service successfully. They are Signal-Neuron FED Controller, New Neuron-PID Controller and Fuzzy-PED Controller based on NN. They obtain excellent control e...