Home Profile Projects Family & Friends Artificial Intelligence
Communications Computer Leisure Farm Spiritual Corner Links

Master Degree Project (at Nanyang Technological University,Singapore)
Neural Networks in ATM Traffic Control

This research project presents an adaptive control scheme using a newly developed Minimal Resource Allocation Network (MRAN) to solve the traffic congestion problem in ATM networks. MRAN generates a minimal radial basis function neural network by adding and pruning hidden neurons based on the input data and is ideal for on-line adaptive control for fast time varying nonlinear systems.

The Asynchronous Transfer Mode (ATM) traffic modeling is carried out using the well-known network simulation package - OPtimized Network Engineering Tools (OPNET) for bursty, variable bit rate (VBR) and multiplexed traffic (combining both speech and video signals) where scenarios have been created for heavy traffic situations. Using these scenarios, performance of MRAN controller is compared with a Back-Propagation (BP) neural network controller and also conventional congestion controller. The aim of the controller is minimizing the congestion episodes and maintaining the quality. Simulation results indicate that MRAN controller performs better than the BP and conventional congestion controller in reducing the heavy congestion and maintaining a better quality of the traffic. Five performance measures have been highlighted in all these simulations namely, the cell loss rate (CLR), queuing delay, link utilization, traffic quality and a total cost function which combines them in a weighted way to make an overall comparison between all the control schemes.

The main objective of this research project is to develop an adaptive traffic control algorithm which is simple, robust and effective to be applied to the ATM network. First of all, analysis of various traffic management methods being developed and implemented have been carried out. Furthermore, the applications of neural networks that have been developed for the traffic management at the User Network Interface (UNI) level of the network has been studied. Then, conventional, BP and MRAN network have been integrated to various type of ATM traffic generated using OPNET network modeler. Here, special focus has been given to the MRAN algorithm which will give the faster and more efficient control. The understanding of Asynchronous Transfer Mode (ATM) is a prerequisite before undertaking such investigation extensively. Meanwhile, the performance of all these control schemes have been evaluated for the time-varying dynamic traffic system.

After implementing the first stage of Neural Congestion Controller Modeling using OPNET, now intensive investigation and study are carry out for the Neural ABR Flow Control. At the same time, the performance of this neural congestion controller will be compared with the Explicit Rate Indication with Congestion Avoidance (ERICA) and EFCI congestion control scheme. The main usage of the OPNET in the research study is the Neural Congestion Controller Design, Modeling which will lead to future possible real implementation.


Ng Hock Soon, N. Sundararajan and P. Saratchandran

Neural Congestion Controller for ATM using OPNET

(Distinguished Paper Award), Proc. of OPNETWORK'99, Washington DC, Aug 1999

Ng Hock Soon, N. Sundararajan and P. Saratchandran

Adaptive Neural Congestion Controller for ATM network with Heavy Traffic

Proc. of  IFIP TC 6 Fifth International Conference on BROADBAND COMMUNICATIONS '99 Hong Kong, Nov 1999 

Ng Hock Soon, N. Sundararajan and P. Saratchandran

ATM Congestion Control using Minimal Resource Allocation Networks (MRAN)

submitted to IEE Communications Proceedings, 2000

Bachelor Degree Project (at Univeristy Technology of Malaysia)
Development of an Education Toolkit Servomotor System
        using Personal Computer and Microcontroller

In this project, a dc servo motor system is developed is for educational purpose which uses a Proportional, Integral, Differentiator (PID) controller and also a Fuzzy Logic Controller. In order to achieve the above objective, both analog and digital control systems have been developed. First, an analog PID controller is built based on some specifications as shown below:

a)  Easy to operate with logical and functional front panel layout.
b)  Built-in 3 ½ digit Digital Voltmeter and low frequency signal generators.
c)  A compact bench mounted dc servo-motor apparatus for investigating the principles of speed control.
d)  Supports a wide range of control principles at all academic levels, including controller design and velocity (rate) feedback.

Analog and digital control techniques are applied to the whole world of control. It is for this reason why it is important to obtain the practice in designing, planning and tuning both analog and digital controller. The dc servo motor system has been chosen to be the test bed of these controllers due to their fast in control system response. Meanwhile, Proportional, Integral & Derivative (PID) and fuzzy logic control algorithm are applied to control the dc servo motor. The whole work is then integrated to build out an educational tool-kit for PID and fuzzy logic control system.

Analog PID controller is built using operational-amplifier, capacitors and resistors while digital PID controller is developed using the universal controller board 80C196 with interface to a Personal Computer.

In the absence of a system mathematical model, a fuzzy system model is described which is analogous to a human operatorís behavior. A set of linguistic fuzzy control rules are set up which are conditional linguistic statements that establish the relationships between the inputs and the outputs. Tools and techniques to generate optimized fuzzy based real time code in C, with short development time, is shown for the Intel 8XC196 microcontroller. Performance and features of the 80C196 for fuzzy-based control are analyzed.

The transient and steady state responses of PID and fuzzy control algorithm have been analyzed to observe their differences.

Back To Top
Hosted by www.Geocities.ws