Artificial Neural Network Based Monitoring, Control and Optimisation

Modelling of Complex Systems using Computational Intelligence Approaches
This project involves study of computational intelligence (CI) approaches for the purposes of modelling the characteristics of
non-linear complex systems. The CI based system modelling consists of a systematic procedure involving:
 appropriate data generation and acquisition,
 data pre-processing,
 normalisation of data,
 selection of appropriate combination of input-output variables
 selection of appropriate topology
 network training using an efficient algorithm
 regularisation to reduce the effect of noise
 and model validation.
The student will be required to research thoroughly all the above aspects. A set of typical example problems will be used to
make a comparative evaluation of important modelling approaches proposed in the literature. Finally, the most appropriate
modelling strategy will be selected for developing a life-long learning system that can facilitate teaching of a robot by
demonstration in the following way:
A learning network can be designed and developed for a mobile robot, so that it emulates the life-long learning behaviour
of a human being. For this, training data needs to be generated by first moving the robot around in a desired way. The data
will include motion commands and reading of all the sensors. A network will then be trained to predict next command, so
that robot can safely move around in its environment. It can also be updated through on-line learning.
Resources Needed
A PC, MATLAB and its toolboxes, a robotic system.