The importance of an automatic aerial place control system is good established. In satellite communicating, it is of import that an aerial should continuously and automatically track the accurate place of an aerial so that the right response becomes possible.
In nomadic communicating systems, normally place of an aerial is set manually to optimise the public presentation of the communicating in a peculiar cell. The manual placement is clip devouring and labourious.Moreover, it may be non optimum. It is hence really of import that the place of an aerial be automated. There are many other applications where aerial place has importance.
In the past, many efforts have been made to automatize the place of an aerial. Proportional accountants, relative plus built-in plus derivative accountants and additive perceivers have been proposed. In this undertaking, we propose unreal nervous webs as the accountants. Artificial nervous webs have already found 1000s of successful applications in control system design.
In this undertaking we intend to utilize Multilayer Perceptron ( MLP ) architecture of nervous webs. These are feed frontward webs and are trained utilizing the mistake back-propagation algorithm.
The importance of the antenna place control system is good established. The aerial is most of import portion in communicating engineering, so it is really of import to automatize this system and besides it should be accurate or else the signals will non be received right.
In nomadic communicating system, the place of an aerial is set manually to optimise the public presentation in a peculiar cell. But manual placement is really clip consuming and arduous, besides the consequences may non be fulfilling. It is hence really of import to automatize the place of an aerial so that the signals can be received right and the system should besides be optimum.
A batch of research has been carried out in past few old ages to automatize this system, by utilizing different types of accountants. Most good known among them are relative accountants, relative plus built-in plus derivative accountants and linear observer based accountants. Each of these accountants have some advantages every bit good as some serious restrictions. In this undertaking we have proposed ANN ( unreal nervous webs ) as accountant. As we know unreal nervous webs has 1000s of successful applications in control field.
Now coming to the overview of this undertaking, the system layout is reasonably simple, the system works on nervous web ‘s mistake back-propagation algorithm, which is used by the MLP ( multilayer perceptron architecture ) .The error signal is produced by the difference of the two electromotive forces ; the input relative electromotive force and the end product relative electromotive force. These two electromotive forces are given to the summing junction from where the mistake or triping signal is given to the accountant, which in this instance is the MLP architecture of nervous webs.
The chief aims of the undertaking are as under:
- survey and analysis of the bing antenna place control systems
- Survey of the jobs associated with these accountants.
- Study of nervous webs
- Design and development of MLP web as antenna place control system
- Testing the web
In chapter 2, we have given a brief about the antenna place control system, how does it works and what type of accountants are used largely with it and what drawbacks they have. Chapter 3 is concentrating on the nervous webs. In chapter 4 all the consequences with the methodological analysis is discussed. And Chapter 5 concludes the consequence with some suggestions for future work.