A New Weighted Voting Algorithm Based on Learning Automata

A New Weighted Voting Algorithm Based on Learning Automata

Elham Bahramsari1 Abbas Karimi2

1) graduate of architecture of computer systems ,
2) Professor of computer system engineer,

IOI Code : XYCC-ABGDD
Publication : International Conference on Science and Engineering(icesconf.com)
Abstract :
    Fault – tolerance is vital in industry, medicine and defensive industry. Fault – tolerance means to assure proper safety and availability. One solution is to apply the redundancy to system modules. In this case, out of different results achieved through input modules, the best of which must be selected by a variable called voter. Therefore, selecting a proper voter in critical system is very important. Voters are divided into different kinds based on their functions, one of the most applicable voters is weighted average voter. In this study, a new weighted average voter is proposed based on learning automata. Weighted average voters require time – consuming calculations and high time complexity. In this study, assigning weight to modules is preformed intelligently, using learning automata. Therefore, there is no need for performing complicated calculations and subsequently, the performance time will be reduced. On the other hand, some voters do not present proper safety and availability which have a critical role in the correct performance of the systems. The experimental results demonstrate that the proposed voter improves the performance of these two parameters strikingly in the presence of permanent and intermittent faults.

Keywords : weighted average voting algorithms learning automata safety