Title :
Evolution of fault-tolerant and noise-robust digital designs
Author :
Hartmann, M. ; Haddow, P.C.
Author_Institution :
Dept. of Comput. & Inf. Sci., Norwegian Univ. of Sci. & Technol., Trondheim, Norway
fDate :
7/18/2004 12:00:00 AM
Abstract :
Artificial evolution has been shown to generate remarkable systems of exciting novelty. It is able to automatically generate digital circuit designs and even circuits that are robust to noise and faults. Extensive experiments have been carried out and are presented here to identify more clearly to what extent artificial evolution is able to generate robust designs. The evolved circuits are thoroughly tested whilst being exposed to noise and faults in a simulated environment and the results of their performance are presented. The evolved multiplier and adder circuits show a graceful degradation as noise and failrate are increased. The functionality of all circuits is measured in a simulated environment that to some extent takes into account analogue electronic properties. Also included is a short overview of some recent work illustrating the robustness and tolerance of bio-inspired hardware systems.
Keywords :
computer architecture; data compression; evolutionary computation; fault tolerance; hardware-software codesign; image coding; bi-level images; context switchable inference processor; design automation; evolutionary algorithm; evolvable hardware; fault-tolerant digital designs; fuzzy inference processor; lossless compression; noise-robust digital designs; very high resolution images;
Journal_Title :
Computers and Digital Techniques, IEE Proceedings -
DOI :
10.1049/ip-cdt:20040014