1997 · 74 citations · 2 references
A Genetic Algorithm is presented which is capable of evolving 100 % functional arithmetic circuits. Evolved designs are presented for one-bit, two-bit adders with carry, and two and three-bit multipliers and details of the 100 % correct evolution of three and four-bit adders. The largest of these circuits are the most complex digital circuits to have been designed by purely evolutionary means. The algorithm is able to re-discover conventionally optimum designs for the one-bit and two-bit adders, but more significantly is able to improve on the conventional designs for the two-bit multiplier. By analysing the history of an evolving design up to complete functionality it is possible to gain insight into evolutionary process. The technique is based on evolving the functionality and connectivity of a rectangular array of logic cells and is modelled on the resources available on the Xilinx 6216 FPGA device. Further work is described about plans to evolve the designs directly onto this device.
2