Hybrid Encoding Evolutionary Algorithm for Reversible Logic Synthesis

Abstract

Abstract Variable-length encoding evolutionary algorithm has been proved effective in small and medium-sized reversible logic synthesis. Variable-length chromosome is adopted because the length of optimal reversible circuit is unknown. However, it leads to huge search space. The improved two-stage hybrid encoding synthesis is proposed to reduce the search space and enhance the feasible ratio. Fixed and variable-length encoding are applied in the two stages respectively. Estimated maximum length is obtained through heuristic information from the positive polarity Reed Muller spectra and the fixed chromosome length in the first stage is set to a value less than the maximum one. When the population converges in the first stage and the feasible solution has not found, the second stage starts to work. It employs the original variable-length encoding algorithm, but the optimization problem is simplified through the reduced positive polarity Reed Muller spectra of the best solution from the first stage. Experimental results show that both the fixed-length encoding and the simplified fitness function in the two stages both can decrease the search space and save the synthesis time.