We introduce a method, based on a genetic algorithm (GA) approach, to design combinational logic circuits. This problem is quite difficult for a traditional GA, but we have overcome these difficulties and have implemented a computer program that can automatically generate high-quality circuit designs. We describe the important issues to consider when solving this circuit design proble: the importance of the representation scheme, the encoding function, and the definition of the fitness function. We present several circuits derived by our system under various assumed constraints, such as the maximum number of allowable gates and the types of available gates. We compare the solutions produced by our system against those generated by a human de-signer. We also show that our representation approach, when compared to a standard binary encoding, produces better performance both in terms of quality of solution and in terms of speed of convergence.
展开▼