Optimization of constellation labeling using genetic algorithm

Constellation labeling is an one-to-one mapping relationship associating 0–1 bit sequences and signal points. Different constellation labeling results in different bit error rate (BER) performance. Some works have certified that the Binary Reflected Gray Code for M-PAM, M-PSK and M-QAM can give the lowest possible average BER. But there are few works considering about the best labeling for other specially designed constellations. This paper applies a genetic algorithm to find the optimal and hypo-optimal labeling for a special constellation. This kind of constellation has been proved to have the lowest average symbol error rate (SER) under AWGN channel.