Bacteria foraging optimization algorithm is a kind of group evolution algorithm inspired from E. coli foraging phenomenon, and it has good global optimization, strong robustness, simple arithmetic, and so on. But it is also easy to premature and slow convergence speed and other shortcomings. According to its shortcomings, it is proposed that an improved bacteria foraging optimization algorithm. The improved algorithm accelerates the convergence speed and avoids the premature in a certain extent. It is applied to on-line self-tuning of PID parameters that the original algorithm and improved algorithm. And through the matlab simulation, it is proved that the superiority of the improved algorithm.