Abstract: Motivated by the high-level regulation mechanism of endocrine system, an agent self-organization algorithm of behavior is put forward. This algorithm receives environmental information using a neural system, produces emotional factors (hormones) through an emotion learning model, then uses these factors to regulate the neural system. At last, the memorization and behavior mode of the neural system are exported to the genetic environment. So it avoids the self-study of neural system and gets better self-organization and self-adaptation power. In order to prove the validity of this algorithm, it is used to control an inverted pendulum. The result shows strong self-adaptation performance of the algorithm.