Abstract: Temperature is an important factor for affecting the accuracy of ring laser gyroscope (RLG). The relationship of RLG's bias to temperature should be established accurately which can improve the precision of RLG strapdown inertial navigation system. Temperature characteristic of RLG was introduced. Temperature compensation method for RLG's bias based on improved artificial fish swarm algorithm (IAFSA) was established and the steps and methods were given. The traditional modeling method of stepwise regression was also investigated to provide a comparison with the IAFSA. The result shows that the temperature compensation model by IAFSA is accurate. The bias instability of RLG output after compensation is 0.001 85 ()/h and its precision is 15.5% which is higher than that of the traditional stepwise regression. The results of two typical temperature tests show that temperature compensation model by IAFSA can compensate RLG's bias real-time.