The invention provides an autonomous positioning method used for a robot and implemented on the basis of an ARMA model. The autonomous positioning method comprises the following steps: performing parameter estimation on data and state of a map in a long-term changing environment by virtue of an autoregressive moving average model (ARMA), and ultimately building a changing map model of the environment and in a probability grid form; performing follow-up prediction on a grid time domain of the built map model, and evaluating prediction confidence, so that a predicted grid map provided with a prediction confidence label is obtained; and finally fusing real-time observation information into the predicted grid map by adopting Bayes rules, so that an updated environment map available for the robot is obtained, and realizing autonomous long-term positioning in the changing environment by the robot by combining a particle filter algorithm under the map. Therefore, the effect that the robot hasrelatively high positioning accuracy and positioning robustness in the long-term changing environment is achieved.