Abstract: To solve the basic particle filter in the unknown space localization method of particle degradation problems, an improved particle filter algorithm based on MCMC was proposed. This particle filter robustness was enhanced. The localization simulation test using improved particle filter was developed which was applied in coal mining robot in unknown underground space. The simulation test results showed that the localization using improved particle filter had more locating accuracy in unknown underground space and better computational real-time ability than extended Kalman filter. This method has solved the pre-localization problem of coal mining robot underground.