Abstract: Mobile robot localization is an important topic in the intelligent robot domain. It includes sensory techniques and localization algorithms. An advantage of Markov localization is that the localization algorithm provides a means of estimating robot position under global uncertainty. In this approach a probability density distribution is employed to represent belief of robot position. The belief distribution is updated by receiving the sensory data and reading the record of robot moving. The maximum of the belief is applied to estimate robot position. However, in a symmetrical environment a robot using distance detector cannot find its position by means of Markov localisation alone. In order to solve this problem, a robot equipped with a gyroscope or a compass is proposed, and the Markov localization algorithm is extended with a new perceptual model that is constructed using an angle Gaussian distribution defined in this paper. Experiments for simulating different symmetrical environments have been done. The experimental results show that the new algorithm is feasible and that a robot with the new algorithm can quickly estimate its position when moving in a symmetrical environment. In addition, the algorithm converges at the real position in an asymmetrical environment faster than regular Markov localization.