Nowadays, Monte Carlo Localization (MCL) algorithm is the most commonly used algorithm for a mobile robot to localize itself automatically, and has shown good performances because of its ability to model arbitrary distributions and robustness towards noisy input data. This paper presents an improved Monte Carlo Localization algorithm incorporating an error correction vector, which is calculated from the sensor's error information, and therefore, can achieve better performances in both accuracy and efficiency. Through practical application of the method which incorporates the error correction vector when conducting the program's prediction stage, the particles move closer to the real robot's position. Experimental results show that the proposed algorithm can increase the probability for particles to better estimate the robot's position.