Abstract: In this paper, the problem of modeling ultrasonic range finder under uncertainty is described.-The inherent uncertainty in the sensor demands a "soft" and robust approach to modeling the problem. Based on adaptive least square (ALS) technology, a robust adaptive least square modeling (RALSM) method is presented. First, we take advantage of on-line adaptive filtering technology to kick out the error data. Then apply with fuzzy technology; we develop a RALS model for ultrasonic ranger finder. Finally, based on error analysis, a evaluation method to the proposed model is also presented. Experiments demonstrate the RLAS model with high feasibility, robustness and high self-adaptability as well.