A quick road-boundary detection algorithm using 2D laser rangefinder based on maximum entropy principle is proposed. By reclassifying observed measurements at different measure angle in limited continuous domain

we calculate fuzzy cluster centers directly and predict the states and predictable error simultaneously

and calculate membership values of current state based on the maximum entropy principle. The method eliminates the dependence of conventional methods on hnowledge of experts and prior information. The weighted factors are designed for calculating prediction results based on the fact that later states have more impacts on the prediction. In the process

the error of every step is also predicted to modify the prediction results of states

and the prediction error is determined by the selected states in the limit domain. Thus the accumulation error overtime is eliminated. The results of experiments indicate the effectiveness of the proposed algorithm.