A single landmark based localization algorithm for unicycle mobile robots was provided in [1]. It is based on the algebraic localizability notion and an efficient differentiation algorithm in noisy environment ([2], [3]). Let us stress that this localization algorithm do not need to know the linear and the angular velocities which are reconstructed by this algorithm using the kinematic model. In this paper, a sensibility study leads to a new fusion algorithm in the multi landmark case us- ing as a basis our posture differentiation based estimator. Some simulations and experimental results are presented in order to prove the effectiveness of the proposed method compared to the well known EKF method.