Aiming at the problem that the performance of ALV (average landmark vector) algorithm for mobile robots is greatly affected by natural landmarks, an optimized algorithm is proposed. By utilizing the image feature detection and matching algorithms (such as scale-invariant feature transform and speeded-up robust feature) to obtain natural landmarks, the optimized algorithm firstly disassembles the original ALV algorithm and obtains the home sub-vectors. Then, the contributions of the home sub-vectors are adjusted and the mismatching landmarks are eliminated by using the statistical theory. Finally, the home sub-vectors that contain weight information are integrated into the home vector pointing to the target location. Experiments show that the optimized ALV algorithm can effectively improve the overall accuracy of the natural landmarks and ensure the correspondence of the landmarks, so as to improve the accuracy of the ALV algorithm and make the robot reach the target location autonomously with a more ideal trajectory.