In Japan, repletion rate of guide dogs for visually-handicapped persons is at very low percentage of roughly 10% nationwide due to required long training period as well as uneconomic price in obtaining a guide dog. Motivated by these two reasons, we are currently developing a Guide-Dog Robot with major objective given to developing intelligence in human-guide-dog robot interface. In this paper, novel human-machine interface algorithm and strategy to lead visually-handicapped people is proposed. A new leading edge searching method intelligently incorporated with one leaser range finder (LRF) is developed and implemented in our prototype guide-dog robot. It is demonstrated in our experiment that with the new method, the robot not only is able to effectively recognize the led-person's walking conditions but also is able to accurately estimate the person's next move. With the new method and a LRF, robustness is improved and validated to allow the guide-dog robot and the led-person to maneuver in complex corridor environment.