A key challenge when developing a mobile robot navigation mechanism is to overcome the dead-end problem with low cost. However, the existing low-cost methods often generate invalid paths or redundant paths that cannot effectively cope with the dead ends. In order to handle them, this paper proposes an online robot path planning mechanism to help the robot successfully escape the dead end by recording low-cost key locations, and reduce the energy consumption of the robot by determining the specific location of the covered cavity. The mechanism consists of key location point recognition, trajectory cluster formation, dead-end escape and route selection. It can also optimize traditional robot mobile deployment algorithm by combining with them. Compared with five well-known robot path planning algorithms, simulation results indicate its efficiency in terms of coverage ratio, path length, energy efficiency and path search efficiency.