Brain-computer Interface (BCI) applications based on steady-state visual evoked potentials (SSVEP) have the advantages of being fast, accurate and mobile. SSVEP is the EEG response evoked by visual stimuli that are presented at a specific frequency, which results in an increase in the EEG at that same frequency. In this paper, we proposed a novel human-guided maze solving robot navigation system based on SSVEP. By integrating human's intelligence which sees the entirety of the maze, maze solving time could be significantly reduced. Our methods involve training an offline SSVEP classification model, implementing the robot self-navigation algorithm, and finally deploy the model online for real-time robot operation. Our results demonstrated such system to be feasible, and it has the potential to impact the life of many elderly people by helping them carrying out simple daily tasks at home with just the look of their eyes.