In recent mobile robotics fields, much attention has been focused on hybrid control methods consisting of local reactive control and deliberative control for global navigation in large unknown environments. In this paper, an improved hierarchical Q-learning algorithm with quantum parallelization computation is proposed for mobile-robot global navigation. The learning process consists of two learning levels for local navigation and topological navigation, respectively. The hierarchial Q-learning bridges these two learning levels with quantum parallelization to combine the learning process into an integral one and speed up the whole learning process as well. Hence the improved hierarchial Q-learning method learns faster and can scale up very well. The results of several group of experiments show the success of the presented approach.