Human aware robot navigation, which takes human factors into account, has been researched more and more for mobile robotics. Navigating in a human existing environment, a robot had better neither neglect potential human cooperation nor always exploit human cooperation. In this paper, we propose the potential co-cost minimization principle (PCCMP) and the generic PCCMP architecture in the context of human aware robot navigation. We describe a concrete instantiation of the PCCMP and demonstrate with simulation results how the PCCMP architecture enables a robot to flexibly and reasonably decide whether to exploit human cooperation according to concrete scenario conditions.