Mobile robots contributed significantly to the intelligent development of human society, and the motion-planning policy is critical for mobile robots. This paper reviews the methods based on motionplanning policy, especially the ones involving Deep Reinforcement Learning (DRL) in the unstructured environment. The conventional methods of DRL are categorized to value-based, policy-based and actorcritic-based algorithms, and the corresponding theories and applications are surveyed. Furthermore, the recently-emerged methods of DRL are also surveyed, especially the ones involving the imitation learning, meta-learning and multi-robot systems. According to the surveys, the potential research directions of motion-planning algorithms serving for mobile robots are enlightened.