Self-driving car navigation with obstacle avoidance problem is a hot topic in academic world.The goal of this problem is to design an autonomous car with learning ability to find a collision-free path in an unknown environment.General solutions to this problem could be divided into two pieces,navigation environment detection and control strategy design.In this paper,we used reinforcement learning(RL)to approach control strategy design and built two different navigation benchmarks with different obstacles situations to verify the control strategy.The simulation results showed that the RL provides an effective solution to self-driving car navigation with obstacles avoidance problem.