Abstract: One of the core issues facing the current ethical dilemma of autonomous vehicles is：whether ethics should be embedded in the structure of the algorithm and how it should be embedded. In the face of possible traffic accidents in the future, there are serious difficulties in screening information and relying on "moral luck" to make random choices and independent information systems based on artificial intelligence. Therefore, a "moral algorithm" should be preset for autonomous vehicles. And for how to decide the problem of "moral algorithm", in view of the conflict between existing moral principles, the complexity of moral decision-making and the contextualization of human moral judgment, it is unrealistic to base on some kind of ethical principle set by mankind, so we can try to simulate the realization of human moral decision in the same situation.