A Reinforcement Learning-Based Approach for Autonomous Vehicle Decision-Making

Nowadays, cars are always the things people cannot live without. With the development and improvement of the car industry and general income, the number of cars equipped with self-driving systems is increasing dramatically in the market. Self-driving has already shown a tendency for a car to have this function as a standard instead of an attractive feature. In contrast, the reliability of this technology is now being doubted. Many drivers point out that sometimes their cars will make silly decisions while in self-driving mode, such as ignoring other cars on the road. This paper aims to come up with a possible approach to help in reducing the errors of the self-driving system in decision-making. The paper mainly uses a quantitative study through the literature review method in data collection and analysis. By summarizing data and reviewing existing materials, a new learning-based approach will be demonstrated in this paper. It discusses a feasible way to let AI in self-driving vehicles memorize and judge road conditions subjectively.