DESIGN AND REALIZATION OF AUTONOMOUS CARS USING DEEP Q LEARNING

摘要

Self driving cars are one of the most acclaimed technologies of the 21st century after the internet, but they have become a bone of contention among orthodox drivers. With the evolution of advances in software such as reinforcement learning algorithms and Q-learning, the world of artificial intelligence has taken a big leap forward. These algorithms are nature inspired to categorize actions through a system of reward points and negative points. Our research reported here focuses on implementation techniques of such reinforcement algorithms in scenarios such as the self-driving car. In this work we refer to the Bellman equation to give rewards for certain actions and the Markov decision processes for decision-making which includes a certain degree of randomness in the self-driving car and make compromises to reach its destination.