Behavioral decision-making for urban autonomous driving in the presence of pedestrians using Deep Recurrent Q-Network

Decision making for autonomous driving in urban environments is challenging to the complexity of the road structure and the uncertainty in the behavior diverse road users. Traditional methods consist of manually designed rules the driving policy, which require expert domain knowledge, are difficult to and might give sub-optimal results as the environment gets complex. , using reinforcement learning, optimal driving policy could be learned improved automatically through several interactions with the environment. , current research in the field of reinforcement learning for autonomous is mainly focused on highway setup with little to no emphasis on urban. In this work, a deep reinforcement learning based decision-making for high-level driving behavior is proposed for urban environments in presence of pedestrians. For this, the use of Deep Recurrent Q-Network (DRQN) is explored, a method combining state-of-the art Deep Q-Network (DQN) a long term short term memory (LSTM) layer helping the agent gain a memory the environment. A 3-D state representation is designed as the input with a well defined reward function to train the agent for learning an behavior policy in a real-world like urban simulator. The proposed is evaluated for dense urban scenarios and compared with a rule-based and results show that the proposed DRQN based driving behavior maker outperforms the rule-based approach.