Switching Deep Reinforcement Learning based Intelligent Online Decision Making for Autonomous Systems under Uncertain Environment

Abstract

In this paper, finite horizon intelligent decision making problem has been investigated for autonomous systems especially under uncertain environment. According to latest studies, the uncertainty of environment will seriously affect the effectiveness of decision making especially for autonomous systems. To handle this issues, transfer learning and deep reinforcement learning has been presented recently. However, those existing Learning algorithms commonly needs a large set of state space which cause the algorithm to be time consuming and not suitable for real-time application. Therefore, in this paper, a library of polices trained using Deep Q-Learning under similar environments are built firstly. Then, a neural network is designed to estimate the environment. Using the learned environment, a novel of switching policy will be developed and integrated with the designed deep reinforcement learning which can efficiently stop learning according to the practical error tolerance. Meanwhile, through the novel policy evaluation method based on the environment estimator, the autonomous agent will select the best policy to followin an online manner. Eventually, simulation results are provided to demonstrate the effectiveness of the designed algorithm.