Development of deep reinforcement learning for inverted pendulum

This paper presents a modification of the deep Q-network (DQN) in reinforcement learning to control angle inverted pendulum (IP). The original DQN method often uses two actions related force states like constant negative and positive values which apply cart IP maintain between Y-axis. Due changing too much value force, may make some oscillation makes performance system could be declined. Thus, modified algorithm is developed based on neural network structure range selections for improve IP. To prove our algorithm, OpenAI/Gym Keras libraries are used develop DQN. All results showed that proposed controller has higher than applied nonlinear system.