RTP-Q: a reinforcement learning system with time constraints exploration planning for accelerating the learning rate

摘要

Reinforcement learning is an efficient method for solving MarkovDecision Processes that an agent improves its performance by usingscalar reward value with higher capa- bility of reactive and adaptivebehaviors. Q-learning is repre- sentative reinforcement learningmethod which is guaranteed to obtain an optimal policy needs numeroustrials to achieve it. k-Certainty Exploration Learning Systemrealizes active ex- rated into two phases and estimate values are notderived during the process of identifying the environment.