Path Planning for Deep Reinforcement Learning with Adaptive Greedy Factor
Although DQN(Deep Q-Network),a pioneering algorithm of deep reinforcement learning,has excellent performance in path planning,it still has some problems such as valuation,experience replay mechanism de-fects,and lack of a good balance between exploration and utilization.A deep reinforcement learning path planning al-gorithm with an adaptive greed factor is proposed.Firstly,the preferential experience replay mechanism is introduced on the basis of the D3QN algorithm,which solves the estimation problem and increases the sampling probability of im-portant samples,thus improving the efficiency of the algorithm.Secondly,a new reward function is designed to improve the differentiation of actions.Finally,an adaptive greed factor is designed to balance the relationship between explora-tion and utilization.The TensorFlow framework in Python and Tkinter library are used to establish the environment map to verify the effectiveness of the algorithm.The results show that the improved algorithm is superior to the DQN algorithm in both the optimal path and the number of algorithm iterations.
Path planningDeep reinforcement learningExploration factorExperience in playback
NETL
NSTL
万方数据
