Navigation Application of Q-learning Neural Network
Yinghao Chen, Hao Wang, Chian‐Song Chiu · 2020
Traditional navigation application has a problem with adaption for changing environment and time complexity. Reinforcement learning gives a better solution for time complexity. Taking Q-learning for example, thought Q-value(score) for every state, Q-learning can choose action instantly, but it should take lots of memory to save Q-table (table for saving Q-value). Neural network is always a good idea to save memory and adapt in every kind of environment. By integrating the above concept, we finally use Deep Q-learning neural network (DQN) to do path planning and give a method to score the action and environment for make the choice of action.