Enhanced image preprocessing method for an autonomous vehicle agent system

Excessive training time is a major issue face when autonomous vehicle agents with neural networks by using images as input. This paper proposes deep time-economical Q network (DQN) input image preprocessing method to train an agent in virtual environment. The environmental information extracted from the A top-view of entire environment then redrawn according information. During DQN model, cropped place at center image. current frame combined previous two iterations. model use this experimental results indicate higher performance and shorter for trained preprocessed compared that without preprocessing.