Duckietown Deep Q learning

Hi , I am working on the lecture 32 duckietown deep q learning problem. The implementation from the lecture isn’t sufficient for making the robot learn move on turns. It’s getting stuck in a local optima and no turning. Does anyone have any solution for the same?


The Deep Q here is really basic and its objective is teach the basic tools. To make it work it needs your work to optimise and make it at the end work.