RL project

Hello everyone ,i want to make a project on self Car using Deep Q network to get best way for destination ,i have 2 problems ? where can i train the model by actually moving the robot to generate data for Q network?how to train model to always do best not just for place i have trained ?

Have you heard about CordConv by uber. Check it out

1 Like

how did it will help train in that situation (reducing times of actually moving the robot to generate ,not just do best for place i have trained),I did some GPT research on it and got that it’s used when i need to inject coordination,