-1

How to create an openai gym custom environment?

zoraiz ali
  • 77
  • 5

2 Answers2

0

For your question How to create an deepbots custom environment for my robot?

You can create your own custom enviromnent in gym and you have exemple on env you can be inspired by. For instance this one https://gym.openai.com/envs/CarRacing-v0/ (the source is available) maybe modifying this env would be enough for you

For your question How to implement DQN with CNN using TensorFlow and keras.

There is just what you need here I think https://keras.io/examples/rl/deep_q_network_breakout/ You can see more algorithm here https://github.com/openai/baselines

Sorry for not providing code but your questions is very open and giving you links sounds like a better ideas.

EDIT 10/11/2020

I think what you are looking for is https://github.com/MattChanTK/gym-maze I strongly advised to use the env like that to write your DQN. Then when it works for you, check the source code and update it for your needs.

Alexandre Catalano
  • 764
  • 4
  • 10
0

@Alexandre Catalano thanks for sharing,enter image description herebut how to create custom environment with obstacles for robot like image above where red circle is robot and green square is target and black z is static obstacle for deep reinforcement learning DQN algorithm, which can be import like others gym environment in python.

I want to create environment, where robot can find target with static obstacles.

zoraiz ali
  • 77
  • 5
  • I edited my answer to avoid having too much answers and questions. You should comment or edit your answers to keep it easy for other people to understand it too. Stack overflow is a community ;) If my answer works for you, please mark it as the correct answer, if not please leave me a comment under it – Alexandre Catalano Nov 10 '20 at 07:42