How to build myself environment?

[zhang-qiang-github]

I want to implement reinforcement learning with tianshou, but I need to build myself environment.

My enviroment is a 3D maze (128x128x128 matrix), and the network should be a 3D convolution neural network.

How can I build myself enviroment? Should I subclass BaseVectorEnv? Or is there any example to build myself environment?

Thanks for any suggestion~~~

[Trinkle23897]

https://tianshou.readthedocs.io/en/master/tutorials/cheatsheet.html#user-defined-environment-and-different-state-representation

[konnovdev]

note that the link to this doc has changed now, here's the updated link: https://tianshou.org/en/stable/01_tutorials/07_cheatsheet.html#user-defined-environment-and-different-state-representation