Skip to content

marmotlab/randenv_nodegen

Repository files navigation

randenv_nodegen

A2C RL + pruning after-processing to plot key nodes in a randomly generated environment.

how to train new models

  1. change EXPERIMENT_NAME and EXPERIMENT_NOTE under alg_parameters.py -> RecordingParameters
  2. create new directory inside complete_map called EXPERIMENT_NAME (from step 1) 3a. If importing pre-existing actor and critic, uncomment lines 82 & 83, specify the actor and critic file path in the input parameter 3b. If training model from scratch, comment out lines 82&83
  3. run driver.py
  4. track results using wandb. Printouts of completed renders will be available at complete_map/EXPERIMENT_NAME. (pruning post-processing is not in this step)

how to verify trained model with post processing

  1. In model_verifier.py, specify trained actor file path in the input parameter for actor_file_name (line 19)
  2. run model_verifier,py. Observe pygame window to see final node position after pruning
  3. press enter to regenerate random map and see new node positions

alt text

NOTE:

  1. reward strcture is within finder_gym.py
  2. ray casting function is within map_tester.py

About

A2C RL + pruning after-processing to plot key nodes in a randomly generated environment.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published