search

OscarHuangWind / DRL-Transformer-SimtoReal-Navigation

[T-ITS] Sim-to-real goal-oriented mapless autonomous navigation (DRL navigation).
MIT License
75 stars 9 forks source link

How to Store Expert Data Before Training in SAC Algorithm #14

Open zahra370 opened 1 month ago

zahra370 commented 1 month ago

I’m having trouble figuring out how to first store expert data into the replay buffer and then start the training process in the SAC algorithm. would you please guide me on the correct sequence for storing expert data and initiating the training phase?

OscarHuangWind commented 1 month ago

Below I list the steps that I would recommend:

  1. Collect the expert data: python3 demonstration.py (https://github.com/OscarHuangWind/DRL-Transformer-SimtoReal-Navigation/blob/master/catkin_ws/src/gtrl/scripts/DIL/demonstration.py)

  2. Load the pre-buffer from expert data: set the "PRE_BUFFER" parameter as True in config.yaml and run python3 main.py

  3. Enable the human-in-the-loop guidance: set the "HUMAN_INTERVENTION " parameter as True in config.yaml and run python3 main.py