search

vita-epfl / CrowdNav

[ICRA19] Crowd-aware Robot Navigation with Attention-based Deep Reinforcement Learning
MIT License
601 stars 172 forks source link

Training Issue #46

Open Watson52 opened 3 years ago

Watson52 commented 3 years ago

Sorry to bother you, I am new to learn reinforcement learning and i' d like to ask you some question.

  1. I wonder how could i change the times of iterations of training?
  2. Could i visualize the process during the training?
  3. Are there exist pre-trained model of IL or RL?

Looking forward to your reply. Thank you very much.

ChanganVR commented 2 years ago

Hi @Watson52 sorry for the late reply. For your questions,

  1. Do you mean the increment for each time step? You should be able to control it with this parameter in the configuration: https://github.com/vita-epfl/CrowdNav/blob/20d678085c06831e658a65b9e20c8bb6f6ecdc10/crowd_nav/configs/env.config#L3

  2. By visualizing the process, do you mean something like tensorboard? It currently only prints out the log. But logging to tensorboard should be very easy to add.

  3. Unfortunately, I don't have any pretrained model weights. But you can re-produce the results very easily following the instructions in the readme file.

Let me know if you have any other questions. Changan