search

reiniscimurs / DRL-robot-navigation

Deep Reinforcement Learning for mobile robot navigation in ROS Gazebo simulator. Using Twin Delayed Deep Deterministic Policy Gradient (TD3) neural network, a robot learns to navigate to a random goal point in a simulated environment while avoiding obstacles.
MIT License
571 stars 119 forks source link

Global Navigation #85

Closed stars113 closed 5 months ago

stars113 commented 10 months ago

Hi Reinis Cimurs,

I hope this message finds you well. I've been exploring your Goal-Driven Autonomous Exploration Through Deep Reinforcement Learning on GitHub and I am quite impressed with the work you've done.

I noticed that there seems to be a lack of code related to global navigation in the repository. Could you kindly provide more information about the absence of global navigation code?

I am particularly interested in understanding how global navigation could be integrated into the project and if there are any specific considerations or challenges you have encountered in this regard.

Thank you for your time, and I appreciate your efforts in making this project open-source.

                                                                                                                      Best regards
reiniscimurs commented 10 months ago

Hi,

Please see the actual repository for GDAE: https://github.com/reiniscimurs/GDAE This is a generic repository for DRL navigation policies. You can use this as a starting point for global planner integration, but if it is to be integrated into GDAE, be aware of discussion in: https://github.com/reiniscimurs/GDAE/issues/15

stars113 commented 10 months ago

Thank you for your reply. You mentioned earlier that "DRL-navigation is a separate repository from GDAE and is not the precise training method used to train the network for this implementation, but a general DRL navigation policy training method. GDAE is a concrete implementation of the physical layer specific repository For Android. This means that there are some items in the code that are not needed in the general solution. In order to use the DRL navigation strategy here, the GDAE code must be updated to be consistent with it." Meaning yes, if I want to use the pytorch framework to Integrate GDAE and DRL. To do the following:

  1. Replace GDAE framework with pytorch
  2. Update DRL policy to GDAE Do you have a better approach to recommend?
reiniscimurs commented 10 months ago
  1. You would essentially just need to update the GDAM.py file there to use pytorch. You could either update that file to use pytorch functionality or unpdate the DRL-navigation test file to pass the arguments and use the env file from GDAE.
  2. Other way around. You would have to update GDAE to use the trained model policy from this repo.

That should replicate GDAE with using policy of the current repo.