This repository is for the paper "Using Deep Reinforcement Learning with Automatic Curriculum Learning for Mapless Navigation in Intralogistics".
The codes includes:
- Soft Actor Critic Algorithm, adapted from [1]
- Our algorithm of NavACL_Q
- Hyper-parameter Settings
Here we demonstrate some videos of our trained agent on successfully reaching goal state, i.e., the blue dolly.
Media2_Trim_1.mp4
Media2_Trim_2.mp4
Media2_Trim_3.mp4
Media2_Trim_4.mp4
Even with a moving target, the trained agent is able to react responsively to navigate to the goal.
