Deep-Reinforcement-Learning-for-mapless-navigation-in-intralogistics

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

Illustration of our results

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.

Illustration of our results

[1] https://github.com/quantumiracle/Popular-RL-Algorithms