Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/nrjsbudhe/robotic-warehouse-automation-using-reinforcement-learning
Implemented Reinforcement Learning Algorithms to Automate Path Finding process in Robotic Warehouse Environments | Q-Learning | TD Learning | Monte Carlo Methods | Dynamic Programming
https://github.com/nrjsbudhe/robotic-warehouse-automation-using-reinforcement-learning
automation reinforcement-learning-algorithms robotics
Last synced: about 1 month ago
JSON representation
Implemented Reinforcement Learning Algorithms to Automate Path Finding process in Robotic Warehouse Environments | Q-Learning | TD Learning | Monte Carlo Methods | Dynamic Programming
- Host: GitHub
- URL: https://github.com/nrjsbudhe/robotic-warehouse-automation-using-reinforcement-learning
- Owner: nrjsbudhe
- Created: 2024-08-21T13:52:59.000Z (4 months ago)
- Default Branch: main
- Last Pushed: 2024-08-21T17:33:30.000Z (4 months ago)
- Last Synced: 2024-08-22T15:37:06.300Z (4 months ago)
- Topics: automation, reinforcement-learning-algorithms, robotics
- Language: Python
- Homepage:
- Size: 2.06 MB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Robotic-Warehouse-Automation-using-Reinforcement-Learning
Implemented Reinforcement Learning Algorithms to Automate Path Finding process in Robotic Warehouse Environments | Q-Learning | TD Learning | Monte Carlo Methods | Dynamic Programming
### Migrated from original repository - [Original Project](https://github.com/jitesh3023/Reinforcement-Learning-Final-Project)This repository contains the implementation of various Reinforcement Learning (RL) algorithms aimed at optimizing robot pathfinding in a custom warehouse environment. The algorithms include Monte Carlo Methods, Q-learning, SARSA and Deep Q-learning.
## Project Overview
In this project, a simulated custom warehouse environment is generated using Gymnasium. Robots navigate this environment using the implemented RL algorithms, with the objective of accessing certain checkpoints in the environment before reaching the goal. The path finding process is automated using Reiforcement Learning algorithms. The performance of each algorithm is compared to determine the most effective approach for warehouse automation.
## Key Features
- **Custom Warehouse Environment**: A simulated custom environment built using Gymnasium. The aim was to approach the problem with the simplest environment possible. We have used _matplotlib_ for rendering the environment.
![alt text](https://github.com/nrjsbudhe/Robotic-Warehouse-Automation-using-Reinforcement-Learning/blob/main/plots/Environment.png)
- **Reinforcement Learning Algorithms**: Implementation of Dynamic Programming, Monte Carlo Methods, Q-learning, and Deep Q-learning.
- **Performance Comparison**: Evaluation and comparison of the effectiveness of each algorithm in optimizing robot paths.
### Q - Learning : 100 episodes
![alt text](https://github.com/nrjsbudhe/Robotic-Warehouse-Automation-using-Reinforcement-Learning/blob/main/demo/Q_100.gif)### Q - Learning : 10000 episodes
![alt text](https://github.com/nrjsbudhe/Robotic-Warehouse-Automation-using-Reinforcement-Learning/blob/main/demo/Q_1000.gif)### Multi Goal Targets
![alt text](https://github.com/nrjsbudhe/Robotic-Warehouse-Automation-using-Reinforcement-Learning/blob/main/demo/MULTI_GOAL.gif)## Analysis
The optimal policy generated by the Reinforcement Learning algorithms was used for determing an optimal path for reaching the goal. The path was compared with traditional path finding algorithms like A\*. The paths provided by RL algorithms were optimal. The approach can be used in large warehouses where there are constraints on mapping and creating pre-planned paths for robot navigation.