Reinforcement learning for construction site robot

Robot that navigate through construction site using Q-learning written in python 3 powered by Pytorc and Robot operating system

How this work

                        +------------------------------------+
                        |           Infrastructure           |
                        +------------------------------------+
                        |               Host OS              |
                        +------------------------------------+
                        |                Docker              |
                        +-----------+------------+-----------+
                        |           |            |           |
                        |    ROS    |   Pytorch  |  RTABMAP  |
                        | Container |  Container | Container |
                        |           |            |           |
                        +-----------+------------+-----------+

Docker image contain Robot Operating System (ROS) and Machine Learning framework such as TensorFlow, Pytorch, Keras, etc. You can switch whatever framework you like.

Hardware

Raspberry Pi 3 Model B with Raspbian image SD card
RGB-D or Stereo image Camera (Microsoft Kinect is highly recommended)
Smart Robot Car kit
Arduino UNO R3

How to assemble robot

Please check wiki page

Prerequisite

You can choose to install it manually or use a script

Docker CE for Raspberry Pi
MeshLab
Python 3.6

Usage

copy & paste in terminal

wget https://raw.githubusercontent.com/kanokkorn/RL-bot/master/run.sh && sudo chmod +x run.sh && ./run.sh

or clone this repository

This script is design to work with Raspberry Pi only.

sudo run.sh

If the setup is done or you have your own setup then run this command

sudo docker-compose up

License

This project is under BSD license