Skip to content

Latest commit

 

History

History
84 lines (57 loc) · 3.9 KB

README.md

File metadata and controls

84 lines (57 loc) · 3.9 KB
Raw

RoboRTS

Build Status Gitter

RoboRTS is an open source software stack for Real-Time Strategy research on mobile robots, developed by RoboMaster. The Real-Time Strategy research that we focus on is multi-agent deep reinforcement learning based decision making.

The motivation for this project is RoboMaster AI Challenge. In this robot challenge, multiple robots should fight with each other on a game field automatically. It would be very convenient to have a unified framework for them to integrate hardware components and implement algorithms.

A modern mobile robot usually contains a simple low-level computing platform runs RT OS (Real Time Operation System) with restricted computing power but real-time output, and a strong high-level computing platform runs non-RT OS (such as Linux and Windows) which dedicates to computational intensive algorithm execution. In our system, the low-level computing platform is MCU (micro-controller), and the high-level computing platform is called the on-board computer.

The framework of RoboRTS consists two parts: autonomous mobile robot layer and intelligent decision-making layer.

The autonomous mobile robot layer alone can let a robot to demonstrate a certain level of intelligence. On its on-board computer runs perception, planning, decision modules, each of which contains some conventional algorithms like A* search and Lidar mapping. On its MCU, an RT low-level robot controller is implemented to govern the robot driving system. This README mainly explains the structure and functionality of this layer.

TODO: Intelligent decision-making layer includes a multi-agent decision-making framework and a game simulator, it will be released in the future.

Get started

You can build a mobile robot described in hardware_setup document and setup embedded system mentioned in RoboRTS-Firmware repository.

Software Requirements

  • ROS kinetic (Ubuntu 16.04)
  • OpenCV 3.0.0 or higher
  • CUDA(Optional)
  • PCL
  • G2O
  • SuiteSparse
  • Google Protocol Buffers 2.6
  • Google Glog
  • CMake 3.1

Install dependencies for Ubuntu 16.04 kinetic

Install ROS

Please follow the installing and configuring ROS environment tutorial on ROS Wiki.

Thirdparty Library

sudo apt-get install ros-kinetic-opencv3 ros-kinetic-cv-bridge ros-kinetic-image-transport ros-kinetic-stage-ros ros-kinetic-map-server ros-kinetic-laser-geometry ros-kinetic-interactive-markers ros-kinetic-tf ros-kinetic-pcl-* ros-kinetic-libg2o protobuf-compiler libprotobuf-dev libsuitesparse-dev libgoogle-glog-dev ros-kinetic-rviz

Build and Install

git clone https://github.com/RoboMaster/RoboRTS.git
cd RoboRTS/
mkdir build
cd build
cmake ..
make messages_generate_messages
make

Setting the RoboRTS environment variable

cd path/to/RoboRTS
echo "export ROBORTS_PATH=${PWD}" >> ~/.bashrc
source ~/.bashrc

Run Examples

./tools/script/run.sh -s

Documents

  • Hardware setup: Instruction to build a RoboMaster infantry robot.
  • Software architecture: The software architecture of RoboRTS framework.
  • Protocol: The protocol between the on-board computer and the MCU.
  • Tutorial: Additional tutorials will be released in the future.

Copyright and License

RoboRTS is provided under the GPL-v3.