This repository contains the ROS workspace and tools needed to run manipulation tasks for the RoBorregos @Home robot. It includes the robot description with URDF and mesh files for the most recent robot including a EAI Dashgo B1 mobile base and a UFactory xArm6. It also contains the MoveIt configuration and packages required for picking, placing and manipulating objects, including 2D and 3D object detector, pointcloud processing and the manipulation pipelines for diverse tasks.
To run the repository, ROS1 Noetic is required. To install ROS Noetic, follow the instructions in the official ROS website.
To install ros dependencies from the workspace, run the following commands:
cd ws/
rosdep update
rosdep install --from-paths src --ignore-src -r -y
Additional dependencies are required for the xArm submodules and the Grasping Pose Detector. For xArm, follow the instructions in the xArm ROS repository, and for the Grasping Pose Detector, follow the instructions in the Grasping Pose Detector repository. For GPD, install eigen and pcl with the following commands:
sudo apt update && apt install -y libeigen3-dev libpcl-dev
The repository contains a Makefile and scripts for running the workspace in a Docker container. It also allows for use with CUDA (which helps in running the object detection YOLO models).
- Docker Engine
- Post-installation steps for Linux If using GPU:
- NVIDIA Driver
- NVIDIA Container Toolkit)
To create the Docker image, run the following command:
make manipulation.build # for CPU
make manipulation.build.cuda # for GPU
To create the Docker container, run the following command:
make manipulation.create # for CPU
make manipulation.create.cuda # for GPU
To run the Docker container, run the following command:
make manipulation.up
make manipulation.shell # for GPU
To remove the Docker container, run the following command:
make manipulation.down
make manipulation.remove
The robot description is located in the ws/src/robot_description
package. It contains the URDF and mesh files for the robot, including the mobile base and arm. After sourcing the workspace, visualize the robot in RViz with the following command:
roslaunch robot_description display.dashgo.launch
To turn on the xArm6 and allow the robot to make movements with MoveIt planning and other planning and movement options, run the following command:
roslaunch dashgo_moveit_config dashgo_xarm.launch robot_ip:=<robot_ip>
The current, most recent and stable manipulation pipelines are located in the ws/src/pick_and_place
package. The pipelines include the object detection, pointcloud processing and manipulation tasks. To run the manipulation pipeline, run the following command:
roslaunch pick_and_place cartesian.manipulation.launch
This command also requires the ZED2 camera, or any other camera that publishes a pointcloud in the /zed2/zed_node/point_cloud/cloud_registered
topic. The camera can be opened with the following command from the zed-ros-wrapper repository in the machine where the camera is connected:
roslaunch zed_wrapper zed2_robot.launch