simple_pybullet_ros2

This repository contains 2 tasks for Pybullet

Stacking Task

This task involves one Franka Panda robotic arm randomly taking 5cm cubes from a table with 21 cubes arranged in a 7 * 3 grid and stacking them. It will stack them into 3 stacks of 5 cubes. The number of columns and the the heigt can be configured in moveit_controller.launch.py

Throw Task

For this, a Franka Panda arm will pick up a cube and throw it toward a pyramid made of 6 cubes. The arm will go as far back as it can and perform a throwing motion toward the ground in front of it. The ball will be released at 50% of the trajectory as it is the moment with the most amount of forces. The pyramid is placed such that a successful throw at full power will collapse it.

Instructions

Requirements

• OS: Ubuntu 20.04 (Focal)
  • Other distributions might work (not tested).

Dependencies

These are the primary dependencies required to use this project.

• ROS 2 Foxy
• PyBullet
• MoveIt 2
  • Install/build a version based on the selected ROS 2 release

Building

Clone this repository and import submodules. Then install dependencies and build with colcon.

# Install pyquaternion
pip install pyquaternion
# Clone this repository and the submodules into your favourite ROS 2 workspace
git clone https://github.com/09ubberboy90/simple_pybullet_ros2.git
# Install external dependencies via rosdep
rosdep install -r --from-paths src --ignore-src --rosdistro ${ROS_DISTRO}
# Build with colcon
colcon build --merge-install --symlink-install --cmake-args "-DCMAKE_BUILD_TYPE=Release"

Sourcing

Before utilising this package, remember to source the ROS 2 workspace overlay.

source ${ROS2_WS}/install/local_setup.bash

Running

The following tasks can be either run as an independant launch or can be recorded

Stacking cube

ros2 launch pybullet_panda stack_cubes.launch.py

Throwing cube

ros2 launch pybullet_panda throw_cubes.launch.py

Recorder

The recorder can be launched with.

ros2 run sim_recorder run_recording 

It has some argument option to configure what the task will be

--help: Gives a list of arguments
-t, --throw: Start the Throwing Task
-i, --iteration: Number of run that should be performed
-s, --start: Start index in case some run need to be rerun after an interuption
--headless: runs without a GUI

The recorder will save the CPU and RAM usage in sim_recoder/data/{sim_name}. sim_name is created automatically based ont the simulation name ignition and on the arguments:

• _gui is added if --headless is not passed
• _throw is added if --throw is passed

An analysis of the run can be performed by running from the source of the repo

python ./sim_recorder/sim_recoder/grapher.py {name_of_folder}

With name of folder being the name of the folder you want to analyse