A robotics simulator is a simulator used to create application for a physical robot without depending on the actual machine, thus saving cost and time. In some case, these applications can be transferred onto the physical robot (or rebuilt) without modifications. — Wikipedia
During application development a robotics simulator can greatly help by cutting down the cost and time associated with managing real robots. This effect is more pronounced during initial phases of development where more code and configuration iterations happen with higher rate of trial-and-error phases.
Gazebo is popular open-source 3D robotics simulator that has been around for more than a decade, and is used by robotics developers around the world. It has an active community and plethora of documentation that helps new users get on-board easily. Gazebo team maintains quite a few 3D models that are readily available for use in projects. The team also maintains ROS packages for applications that want to interact with Gazebo through ROS interfaces (topics, services, and actions).
rapyuta.io lets users run their simulation application with Gazebo simulator on the cloud. It provides a web interface to watch and interact with Gazebo GUI client running on the cloud. ROS interfaces provided by its ROS packages complements rapyuta.io’s ROS communication support, and enhances the ways users architect their simulation and robotics applications. For instance, robotics application running on a device can interact with Gazebo running on cloud through ROS interfaces (e.g. to get model state, set physics properties). This requires, however, that launch files for simulation are not tightly coupled with the rest of the application so they can be run independently.
The following sample-walkthrough demonstrates how to create such a setup keeping in mind that it should still run on local machine besides rapyuta.io: Separating Simulation and Application.
As of now rapyuta.io supports Gazebo version 9, and only for executables with Catkin recipe. Users are expected to use gazebo9 ROS packages for their corresponding ROS distros. For instance,
ros-melodic-gazebo-ros-pkgs on ROS melodic, and
ros-kinetic-gazebo9-ros-pkgs on ROS kinetic.
One limitation while running the above setup is that the ROS parameter server is not shared. This means, for instance, robot_description loaded by robotics application on device (for use by robot_state_publisher) will not be available to Gazebo simulation running on cloud (to spawn URDF model). Since such parameters that need to be shared are not modified after being set, both simulation launch files and application launch files can set them for their respective ROS masters. Controller parameters are another example besides robot_description.
When communicating over the public internet through ROS interfaces, communication can incur significant delays. The delays may or may not affect simulation and robotics applications, depending on how much tolerant application is towards these delays. The following ways can be used to account for such delays:
Run Gazebo at a slower speed by reducing real-time factors (which in turn reduces /clock frequency). This can be done by modifying max_step_size and real_time_update_rate to decrease the real-time factor. In gzclient, these are updated by selecting Physics in the World tab. Or, in SDF world file, they can be updated in <physics> tag. The following changes in SDF world file will reduce the real-time factor from 1 (the default) to 0.5, thus running Gazebo simulation at half the normal speed:
<physics type='ode'> <real_time_update_rate>100</real_time_update_rate> <!-- default: 1000 --> <max_step_size>0.005</max_step_size> <!-- default: 0.001 --> <real_time_factor>0.5</real_time_factor> <!-- default: 1 --> </physics>
<param name="/gazebo/pub_clock_frequency" value="500" />
If none of this helps and sharing ROS master is the only way, then simulation and the application must be run in the same Package Component.
rapyuta.io only supports Gazebo version 9. When using Gazebo ROS packages, it assumes those to be gazebo9. When working on Ubuntu xenial, specifying Gazebo ROS dependencies in package.xml will fetch gazebo7 ROS packages. For instance,
<depend>gazebo_ros</depend> in package.xml will fetch
ros-kinetic-gazebo-ros, which is a package for gazebo7. To install gazebo9 packages, the user will need to add the rosdep source-list from https://github.com/osrf/osrf-rosdep/blob/master/gazebo9/00-gazebo9.list. On the local machine execute in terminal:
wget -qP /etc/ros/rosdep/sources.list.d/ https://raw.githubusercontent.com/osrf/osrf-rosdep/master/gazebo9/00-gazebo9.list rosdep update
After this, installing
<depend>gazebo_ros</depend> should fetch you
ros-kinetic-gazebo9-ros. While building your simulation application, this is what rapyuta.io does too.
Follow the sample walkthrough: Basic Simulation