Skip to content

Latest commit

 

History

History
85 lines (67 loc) · 3.97 KB

README.md

File metadata and controls

85 lines (67 loc) · 3.97 KB
Raw

Orocos-RTT / Gazebo Integration

This repository contains tools for prototyping robot control systems designed with the Orocos Toolchain Real-Time Toolkit (RTT) in the Gazebo Simulator. The goal is to make it easy to run the exact same collections of RTT components in simulation as are run on real hardware.

Building

NOTE: The Orocos Toolchain packages (RTT amd OCL) need to be built with the following CMake arguments to enable use of the remote console (rtt_gazebo_console). These are the default arguments used in the ROS-distributed Orocos binary packages:

-DENABLE_CORBA=ON -DCORBA_IMPLEMENTATION=OMNIORB

Packages

  • rtt_gazebo_system A Gazebo SystemPlugin for initializing RTT in-process and overriding the RTT::os::TimeService with the Gazebo clock, as well as triggering any rtt_rosclock::SimClockActivity-bound RTT components.
  • rtt_gazebo_deployer A Gazebo ModelPlugin for running an RTT deployment component associated with a given Robot Model. This deployment component can then be used to load other RTT components. This will also instantiate an RTT TaskBrowser associated with each Gazebo Model.
  • rtt_gazebo_console A CORBA-based console for connecting to an Orocos TaskBrowswer running in the Gazebo Simulator (via rtt_gazebo_deployer).
  • rtt_gazebo_examples Example ROS launchfiles, Gazebo worlds, and RTT components which can interact with Gazebo.

See the README.md files in each package for more details.

Theory of Operation

In order to simulate a piece of hardware, you will write a custom end-point component (like a driver) which will expose the same RTT interfaces as your actual hardware. Instead of calling a lower-level API for reading the device, this component will define a pair of RTT operations which will be called each time Gazebo's simulation engine updates.

In this callback, the compoents can query any information about the world, and store it as necessary. Note that this execution path is different than the way in which the component interacts with Orocos/RTT.

This component will be associated with a Gazebo SDF model, and when the SDF is loaded, the rtt_gazebo_deployer Gazebo model plugin will create an RTT DeploymentComponent in which the simulation component is instantiated. This will allow you to load numerous other components as its peers.

These deployers can be accessed via the rtt_gazebo_consol over the standard RTT CORBA deployer interfaces. This enables quick and easy introspection into components which are technically instantiated in the Gazebo process.

In order for your periodic tasks to run at the right rates, you can use the SimClockActivity in the rtt_rosclock package. When loading a component in the rtt_gazebo_deployer, RTT's system clock will be driven by the Gazebo clock and SimClockActivity activities will run at rates which observe this simulated time.

Note that if you want to run your components out-of-process and communicate with Gazebo asynchronously over ROS topics, you can still use rtt_rosclock and communicate with Gazebo

For more usage details, see the README.md files in each package in this repository.

Performance (thorough analysis forthcoming)

Obviously, when using the tools in this package to run RTT components synchronously in the Gazebo process, you are able to force the system to enable you to read and write as fast as is desirable. This means that a collection of RTT components can slow the Gazebo simulation to the point where they have enough time to compute each control update. This can be useful in cases where the simulation machine is computationally underpowered compared to the real system.

The expectation is that running RTT components in-process enables lower latencies and higher closed-loop control frequencies. A close analysis of such performance has yet to be performed.