备注

You're reading the documentation for an older, but still supported, version of ROS 2. For information on the latest version, please have a look at Kilted.

Ardent Apalone (ardent)

Welcome to the first non-beta release of ROS 2 software named Ardent Apalone!

Supported Platforms

This version of ROS 2 is supported on three platforms:

• Ubuntu 16.04 (Xenial)

• Mac macOS 10.12 (Sierra)

• Windows 10

Binary packages as well as instructions for how to compile from source are provided for all 3 platforms (see install instructions as well as documentation).

Required support
Architecture	Ubuntu Xenial (16.04)	MacOS Sierra (10.12)	Windows 10 (VS2015)
amd64	X	X	X
arm64	X

Minimum language requirements:

• C11[^2]

• C++14

• Python 3.5

[^2]: C11 is required, but support for some non-compliant systems is

also provided, e.g. MSVC.

Dependency Requirements:

Package	Ubuntu Xenial	MacOS**	Windows 10**
CMake	3.5.1	3.11.0	3.10.2
EmPY	3.3.2	3.6.5	3.3.2
Ogre	1.10*	1.10*	1.10*
OpenCV	2.4.9	3.4.1	2.4.13.2*
Poco	1.7.7*	1.7.7*	1.7.7*
Python	3.5.1	3.6.5	3.6.4
Qt	5.5.1	5.10.0	5.10.0
Linux only (used for turtlebot demo)
PCL	1.7.2	N/A	N/A

" * " means that this is not the upstream version (available on the official Operating System repositories) but a package distributed by OSRF or the community (package built and distributed on custom repositories).

" ** " Rolling distributions will see multiple version changes of these dependencies during their lifetime.

This document only captures the version at the first release of a ROS distribution and will not be updated as the dependencies move forward. These versions are thus a low watermark.

Package manager use for dependencies:

• Ubuntu Xenial: apt

• MacOS: Homebrew, pip

• Windows: Chocolatey, pip

Build System Support:

• ament_cmake

• cmake

• setuptools

Middleware Implementation Support:

• eProsima Fast-RTPS

• RTI Connext

• ADLINK OpenSplice

Features

New features in this ROS 2 release

• Distributed discovery, publish / subscribe, request / response communication

  • Provided by a C API

  • Implemented using different vendors:

    • eProsima's Fast RTPS as well as ADLINK's OpenSplice (from binary and source)

    • RTI's Connext (only from source)

  • Numerous quality of service settings for handling non-ideal networks

  • DDS Security support (with Connext and Fast RTPS)

• C++ and Python 3 client libraries

  • Sharing common code in C to unify the implementation

  • Execution model separated from the nodes, composable nodes

  • Node-specific parameters (only in C++ atm)

  • Life cycle (only in C++ atm)

  • Optionally intra-process communication using the same API (only in C++)

• Message definitions (with bounded arrays and strings as well as default values)

• Command line tools (e.g. ros2 run)

• rviz with a few display types (the Windows version will likely follow in a few weeks)

• File system-based resource index (querying information without recursive crawling)

• Realtime safe code paths for pub / sub (with compatible DDS implementations only)

• Bridge between ROS 1 and ROS 2

• HSR demo see Beta 3

• Turtlebot demo see Beta 2

For a more detailed description please see the Features page.

Changes since Beta 3 release

Improvements since the Beta 3 release:

• rviz

• Different initialization options for message data structures in C++ (see design doc)

• Logging API improvements, now also used in the demos

• Time support in C++ with different clocks

• wait-for-service support in the Python client library

• Draft implementation of REP 149 specifying format 3 of the package manifest files

Known Issues

• Fast RTPS performance with larger data like the image demo

• Using Connext it is currently not allowed for two topics with the same base name but different namespaces to have a different type (see issue).

• Listing of node names (e.g. using ros2 node list) does not work across some rmw implementations.

• On Windows Python launch files might hang when trying to abort using Ctrl-C (see issue). In order to continue using the shell which is blocked by the hanging command you might want to end the hanging Python process using the process monitor.