Robotics - [ROS2 Foundation - 1] DDS Notes

What’s Middleware

Middleware abstracts the complexity of managing communication, serialization, discovery, and other low-level functionalities, allowing developers to focus on application-level logic.

• ROS1’s middleware is TCPROS: it’s a custom serialization format, a custom transport protocol as well as a custom central discovery mechanism

• ROS2 uses an existing middleware interface, DDS. There are multiple DDS implementations. ROS has a unified interface for these implementations

  • In DDS, each ROS 2 node is equivalent to a DDS “participant”. There could be multiple ROS nodes in the same process. Each node is still a DDS participant
  • Under the hood, there are DDS datareader, datawriter, and DDS topics. They are not exposed to ROS users

1
2
3
4
5
6
7
8
9
10
11
+-----------------------------------------------+
|                   user land                   |   No middleware implementation specific code
+-----------------------------------------------+
|              ROS client library               |
+-----------------------------------------------+
|             middleware interface              |   No DDS specifics
+-----------------------------------------------+
| DDS adapter 1 | DDS adapter 2 | DDS adapter 3 |
+---------------+---------------+---------------+
|    RTI DDS impl 1 | PrimsTech DDS impl |    DDS impl 3 |
+---------------+---------------+---------------+

Why DDS (Summary)

When trying to develop ROS 2 communication systems, Open Robotics faced two choices: 1. improving on the existing ROS1 infrastructure 2. Use ZeroMQ, Protocol Buffers, and Zeroconf (see below). Both choices need a middleware.

Some Salient Features of DDS:

1. The default implementation of DDS is over UDP, and only requires that level of functionality from the transport.
2. Because it’s based on UDP, Quality of Service (QoS) needs to be introduced (unlike TCP)
   • for soft real-time, you can basically tune DDS to be just a UDP blaster.
3. DDS would completely replace the ROS master based discovery system.
4. DDS has publisher and subscriber. However, they are implemented by DataWriter and DataReader, which hides the implementations of communication
5. Shared Memory Support
   • In ROS 1, nodelets were used. Nodelets allow publishers and subscribers to share data by passing around boost::shared_ptrs to messages

Disadvantages of DDS:

• ROS must work within that existing design. If the design did not target a relevant use case or is not flexible, it might be necessary to work around the design.

[1] DDS Implementations

There are multiple middlewares because you might have considerations such as license, platform availability, or computation footprint:

• Fast DDS(by eProsima) whose wire protocol is Fast RTPS. Package: rmw_fastrtps_cpp
  • ROS2 Humble’s default
  • Check echo $RMW_IMPLEMENTATION.
• Cyclone DDS (Eclipse): slightly lighter weight, but less configurable than Fast RTPS. Can be explicitly installed by: sudo apt install ros-humble-rmw-cyclonedds-cpp. Package: rmw_cyclonedds_cpp
  • ROS2 Galactic’s default
• ConnextDDS (RTI). Package: rmw_connext_cpp

DDS is controlled by:

1
export RMW_IMPLEMENTATION=rmw_fastrtps_cpp

• rmw stands for ROS Middleware. It’ a package that implements the abstract ROS middleware interface using DDS or RTPS’s API and tools.

Some RMW implementations can communicate:

• Fast DDS <-> Cyclone DDS
• Fast DDS <-> Connext

[2] Zeroconf

ROS 2 and many other IoT devices heavily relies on Zeroconf. Zero Configuration Networking (Zeroconf) allow devices on local network to discover and communicate with each other without requiring manual configuation. It does:

• Automated Addressing:
  • Devices assign themselves a unique IP address without requiring a DHCP server. This is usually in the link-local address range (169.254.x.x in IPv4)
  • It tries to use DHCP first. If no DHCP server is found, the device assigns itself a link-local IP address.
• Service Discovery
  • Each device broadcasts and discovers services (printers, file shares, media servers) on the local network
  • Often uses Multicast DNS, DNS Service Discovery (DNS-SD)
• Name resolution
  • Based on the discovered devices, names <-> IP mapping can be broadcasted

Common protocols used inZeroconf :

• Apple Bonjour: A popular Zeroconf implementation used in macOS and iOS.
• Avahi: A Linux-based Zeroconf implementation, commonly used for service discovery.
• DNS-SD (Service Discovery): Advertises and discovers services (e.g., “_http._tcp.local” for a web server).
• Multicast DNS (mDNS): Resolves hostnames to IP addresses in the absence of a DNS server.

Advantages:

• No network configuration needed. It uses dynamic and flexible local networking
• Reduces dependency on network infrastructure like DHCP and DNS servers

Disadvantages:

• Scalability. In larger network, multicast could be noisy
• Security: open broadcasting could be intercepted
• Network scope: limited to a single subnet and does not cross routers

Reference: ROS2 Design Documentation

Backgrounds of DDS

DDS comes out of a set of companies which are decades old, was laid out by the OMG which is an old-school software engineering organization, and is used largely by government and military users.

• They have limited forum support
• They do not have extensive user-contributed wikis or an active Github repository.

[1] What is Fast RTPS

Fast RTPS was designed by Object Management Group (OMG). While DDS defines overall framework and API, Fast RTPS defines the nitty-gritty of the publisher-subscriber model used by DDS:

• Serialization, deserialization
• publishers, subscribers, topics
• QoS policies: reliability, durability, and liveliness
• Automatic Discovery
• Transport Independence uses UDP and shared_memory

How does RTPS work?

First, RTPS has Participants, which represent applications or nodes in the distributed system. They correspond to DDS participants.

Each participant has Writers and Readers:

• Writers: Publish data (similar to DDS DataWriters).
• Readers: Subscribe to data (similar to DDS DataReaders).

Each writer and reader has topics: Logical channels through which writers and readers exchange data.

RTPS Communication Workflow

1. discovery process: RTPS uses Simple Endpoint Discovery Protocol (SEDP) to discover participants, topics, and endpoints (readers/writers).

2. Data Exchange phase:

• Writers publish serialized data to the network.
• Readers listen for data on the corresponding topic and deserialize it upon reception.

1. QoS Enforcement

[2] What is IDL

DDS uses the “Interface Description Language (IDL)” as defined by the Object Management Group (OMG) for message definition and serialization. For example:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
module RobotSystem {
  struct Pose {
    float x;       // Position in X-axis
    float y;       // Position in Y-axis
    float z;       // Position in Z-axis
    float roll;    // Rotation around X-axis
    float pitch;   // Rotation around Y-axis
    float yaw;     // Rotation around Z-axis
  };

  struct RobotState {
    string<255> name;  // Robot name (max 255 characters)
    Pose pose;         // Pose of the robot
    bool is_active;    // Is the robot active?
  };
};

DDS implementations (e.g., Fast DDS, Cyclone DDS, or RTI Connext) provide tools to generate code from the IDL file. The generated code includes:

1
2
Serialization and deserialization functions.
Definitions for the corresponding data types in the target programming language (e.g., C++, Python, etc.).

What is ROS domain ID

ROS_DOMAIN_ID is an environment variable (default 0). It namespaces the DDS discovery ports so you can safely run multiple independent ROS 2 networks on the same physical LAN. ROS 2 uses it to compute DDS ports; pick 0–101 to avoid ephemeral-port conflicts

Messages, Services and Actions

[1] Message Conversions Base Mechanism

1. .msg files still keep its format and in-memory representation because it’s proven to be functional in ROS1.
2. .msg file are converted into .idl files so that they could be used with the DDS transport.
   • Language specific files, conversion functions are generated as well.
     • Conversion functions convert ROS <-> DDS in memory instances
   • ROS2 API only converts a message field-by-field into an idl object. DDS in-memory instances are not generated and consumed except only in .publish() and topic callbacks.
     • This is because serialization is at least 10x longer than field-to-field copy.
     • ROS2 build can do msg -> idl outside publish() and receiver callbacks instead of doing a zero-copy msg conversion.
3. .idl files are published

So during c++ compilation, msg file Pose.msg:

1
2
3
4
5
6
float32 x
float32 y
float32 z
float32 roll
float32 pitch
float32 yaw

will be converted to:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
// Generated C++ code
namespace custom_msgs {
  struct Pose {
    float x;
    float y;
    float z;
    float roll;
    float pitch;
    float yaw;

    // Constructor
    Pose() : x(0.0), y(0.0), z(0.0), roll(0.0), pitch(0.0), yaw(0.0) {}

    // Comparison operators, serialization methods, etc., are also included
  };
}

So you can create:

1
2
3
custom_msgs::Pose pose;
pose.x = 1.0;
...

In Python, similar thing happens during build:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
# Generated Python code
class Pose:
    def __init__(self):
        self.x = 0.0
        self.y = 0.0
        self.z = 0.0
        self.roll = 0.0
        self.pitch = 0.0
        self.yaw = 0.0

# User code
from custom_msgs.msg import Pose

pose = Pose()
pose.x = 1.0

• In ROS2 terms, each C++/Python variable is an in-memory representation
• The in-memory representation includes methods for serialization (converting to a binary format for transmission) and deserialization (reconstructing the object from the binary format).

[2] Service and Actions

A service in ROS 2 is a mechanism for synchronous communication between a client and a server. It allows a node to send a request and receive a response. E.g,

1
2
3
4
5
# AddTwoInts.srv
int64 a
int64 b
---
int64 sum

An action is a mechanism for asynchronous, long-running tasks that provides periodic feedback and the ability to cancel the task. It uses Goal-Feedback-Result model:

1
2
3
4
5
6
# Fibonacci.action
int32 order  # Goal
---
int32[] sequence  # Feedback
---
int32[] sequence  # Result

Both Service and Actions are built on top of the publish-subscribe model.

• Service uses two topics: /service_name_request and /service_name_response
• Action has:
  • Goal service: send the goal to the action server
  • Cancel service: cancel the current goal
  • Feedback and status are communicated using topics:
    • A feedback topic for periodic updates.
    • A status topic to inform about the state of the goal (e.g., “active,” “canceled,” “succeeded”).
  • Result service: retrieve the final result

• Previous
  Robotics General Design Notes
• Next
  Robotics - [ROS2 Foundation 2] ROS2 in Docker