tf2_server

An upgraded tf2_ros/buffer_server. Main features are:

• It supports the same action API as tf2_ros/buffer_server node for on-demand transform querying (which is however not suitable for fast queries).
• It allows clients to request only subparts (or “streams”) of the TF tree, which results in lowering the data transfer rates and computational burden of TF clients.
• It provides C++ and Python helper libraries that make the subtree subscription easier. However, use of these libraries is optional and client code can make full use of subscription to a TF subtree just by correct configuration in .launch files.
• It can be used as a TF concentrator which reduces the usual M:N connection nature of the TF topic to a M:1 and 1:N.

The subtree subscription mode is useful as a middle way between standard TransformListener and the action interface of the buffer_server - it transmits only the requested transform(s), but isn’t burdened with the overhead the action interface brings in.

The server publishes the so called “TF streams”, which are internally utilized by the helper libraries, but can also be used by custom user code. The streams can either be requested dynamically by calling service ~/request_transform_stream or by setting an initial set of streams using the streams parameter described in the following section.

Nodes and nodelets

This package provides node tf2_server_node and nodelet tf2_server/nodelet with equal functionality. Try running the nodelet version with your TF-hungry nodes.

Actions

• tf2_msgs/LookupTransform: Depending on value of parameter ~use_node_namespace, this action is either published in namespace of the node, or in namespace tf2_buffer_server.

Parameters

The first 3 parameters are taken from the original TF2 buffer server. To stay compatible, the parameters are not node-private, but should be set “one level” higher, e.g. besides the node and not inside it.

• double buffer_size: Duration of the buffer memory (in seconds). Default is 120.0.
• bool publish_frame_service: If True, the server will enable service ~/tf2_frames which shows debug information about the contents of the buffer. Default is False.

• bool use_node_namespace: If True, the lookup transform action of this server will be published under the namespace of this tf2_server. If False, the action server will be published in namespace tf2_buffer_server. Default is False.

• double ~transforms_update_period: The period at which the server looks for newly added TF frames.
• double ~initial_streams_wait_time: Specifies how long to wait before registering streams from parameter streams (in seconds). This helps in case not all frames are right away available and you do not want to dynamically update the lists of available frames.
• dict ~streams: If nonempty, specifies some TF streams that should be registered right after starting the server. Each stream has a name (its key in this dict). The stream will be published at ~/stream_key and ~/stream_key/static. The structure is the following and is described in more detail in the following section:

streams:
  body:
    parent_frame: 'base_link'
    child_frames: []
    intermediate_frames: True
    publication_period: 0.1
    publisher_queue_size: 11
    allow_transforms_update: True

Services

• tf2_server/RequestTransformStream ~/request_transform_stream: Requests a transform stream satisfying the given parameters.
• tf2_msgs/FrameGraph ~/tf2_frames: Returns a list of all available frames seen by the buffer server.

tf2_server_node publications and subscriptions

• /tf, /tf_static: The original TF topics.
• ~/stream_key, ~/stream_key/static for each stream_key

RequestTransformStream

The subtree subscription API is built around ReuqestTranformStream service type. The subtree listeners are configured by RequestTransformStreamRequest objects. The meaning of the individual fields is as follows:

• string parent_frame: The top-most frame in TF tree you’re interested in.
• string[] child_frames: In case a nonempty list is given, this specifies the child transforms you are interested in. They don’t need to be direct children of parent_frame, but they should be in the same TF tree. If an empty list is given, stream the whole subtree of parent_frame. This mode requires intermediate_frames set to True.
• bool intermediate_frames: If False, stream only direct parent_frame->child_frame transforms (no matter if they were originally direct neighbors). If True, stream all the transforms between parent_frame and all child_frames.

File truncated at 100 lines see the full file

Changelog for package tf2_server

1.1.3 (2024-01-10)

• Fixed errors printed by pluginlib when tf2_server is installed
• Contributors: Martin Pecka

1.1.2 (2022-06-10)

• Try to de-flake test.
• Contributors: Martin Pecka

1.1.1 (2022-06-08)

• Fixed Noetic CI
• Disable Noetic CI
• Contributors: Martin Pecka

1.1.0 (2022-06-08)

• Noetic compatibility
• Improved readme.
• Added CI status badge.
• Fixed a race condition when creating streams (if the timer were created later than onSubscribe() is called.
• Update README.md (cherry picked from commit 3253416eff8f88330411ec8732b4cacd612923a8)
• Add CI via Github Actions (cherry picked from commit a60dcad82e19caebd6481c6c2af96e6cfbe3a57c)
• Contributors: Martin Pecka

1.0.6 (2020-06-09)

• Minor: catkin_lint.
• Fix: prevent deadlocks on stopping timers.
• Fix: Ignore stale frames in subtree search.
• Fix: Fix nodelet namespace.
• Contributors: Martin Pecka

1.0.5 (2020-02-09)

• Added support for streams whose TF tree can be updated during runtime.
• Added support for initial streams.
• Added possibility to specify desired names of the published transform streams.
• Contributors: Martin Pecka

1.0.4 (2020-02-04)

• Added a parameter that can help distinguishing between the original and the upgraded TF2 server.
• Added support for running the server as nodelet.
• Added support for updating the requested subtree during runtime.
• Contributors: Martin Pecka

1.0.3 (2020-01-21)

• Create LICENSE
• Contributors: Martin Pecka

1.0.2 (2020-01-10)

• Added url tags to package.xml
• Contributors: Martin Pecka

1.0.1 (2020-01-10)

• Initial version
• Contributors: Martin Pecka