mola_input_rosbag1

MOLA DataSource from ROS1 bag files that does not need a ROS1 installation

Links
C++ API

README

mola_input_rosbag1

Distro	Build dev	Builds	Stable release
ROS 2 Humble (u22.04)		amd64 arm64
ROS 2 Jazzy (u24.04)
ROS 2 Kilted (u24.04)
ROS 2 Lyrical (u26.04)
ROS 2 Rolling (u26.04)

A MOLA RawDataSource module that reads ROS 1 .bag files and exposes their contents as MOLA observations, without requiring a ROS 1 installation.

The ROS 1 bag (de)serialization, message definitions, and rosbag_storage reader are vendored inside this package (see the ros1/ and mrpt_ros_bridge/ subdirectories), so the module builds and runs in a pure ROS 2 (or even non-ROS) colcon workspace.

Typical uses:

Feed a legacy ROS 1 dataset into MOLA LiDAR Odometry or any other MOLA consumer.
Turn MOLA into a ROS 1 -> ROS 2 bridge: read a .bag and re-publish its streams as ROS 2 topics + /tf, in real time (see demo below).
Inspect raw sensor streams in the MOLA visualizer.

Supported message types

ROS 1 message type	MOLA / MRPT observation
`sensor_msgs/Imu`	`CObservationIMU`
`sensor_msgs/Image`	`CObservationImage` (mono8, mono16, rgb8, bgr8, rgba8, bgra8, bayer_rggb8, bayer_bggr8, bayer_gbrg8, bayer_grbg8)
`sensor_msgs/CompressedImage`	`CObservationImage` (JPEG, PNG, and any format supported by OpenCV `imdecode`)
`sensor_msgs/PointCloud2`	`CObservationPointCloud` (XYZ / XYZI / XYZIRT) or `CObservationRotatingScan`
`sensor_msgs/LaserScan`	`CObservation2DRangeScan`
`sensor_msgs/NavSatFix`	`CObservationGPS`
`nav_msgs/Odometry`	`CObservationOdometry`
`tf2_msgs/TFMessage` (`/tf`, `/tf_static`)	transform tree (sensor pose lookup)

Topics with no known mapping are ignored (a one-time warning is logged).

Sensor poses and `/tf`

The pose of each sensor in the robot body frame is looked up from the bag’s /tf and /tf_static tree, as the transform base_link_frame_id -> <message frame_id>.

Set base_link_frame_id to the name of your robot body frame. Note that many ROS 1 bags use namespaced frames (e.g. r1/base_link instead of base_link).
If the relevant transform is not yet available when a message is read (e.g. the first few messages before any /tf), that single observation is dropped and a throttled warning lists the currently known tf frames, which is handy for discovering the correct base_link_frame_id.
If your bag has no /tf, override the pose per sensor with fixed_sensor_pose (see the example in ros1bag_just_view.yaml).

Parameters

Param	Required	Default	Description
`rosbag_filename`	yes	-	Path to the input `.bag` file.
`base_link_frame_id`	no	`base_link`	Robot body frame for tf pose lookup.
`time_warp_scale`	no	`1.0`	Playback speed multiplier.
`read_ahead_length`	no	`15`	Number of messages pre-read ahead.
`start_paused`	no	`false`	Start playback paused.
`sensors`	no	auto	Explicit list of topics, types, and pose overrides. If omitted, all topics with a known mapping are exposed automatically using the topic name as `sensorLabel`.

See mola-cli-launchs/ros1bag_just_view.yaml for a fully documented sensors example.

Demos

Two ready-to-use mola-cli launch files are provided under mola-cli-launchs/.

1. Visualize raw streams

ROSBAG1_FILE=/path/to/dataset.bag \
  mola-cli src/mola_input_rosbag1/mola-cli-launchs/ros1bag_just_view.yaml

2. ROS 1 -> ROS 2 bridge

ROSBAG1_FILE=/path/to/dataset.bag \
  mola-cli src/mola_input_rosbag1/mola-cli-launchs/ros1bag_to_ros2.yaml

Then, in another terminal with ROS 2 sourced:

ros2 topic list
ros2 topic echo /your_topic
rviz2

If your bag uses namespaced frames, point the body frame at the right one, e.g.:

MOLA_BASE_LINK_FRAME=r1/base_link \
ROSBAG1_FILE=/path/to/dataset.bag \
  mola-cli src/mola_input_rosbag1/mola-cli-launchs/ros1bag_to_ros2.yaml

CLI tools

`rosbag1-info`

An equivalent to ROS 1’s rosbag info, but built on the vendored ROS 1 bag reader so it runs in a pure ROS 2 (or non-ROS) environment. Once this package is built and the workspace sourced, the tool is on the PATH:

rosbag1-info /path/to/dataset.bag

It prints the bag path and version, recording duration, start/end wall-clock times, file size, total message count, compression, the message types (with their md5sums), and, per topic, the message count, average frequency and type:

path:        /path/to/dataset.bag
version:     2.0
duration:    208.8s (3min 28s)
start:       Oct 18 2022 06:12:24.58 (1666066344.58)
end:         Oct 18 2022 06:15:53.36 (1666066553.36)
size:        1.9 GB
messages:    150843
compression: none
types:       nav_msgs/Odometry [cd5e73d190d741a2f92e81eda573aca7]
             sensor_msgs/Imu [6a62c6daae103f4ff57a132d6f95cec2]
             sensor_msgs/PointCloud2 [1158d486dd51d683ce2f1be655c3c181]
topics:      /imu/data          73279 msgs @   399.6 Hz : sensor_msgs/Imu
             /odom              37195 msgs @   200.0 Hz : nav_msgs/Odometry
             /velodyne_points    2071 msgs @     9.9 Hz : sensor_msgs/PointCloud2

Building

cd ~/ros2_ws
colcon build --packages-select mola_input_rosbag1

The vendored ROS 1 reader needs Boost, BZip2 and lz4 development packages (already present in a standard ROS 2 desktop install).

Credits

The vendored ROS 1 rosbag_storage / roscpp_core sources under ros1/ are taken from Traj-LO (originally Willow Garage, BSD license).
ROS <-> MRPT message conversions come from the mrpt_ros_bridge ros1 branch, vendored as a git submodule.
Bag iteration logic is adapted from MRPT’s rosbag2rawlog.

mola_input_rosbag1

README