baglab

ROS2 rosbag data analysis library. Load, explore, and visualize rosbag data with MATLAB-like simplicity using pandas DataFrames.

Features

• Lazy loading — open a bag and read only the topics you need
• Dot-access — navigate message fields like msg.twist.linear.x
• Geometry — quaternion/euler conversion, distance, pose interpolation
• Signal processing — lowpass filter, FFT, differentiation, integration
• Diagnostics — topic rate, message gaps, latency chain analysis
• Plotting — time-series, trajectory, step response, error band

Installation

Prerequisites

• Python 3.10+
• ROS2 (Humble / Jazzy)

baglab (Python only)

pip install baglab

For development:

git clone https://github.com/Kotakku/baglab.git
cd baglab
pip install -e ".[dev]"

MCAP acceleration backend (optional)

Installing baglab_mcap_backend enables high-speed MCAP reading without ROS 2 dependency.

cd baglab
sudo apt install liblz4-dev libzstd-dev
pip install --no-build-isolation ./baglab_mcap_backend

Check whether the backend is available:

import baglab
print(baglab.has_mcap_backend())  # True / False

baglab works without the backend — it falls back to rosbags (pure Python).

Usage

import baglab

# Load a rosbag (lazy loading)
bag = baglab.load("path/to/rosbag")

# List topics
print(bag.topics)
# {"/motor/angle": "sensor_msgs/msg/JointState", ...}

# Access topic data (loaded on first access)
twist_df = bag["/cmd_vel"]

# Dot-access for message fields
twist_df.msg.twist.linear.x         # Series
twist_df.msg.twist.linear.df        # DataFrame (columns: [x, y, z])

# Timestamps
t = baglab.stamp_to_sec(twist_df)                  # absolute [s]
t = baglab.stamp_to_sec(twist_df, relative=True)   # relative (starts at 0)

# Reindex by publication timestamp
twist_df = baglab.reindex_by_stamp(twist_df)

Plot example

import baglab
import matplotlib.pyplot as plt

bag = baglab.load("path/to/rosbag")
twist_df = bag["/cmd_vel"]

t = baglab.stamp_to_sec(twist_df, relative=True)
vel = twist_df.msg.twist.linear.df

fig, axes = plt.subplots(3, 1, sharex=True)
for ax, axis_name in zip(axes, ["x", "y", "z"]):
    ax.plot(t, vel[axis_name])
    ax.set_ylabel(axis_name)
axes[-1].set_xlabel("time [s]")
plt.show()

Documentation

https://kotakku.github.io/baglab/

License

Apache-2.0