What is robologs-timeseries-utils

A collection of timeseries utility functions to analyze, search and transform robotic sensor data.

Quickstart

Try the Timeseries Similarity Search Jupyter Notebook

---

Context

Our Timeseries Similarty Search notebook demonstrates how to perform a multi-signal timeseries similarity search in a rosbag.

Goal

Find the most similar looking subsequences for a query signal in a given IMU signal.

Steps

• Extract IMU and image data from a rosbag using robologs-ros-utils
• Plot IMU angular velocities (angular_velocity.x, angular_velocity.y, angular_velocity.z)
• Select a region of interest in the IMU plot as a query signal, show corresponding video sequence
• Search for the n-most similar subsequences in the IMU plot
• Visualize the n-most similar subsequences against the query signal
• Show video of n-most similar subsequences

Data

The drone racing dataset used in this notebook was recorded by the Robotics and Perception Group (RPG) at the University at Zurich and can be found here. It is released under the Creative Commons license (CC BY-NC-SA 3.0), which is free for non-commercial use (including research).

Example

IMU plot with query signal (black) and 5-most similar subsequences.

Query Sequence

Most Similar Subsequence

Install Jupyter

pip3 install jupyter
jupyter notebook

Run Timeseries Similarity Search Notebook

cd ~/Code/robologs-timeseries-utils/
jupyter notebook notebooks/TimeseriesSimilaritySearch.ipynb