About

The IMU Odometry Challenge is hosted by CMU AirLab, with organization led by the TartanIMU and Super Odometry contributors.

The challenge is also designed as a technical prelude to the IROS’26 Workshop: Beyond Exteroception, which centers on interoceptive perception for resilient robotics.

Motivation

Learning-based inertial methods have shown strong progress, but robust cross-platform generalization remains a major open problem.

This challenge is designed to provide:

• A common benchmark for cross-platform inertial odometry
• Shared train/val resources and hidden test evaluation
• A transparent evaluation path for comparing new methods

Technical Context

This challenge is aligned with the Tartan IMU direction:

• Large-scale multi-platform IMU pretraining
• Efficient adaptation to unseen domains
• Online adaptation for deployment-time robustness

The benchmark is intended to surface concrete modeling, adaptation, and evaluation questions that can feed into the workshop discussion around robot self-sensing and robustness beyond exteroceptive perception alone.

Contact

For questions, reach out to the CMU AirLab organizers via the challenge registration page.