Data in Field Robotics: From State Estimation to Navigation

About the Workshop

Mobile robots, including legged robots, humanoids, and drones, are increasingly deployed in unstructured and demanding environments—from scaling rubble and inspecting industrial facilities to navigating disaster zones. These open-ended tasks require robustness to terrain, resilience under degraded sensing, and adaptability to diverse task complexity. On the path to maturity in mobile robotics, data and benchmarks have proven to be the core driving factor, as demonstrated in autonomous driving. There, large-scale datasets and task-specific benchmarks have propelled tasks such as localization, mapping and detection to reach remarkable levels of precision. However, for mobile robots in the field, we are still missing this ecosystem. Unlike passenger vehicles, collecting data in the field with these platforms poses significant challenges. The cost of collecting data is high, and these platforms cannot easily carry the full payload of sensors and ground-truth systems. This raises several key open questions: What kind of datasets will have lasting impact and be most useful for robotics? Which sensors are truly essential to capture the task complexity of these platforms? What level of calibration and ground-truth effort is required to advance research in these tasks? Most importantly: how do we design benchmarks that are apt for the unique tasks and failure modes of these robots beyond navigation on suburban streets? This workshop will bring together researchers from robotics, computer vision, and machine learning to define principles for the next generation of datasets and benchmarks. By rethinking dataset and benchmark design in light of the distinct requirements of embodied intelligence, we aim to accelerate progress not only in localization, but also in perception, mapping, and long-term autonomy in the wild.

Speakers

Navigation Challenges in Subarctic Forests: From Data Collection to Evaluation

Prof. Dr. François Pomerleau

Université Laval

Webpage Google Scholar

Learning Navigation from World-Scale Street-View Platforms

Dr. Cherie Ho

Stanford University

Webpage Google Scholar

Online Refinement with Real Data and Realistic Modeling in Simulation—a Virtuous Cycle

Dr. Laura Herlant

Robotics and AI Institute

Webpage Google Scholar

Seeing Through the Noise: Radar Data for Robust State Estimation in Field Robotics

Ayoung Kim

Seoul National University

Webpage Google Scholar

Hiking the Alps

Marco Hutter

ETH Zürich / RAI

Webpage Google Scholar

Rising Star Talk: Frontiers in Robot Data: Examples from 4D Perception

Lukas Rosenberger Schmid

University of Technology Nuremberg

Website Google Scholar

Tentative Program

Thursday, October 1, 2026 · 1:30–5:30 PM

The afternoon program combines invited talks, interactive poster presentations, and hands-on robot demonstrations. Times and talk titles are subject to change.

Time	Session	Details
13:30–13:40	Organizing Committee	Welcome and introduction
13:45–14:05	Marco Hutter · Keynote Talk	Hiking the Alps
14:10–14:30	Cherie Ho · Rising Star Talk	Learning Navigation from World-Scale Street-View Platforms
14:35–14:55	Ayoung Kim · Keynote Talk	Seeing Through the Noise: Radar Data for Robust State Estimation in Field Robotics
15:00–15:20	Lukas Rosenberger Schmid · Rising Star Talk	Frontiers in Robot Data: Examples from 4D Perception
15:20–15:30	Poster Session and Robot Interactions	Interactive research presentations and hands-on demonstrations
15:30–16:00	Coffee Break, Extended Poster Session, and Robot Interactions	Live robotic deployments
16:00–16:20	Laura Herlant · Industry Talk	Online Refinement with Real Data and Realistic Modeling in Simulation—a Virtuous Cycle
16:25–16:45	François Pomerleau · Keynote Talk	Navigation Challenges in Subarctic Forests: From Data Collection to Evaluation
16:50–17:10	COMFORT Challenge Spotlight	Challenge spotlight and selected results
17:15–17:30	Best Poster and Challenge Awards	Awards and closing remarks
17:30	End of Workshop	—

COMFORT Localization Benchmark Challenge

The COMFORT Localization Benchmark evaluates real-time state-estimation and localization methods using the GrandTour Dataset. Explore the GrandTour dataset, quick-start guide, and localization task for data access and technical details.

The benchmark is hosted on Codabench. Participants must create a Codabench account before they can register for the competition and submit results.

The challenge follows a two-phase structure:

Validation Phase (July 13, 2026 at 00:00 UTC–August 3, 2026 at 00:00 UTC): The leaderboard is open, and participants can develop and evaluate their methods using the training data.
Test Phase (August 3, 2026 at 00:00 UTC–October 30, 2026 at 23:59 UTC): Reference trajectories remain hidden. Participants can view their own scores, choose whether to make a submission public, and add one result to the leaderboard. The benchmark remains open after the workshop for continued evaluation.

Final ranking and award eligibility: Only test-phase submissions count toward the final benchmark ranking, which is determined by average Absolute Trajectory Error (ATE); lower is better. To be eligible for an award, a submission must be made from August 3, 2026 at 00:00 UTC through September 21, 2026 at 23:59 Anywhere on Earth (AOE), equivalent to September 22, 2026 at 11:59 UTC.

Method requirements: Post-processing is not allowed, and submitted methods must be applicable in real time.

Award selection: Awards are intended for research participants, including students and researchers in academia or industry. Novel methods will be prioritized. Award recipients must present their results at the workshop.

Awards

1st Place: TBD
2nd Place: TBD
3rd Place: TBD

Further details on awards, submission process, evaluation, and additional constraints will be announced soon.

Call for Papers

We invite contributions related to data, benchmarking, state estimation, perception, and navigation for robots operating in challenging real-world environments. Relevant topics include, but are not limited to:

Topics of Interest

Data for field robotics
State estimation, LIO, VIO
Benchmarking and evaluation of field robotics
Multimodal sensing
Navigation and traversability
Autonomy stacks
Real-world deployment at scale

Submission Types and Review Process

We accept two types of submissions:

Novel work: Submit an anonymous 2–4-page, double-column short paper, with unlimited references and appendices. Papers must follow the IROS 2026 formatting guidelines; we recommend this LaTeX template. Submissions will undergo double-blind peer review, with at least one review per paper. Authors of accepted papers will then prepare a poster for an interactive presentation at the workshop.
Previously published work: Work already published or accepted at IROS or another venue should be submitted as a one-page poster. These submissions will be reviewed only for their relevance to the workshop.

Workshop papers will not appear in the official IROS proceedings and generally do not conflict with dual-submission policies. Submit through the DFR Workshop venue on OpenReview. Additional venue information is available on the OpenReview forum.