{# This is crucial for text visibility #}

Resilient SLAM in Challenging Settings

A Full-Day Workshop held at ECMR 2025

About the Workshop

Resilient SLAM in Challenging Settings

Recent Advancements, Open Problems, and Competitions

Despite decades of research, deploying SLAM systems in the real world remains a major hurdle. This full-day workshop, held in conjunction with ECMR 2025 conefernce (Padua, Italy). The workshop is aimed to unite researchers across Europe to tackle the challenges of robust localization and mapping head-on.

Visit ECMR Conference Site

SLAM Competition

A competition to benchmark new methods on unique datasets from quadruped and wheeled robots.

Published Papers

The workshop provided an opportunity to discuss late-breaking results and novel ideas about resilient SLAM.

Invited Talks

Sessions with leading experts from academia and industry shared their latest insights.

Interactive Sessions

Engage directly with peers during poster sessions and a World Café-style discussion.

Workshop Timeline

All deadlines have now passed. Thank you to all our contributors!

Paper submission deadline: August 20, 2025
Notification of acceptance: August 28, 2025
Competition deadline: August 28, 2025
Workshop Day: September 2, 2025

Benchmark

Robots in a challenging outdoor environment

Open SLAM Benchmark Platform

An ongoing resource for the robotics community.

Go to Benchmark Platform

The ECMR 2025 workshop competition has concluded. We now proudly present the evaluation infrastructure as an open and ongoing benchmark platform for the SLAM community.

A Novel Evaluation Approach

This platform uses a unique server-side evaluation. You submit your entire Dockerized ROS1 SLAM system, and we run it on our servers against private test data.

An Open Platform

This platform is a permanent resource, open globally for anyone to benchmark their SLAM systems against our challenging datasets at any time.

Evaluation Metrics

Submissions are evaluated using standard metrics, including Absolute Trajectory Error (ATE) and Relative Pose Error (RPE).

Host Your Own Benchmark Challenge!

We invite research groups to expand this platform. You can create your own challenge by providing a new, challenging dataset. We will help you integrate it, and you can then evaluate submitted solutions on your own hardware. This is an excellent opportunity to promote your research and establish a community standard.

Call for Collaboration: ROS2 Upgrade

Our current evaluation system is built on ROS1. We are actively looking for collaborators to help upgrade the platform to ROS2 to support the latest advancements in the robotics community. If you are experienced with ROS2 and interested in contributing, please get in touch!

Collaborate With Us

Integrated Datasets for Auto-Evaluation

Outdoor Degenerate LiDAR

Open fields with a 3D LiDAR. Tunnel and alpine environments are upcoming.

integrated View on GitHub

Viking Hill Radar

Challenging navigation in tall grass using a radar sensor.

integrated View on GitHub

Indoor Quadruped

An indoor dataset from a quadruped robot with RGB-D and motion capture ground truth.

integrating View on GitHub

Community-Provided Datasets

The following datasets were provided by our workshop participants. They are not integrated into the auto-evaluation platform.

Odyssey

Geo-referenced LiDAR-Inertial data set

community View Dataset

Natural and Semi-Urban Dataset

Weekly Multi-Temporal Dataset for Short-Term Localization and Environmental Change Analysis.

community View Dataset

Program

Full-day schedule with keynotes, technical sessions, and competitions: September 2, 2025

View the detailed schedule here.

Presentations & Materials

Thank you for attending! All workshop materials, including accepted papers and speaker presentations, are now available in our shared folder. Recordings are in the process of uploading, and will appear within several weeks after the workshop.

Access All Materials

09:00–09:10Welcome
09:10–10:15 Invited Speakers 1
- 09:10-09:40
  Unal Artan online"An Unsupervised Approach to Map Quality Assessment"
- 09:40-10:15
  Luca Di Giammarino online"Active Perception and Mapping: Learning to Localize in Space and Time"
10:30–11:00Coffee Break
11:00–11:40 Invited Speaker 2
- 11:00-11:40
  Martin Saska"Long-term UAV deployment in challenging GNSS-denied conditions"
11:40–12:40 Accepted Papers
- 11:40-12:00
  Sai Puneeth Gottam online “Good Deep Feature to Track: Self-Supervised Feature Extraction and Tracking in Visual Odometry”
- 12:00-12:20
  Aaron Kurda “Reducing Drift of Lidar Odometry by Incorporating OpenStreetMap Building Data”
- 12:20-12:40
  Abdel-Raouf Dannaoui “When and Where Localization Fails: An Analysis of the Iterative Closest Point in Evolving Environment”
13:00–14:00Lunch Break
14:00–15:30 Invited Speakers 3
- 14:00-14:30
  Matteo Frosi "Beyond Vision: Radar-based SLAM for Robust Localization and Mapping"
- 14:30-15:00
  Mihir Dharmadhikari online "Resilient Multi-modal Perception"
- 15:00-15:30
  Johann Laconte "Towards Robust and Resilient SLAM for Field Robotics"
15:30–16:00Coffee Break & Poster Session
16:00-16:20 Accepted Paper: Dyuman Aditya
“Robust Localization, Mapping, and Navigation for Quadruped Robots”
16:20–16:30 Competition Presentation
Go to Section

We will present our system for benchmarking and auto-evaluation of the dockerized SLAM systems on challenging private datasets.
16:30–17:05 World Cafe Discussion
The World Café is a discussion format where attendees explore different facets of a topic in breakout groups, using the day's talks and posters as stimuli. The goal is to connect participants from various backgrounds to discuss relevant community challenges on even ground. The format consists of 5-minute discussion rounds on each topic, followed by a final conclusion session.
17:05–17:15Closing Remarks

Invited speakers and Materials

Martin Saska Czech Technical University in Prague Topic: "Long-term UAV deployment in challenging GNSS-denied conditions"

Matteo Frosi Politecnico di Milano Topic: "Beyond Vision: Radar-based SLAM for Robust Localization and Mapping"

Johann Laconte French National Institute for Agriculture, Food, and Environment (INRAE) Topic: "Towards Robust and Resilient SLAM for Field Robotics"

Luca Di Giammarino online Sapienza University of Rome Topic: "Active Perception and Mapping: Learning to Localize in Space and Time"

Mihir Dharmadhikari online Norwegian University of Science and Technology (NTNU) Topic: "Resilient Multi-modal Perception"

Post-Workshop Q&A

How do you handle multi-sensor calibration? Is it part of your state space or more of an external calibration?

A: We do not estimate the sensor extrinsics as part of the state space. We perform sensor extrinsic calibrations based on the modality. We use Kalibr for camera-IMU calibration. Similarly, we have explored various LiDAR-IMU and radar-IMU calibration techniques. As our sensors have rigid transformations between them and are placed close by, a transformation from CAD often can work fairly well, if nothing else as an initial guess.
What about the time synchronization in the factor graph optimization? Do you consider discrete measurements or something else?

A: We indeed consider discrete measurements for our SLAM methods. Regarding time synchronization, we have developed a synchronization hardware and software system using a Teensy microcontroller that synchronizes the camera, LiDAR, radar, and IMU. Briefly speaking, the microcontroller provides triggering signals to the IMU, radar, and camera for synchronized sampling. Furthermore, the same microcontroller passes an NMEA message once per second to the LiDAR via UART, synchronizing the LiDAR timing the same way a GPS would. In the factor graph, the LiDAR point cloud is deskewed to the latest radar measurement timestamp. Through this synchronization system we achieve accurate time stamping of the measurements that are added to the factor graph.More details about the time synchronization system can be found here: https://arxiv.org/pdf/2507.05717

Ask a Question

Unal Artan online Örebro University Topic: "An Unsupervised Approach to Map Quality Assessment"

Archived Call for Papers

The workshop invited contributions on topics related to resilient SLAM in challenging real-world conditions. The topics of interest included:

Organizers

Vsevolod Hulchuk

CTU in Prague

Contact

Matteo Luperto

University of Milano, Italy

Dyuman Aditya

Ecole Centrale de Nantes, France

Vladimír Kubelka

Örebro University, Sweden

Rudolf Szadkowski

CTU in Prague

Martin Magnusson

Örebro University, Sweden

Miloš Prágr

CTU in Prague

Konstantinos Alexis

NTNU

Tomasz Kucner

Aalto University, Finland

Piotr Kicki

Poznan University of Technology, Poland

Venue and Location

Workshop Location

Palazzo della Salute
Via S. Francesco, 90
I-35121 Padua (Italy)

Open in Google Maps

Room: ELETTRA 1 BLUE

Information on how to get to the room will be provided at the main conference registration desk.

Resilient SLAM in Challenging Settings

About the Workshop

Resilient SLAM in Challenging Settings

SLAM Competition

Published Papers

Invited Talks

Interactive Sessions

Workshop Timeline

Benchmark

Open SLAM Benchmark Platform

A Novel Evaluation Approach

An Open Platform

Evaluation Metrics

Host Your Own Benchmark Challenge!

Call for Collaboration: ROS2 Upgrade

Integrated Datasets for Auto-Evaluation

Outdoor Degenerate LiDAR

Viking Hill Radar

Indoor Quadruped

Community-Provided Datasets

Odyssey

Natural and Semi-Urban Dataset

Program

Presentations & Materials

Invited speakers and Materials

Post-Workshop Q&A

Archived Call for Papers

Submission Topics

Organizers

Venue and Location

Workshop Location