Challenges in Sensor Calibration for Robotics Applications

Introduction

In order to ensure robust autonomy, the modern robotic perception suite includes a variety of exteroceptive and interoceptive sensors. The data from these sensors are fused together to eliminate the failure conditions of a single sensing modality. However, sensor fusion requires accurate knowledge of the (i) parameters describing the individual sensor models (e.g., camera lens properties), (ii) parameters expressing spatial relations among different sensors (i.e., relative poses of sensor reference frames), and (iii) the time offsets between the different sensor measurement streams. The tasks of estimating these parameters are known as intrinsic, extrinsic, and temporal calibration, respectively. Often, these tasks are referred to as sensor calibration.


Sensor calibration has been extensively studied, which has led to a diverse range of calibration strategies. While these strategies can be divided based on a variety of criteria, the most fundamental set of criteria is whether they require special targets, and whether they can run online.


Due to the unique and vast options for calibration, a discussion of the calibration methods would best be explored in a survey paper. As a result, the aim of this workshop is not to traverse the whole research field but to tackle sensor calibration questions that are often left inconclusive in modern research publications. This workshop will provide a forum to discuss six themes that often go unmentioned: sensor calibration requirements, sensor calibration simulations, sensor calibration performance, long-term sensor calibration, challenging calibration environments, and sensor calibration method selection. Example discussion points could be the following:. How can we assess a long-term stability of calibration and how often is recalibration required? Which environmental conditions cause the system’s sensor parameters to become inaccurate and to what extent? How can one validate sensor calibration results online? What role does simulation play in the development of calibration algorithms and what are the sim2real gaps? How are downstream tasks, such as SLAM, influenced by calibration inaccuracies and what is the acceptable error tolerance? What are the best methods to compare different calibration methods? Finally, how does one choose the most suitable calibration technique for a specific robotic system?

Invited speakers

Teresa Vidal-Calleja

Associate Professor

University of Technology Sydney

Mathias Gehrig

PhD Student

University of Zurich


Paper Submission

  • Maximum of 4 pages IAS template

  • Submit using one of the available templates

  • When submitting your paper please choose an option "Challenges in sensor calibration for robotics applications" under "Other Information and Files" card.

  • Accepted papers will be published on the workshop website.

  • Accepted papers will not be published in the conference proceedings

  • Authors will be asked to present a poster at the conference workshop

  • Posters will be published to the workshop website after the conference.

Important dates

  • Paper submission deadline: May 13 May 20

  • Notification of acceptance: May 16 May 23

  • Poster submission: May 27

  • Workshop date: June 13th

Program timeline (June 13, 2022)

Sponsors

For more information visit the conference sponsoring page.

Organizing committee

Juraj Peršić

University of Zagreb

General Co-Chair

Emmett Wise

University of Toronto

General Co-Chair