Martin Magnusson

Welcome!

I am a researcher working here at the Learning Systems lab of AASS.

My main research is concerned with 3D range sensing, and efficient and versatile 3D surface representations. I have investigated 3D scan registration and mobile-robot applications such as localisation, loop detection, and semantic analysis of 3D scenes.

Teaching

I currently teach in the Probabilistic Robotics course, which is part of the IMRIS program.

Brief CV

I am currently a post-doc researcher working at the Learning Systems lab of AASS. I am mainly involved in the ALL-4-eHAM project together with Volvo Construction Equipment, with the target application being autonomous wheel loaders. In this project, we study methods for mobile-robot perception and navigation in dynamic outdoor environments, as well as methods for efficient autonomous handling of piled materials.

In 2009, I received a tekn. dr. (Ph. D.) degree from Örebro University. Between 2004 and 2009, I have been an industrial graduate student at Örebro University in cooperation with Atlas Copco Rock Drills. The focus of my research has been the 3D normal-distributions transform and its applications for scan registration, surface analysis, and loop detection.

I received my undergraduate education in computer science at Uppsala University, interleaved with periods of work and play. I wrote my Master's thesis on the subject of hierarchical reinforcement learning for balancing a bipedal robot, and received my fil. mag. (M. Sc.) in computer science from Uppsala University in 2004.

Selected publications

• Martin Magnusson, Henrik Andreasson, Andreas Nüchter and Achim J. Lilienthal, Automatic Appearance-Based Loop Detection from 3D Laser Data Using the Normal Distributions Transform. Journal of Field Robotics, 26:11-12, 2009, 892-914. [BibTex] [PDF (0.9MB)]

  Animation of results, showing correctly detected "new places" and "loop detections" [MPEG-4, 6.9MB].

  Affinity matrices: outcome of the algorithm vs ground truth.

  Loop detection result (excerpt).

• Martin Magnusson, Tom Duckett and Achim J. Lilienthal, Scan Registration for Autonomous Mining Vehicles Using 3D-NDT. Journal of Field Robotics, 24:10, 24 Oct 2007, pp. 803-827. [BibTex] [PDF (1.0MB)].

  Underground mine data set used for evaluation.

• Martin Magnusson, Andreas Nüchter, Christopher Lörken, Achim J. Lilienthal and Joachim Hertzberg, Evaluation of 3D Registration Reliability and Speed – A Comparison of ICP and NDT. Proceedings of the IEEE International Conference on Robotics and Automation (ICRA), 2009, pp. 3907-3912. [BibTex] [PDF (0.9MB)] [Digest (0.2MB)]

  Registration results, ICP (left) vs NDT (right). Green: success. Red: failure.

• Benjamin Huhle, Martin Magnusson, Wolfgang Straßer and Achim J. Lilienthal, Registration of Colored 3D Point Clouds with a Kernel-based Extension to the Normal Distributions Transform. Proceedings of the IEEE International Conference on Robotics and Automation (ICRA), 2008, pp. 4025-4030. [BibTex] [PDF (1.0MB)] [Digest (0.5MB)]

  Failed registration, using visual features only.

  Registration result when additionally using Color-NDT.

All publications

Journal articles

• Martin Magnusson, Henrik Andreasson, Andreas Nüchter and Achim J. Lilienthal, Automatic Appearance-Based Loop Detection from 3D Laser Data Using the Normal Distributions Transform. Journal of Field Robotics, 26:11-12, 2009, 892-914. [BibTex] [PDF (0.9MB)].
• Martin Magnusson, Tom Duckett and Achim J. Lilienthal, Scan Registration for Autonomous Mining Vehicles Using 3D-NDT. Journal of Field Robotics, 24:10, 24 Oct 2007, pp. 803-827. [BibTex] [PDF (1.0MB)].

Conference papers

• Martin Magnusson and Håkan Almqvist, Consistent Pile-Shape Quantification for Autonomous Wheel Loaders. Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2011, pp. 4078-4083. [BibTex] [PDF (1.3MB)] [Digest (0.1MB)]
• Martin Magnusson, Henrik Andreasson, Andreas Nüchter, and Achim J. Lilienthal, Appearance-Based Loop Detection from 3D Laser Data Using the Normal Distributions Transform. Proceedings of the IEEE International Conference on Robotics and Automation (ICRA), 2009, pp. 23-28. [BibTex] [PDF (0.3MB)] [Digest (0.1MB)]
• Martin Magnusson, Andreas Nüchter, Christopher Lörken, Achim J. Lilienthal and Joachim Hertzberg, Evaluation of 3D Registration Reliability and Speed – A Comparison of ICP and NDT. Proceedings of the IEEE International Conference on Robotics and Automation (ICRA), 2009, pp. 3907-3912. [BibTex] [PDF (0.9MB)] [Digest (0.2MB)]
• Benjamin Huhle, Martin Magnusson, Wolfgang Straßer and Achim J. Lilienthal, Registration of Colored 3D Point Clouds with a Kernel-based Extension to the Normal Distributions Transform. Proceedings of the IEEE International Conference on Robotics and Automation (ICRA), 2008, pp. 4025-4030. [BibTex] [PDF (1.0MB)] [Digest (0.5MB)]
• Henrik Andreasson, Martin Magnusson and Achim J. Lilienthal, Has Something Changed Here? Autonomous Difference Detection for Security Patrol Robots. Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2007, pp. 3429-3435. [BibTex] [PDF (0.8MB)] [Digest (0.2MB)]
• Martin Magnusson and Tom Duckett, A Comparison of 3D Registration Algorithms for Autonomous Underground Mining Vehicles, Proceedings of the European Conference on Mobile Robotics (ECMR), 2005, pp. 86-91. [BibTex] [PDF]
• Martin Magnusson, Rolf Elsrud, Lars-Erik Skagerlund and Tom Duckett, 3D Modelling for Underground Mining Vehicles, Proceedings of the Conference on Modeling and Simulation for Public Safety (SimSafe), 2005, pp. 19-25. [BibTex] [PDF]

Workshop papers

• Martin Magnusson, Andreas Nüchter, Christopher Lörken, Achim J. Lilienthal and Joachim Hertzberg, 3D Mapping the Kvarntorp Mine — A Field Experiment for Evaluation of 3D Scan Matching Algorithms. Proceedings of IROS Workshop "3D Mapping" (IROS), 2008. [BibTex] [PDF (1.0MB)] [Abstract (0.1MB)] [Presentation (4.9MB)].

Theses

• Martin Magnusson, The Three-Dimensional Normal-Distributions Transform — an Efficient Representation for Registration, Surface Analysis, and Loop Detection. Doctoral Thesis, Örebro University, 2009. [BibTex] [PDF (6.4MB)].
• Martin Magnusson, 3D Scan Matching for Mobile Robots with Application to Mine Mapping. Licentiate Thesis, Örebro University, 2006. [BibTex] [PDF (2.9MB)].
• Martin Magnusson, Hierarchical Reinforcement Learning for Balancing a Humanoid Robot. Master Thesis, Uppsala University, 2004.

Data

This section contains links to 3D scan data used for the experimental results in the publications listed above. See also the Osnabrück Robotic 3D Scan Repository, which contains some of the data sets used here.

• Gravel pile data. These are scans of gravel piles from an asphalt production site, used to evaluate methods for pile segmentation and determining efficient attack poses for autonomous construction vehicles. Also included is a set of simulated pile scans. More pile scans are available on request.

Software

This section contains precompiled releases of some of the software used for the results in my publications. Source code is available on request.

• Registration suite. This is the 3D scan registration test suite used for experimental evaluation of 3D-NDT. Both NDT and ICP registration are included.

Contact

Martin Magnusson
AASS Research Centre, School of Science and Technology, Örebro University,
SE - 70182 Örebro, Sweden
Room: T-1216,
Phone: (+46/0) 19 - 303870
Fax: (+46/0) 19 - 303463
Email (masked to avoid spamming): martin {dot} magnusson {at} oru {dot} se

This web site also features a blank page (see the TPILB Project).