jlblanco's blog

I'm glad to announce that MRPT has just finished moving from our "home-made" Linear Algebra classes to the great Eigen library. The changes are now only in SVN HEAD, and the feature will appear in the next stable release 0.9.3 (see changelog).

I'm glad to announce the release of MRPT 0.9.2, which comes with dozens of new features and bugfixes (see the CHANGELOG).

The most important changes are:
- Basic support for Graph-SLAM. Among others, see mrpt::poses::CNetworkOfPoses, the new namespace mrpt::graphslam and the program graph-slam
- First working version of Sparse Bundle-Adjustment (see mrpt::vision::bundle_adj_full).
- New hardware supported: Phidget kits & Fiber Optic Gyro KVH DSP3000 (both contributed by Adrien Barral, Robopec).

The MRPT book has been updated with two new chapters:

• Chapter 5: Description of existing MRPT libraries.
• Chapter 11: Smart pointers

The chapter on smart pointers will be specially interesting to those beginning with MRPT, since there are a number of subtle concepts that should be fully understood for an efficient usage of those objects.

From today on, Ubuntu users won't have to keep an up-to-date SVN local copy and rebuild MRPT from sources to enjoy the latest version of the libraries.

As explained here, an Ubuntu PPA repository has been created so installing the latest MRPT and all its dependencies can be done with just three commands.

There is a bug in OpenCV 2.0 packaged in Ubuntu Lucid (10.04), since the function cv::FAST(), declared in the headers, is not actually exported in any library, so that's the reason for the following error building MRPT against that version of OpenCV:

../../lib/libmrpt-vision.so.0.9.1: undefined reference to `cv::FAST(cv::Mat const&, std::vector<:keypoint std::allocator=""> >&, int, bool)'
collect2: ld returned 1 exit status

Solutions:
1. Remove opencv packages, download opencv sources (2.1 or above) and build them.

Just a quick update:
A new technical report has been published in [1] with tons of facts on SE(3) geometry (6D poses), including the relation with existing MRPT functions.

Enjoy!

[1] http://www.mrpt.org/6D_poses%3Aequivalences_compositions_and_uncertainty

FYI, the web section with the repository of datasets has just been finished, as part of the (still!) on-going port from our former website.

Hope you enjoy it!

Hi all,

The latest stable release of MRPT was published yesterday, and is now available for download. This new version comes with dozens of changes and improvements.

The most important ones:

• Most important changes:
  • CRITICAL REGRESION FIXED: The program rbpf-slam did always crash at start-up in MRPT 0.9.0.
  • Changes in libraries:

Up to now, the Subversion repository used to develop MRPT was sited and maintained at the Málaga University, but to assure a greater reliability in the future, we just finished moving it to the servers of Google Code.

Q: I downloaded MRPT as zip or tar.gz files, I try to be away of weird stuff such as that SVN thing...

A: That's your case, then can stop reading this! You're not affected at all.

A new algorithm, "champion detector" (see attached picture! ;-) ) will be available in the forthcoming MRPT 0.9.1.

Joking aside, Raúl Raul Ruiz Sarmiento, a developer here at MAPIR has been working in a hierarchy of C++ classes for very generic detectors/detectable objects for computer vision, which may be extended in the future for more sensors (e.g. 3D cameras).

For more information, see the (under development!!) class hierarchies: