[robotics-worldwide] GTSAM 2.0

Frank Dellaert dellaert at gmail.com
Tue Oct 9 14:00:08 PDT 2012


Dear All,

We are pleased to announce GTSAM 2, a high-performance, easy-to-use, and modern C++ library for SLAM, SAM, bundle adjustment, and general nonlinar optimization. It comes with a comprehensive object-oriented MATLAB wrapper (pre-compiled for Mac, Windows, and Linux), and Hands-on introduction and tutorial on smoothing-and-mapping and factor graphs.

GTSAM 1.0beta was released during last year's IROS, and since then we have made significant changes to the library, based on your feedback:

Changes in 2.0.0 from 1.0-beta
	• Large runtime speedup and less memory usage
	• Greatly simplified and cleaner API
	• New discrete optimization framework
	• Dogleg nonlinear optimization (in addition to Levenberg-Marquardt and Gauss-Newton algorithms)
	• iSAM2 algorithm for incremental nonlinear estimation
	• Incremental nonlinear Dogleg optimization using ISAM2
	• CMake build system
	• BSD License

GTSAM 2.1.0 Changes from 2.0.0

	• Hands-on introduction and tutorial on smoothing-and-mapping and factor graphs.
	• Comprehensive object-oriented MATLAB wrapper: rapid application-building or prototyping, easy access to internal numerical data, and access to GTSAM via the interactive MATLAB command line.  Syntax and classes match the C++ API.  A set of MATLAB examples is included in the matlab directory.
	• Sparse Preconditioned Conjugate Gradient Optimization for graphs that are too densely-connected for direct solution methods.
	• API refinements and minor performance improvements.
	• Full support (editing and compiling) for Windows / Visual Studio using CMake-generated solution files.

GTSAM 2.2.0, changes from 2.1.0:

	• Orders-of-magnitude speedup in computing single-variable marginals.
	• Reverted to basic method for computing multiple-variable joint marginals due to occasional information duplication problem.
	• Revamp of the GTSAM timing library - code instrumentation and bookkeeping to produce outlines for timing and profiling.

Best !
Frank Dellaert and the members of the BORG lab.
Center for Robotics and Intelligent Machines
Georgia Institute of Technology
http://www.cc.gatech.edu/~dellaert 



More information about the robotics-worldwide mailing list